CN110146100A - Trajectory prediction method, device and storage medium - Google Patents

Abstract

Translated fromChinese

本公开提供了一种轨迹预测方法、装置及存储介质，属于车辆控制技术领域。方法包括：获取目标车辆在当前时刻的多个第一驾驶行为特征值，目标车辆为本车周围的车辆，多个第一驾驶行为特征值与多个驾驶行为特征一一对应，多个驾驶行为特征为根据多个车辆的历史运动信息得到的与驾驶行为相关的特征，所述驾驶行为包括换道或者直行；根据所述多个第一驾驶行为特征值，识别所述目标车辆在所述当前时刻是否有换道意图；当识别到所述目标车辆在所述当前时刻有换道意图时，获取所述目标车辆与所述本车之间的相对距离；根据所述相对距离，通过预设的轨迹预测方程，预测所述目标车辆的运动轨迹。本公开能够准确识别目标车辆的换道意图，并准确预测运动轨迹。

The disclosure provides a trajectory prediction method, device and storage medium, which belong to the technical field of vehicle control. The method includes: acquiring a plurality of first driving behavior characteristic values of the target vehicle at the current moment, the target vehicle is the vehicles around the own vehicle, the plurality of first driving behavior characteristic values correspond to the plurality of driving behavior characteristics one by one, and the plurality of driving behavior characteristics The feature is the feature related to the driving behavior obtained according to the historical motion information of a plurality of vehicles, the driving behavior includes changing lanes or going straight; Whether there is an intention to change lanes at any time; when it is recognized that the target vehicle has an intention to change lanes at the current moment, obtain the relative distance between the target vehicle and the vehicle; according to the relative distance, by preset The trajectory prediction equation predicts the trajectory of the target vehicle. The present disclosure can accurately identify the lane-changing intention of the target vehicle and accurately predict the trajectory.

Description

Translated fromChinese

轨迹预测方法、装置及存储介质Trajectory prediction method, device and storage medium

技术领域technical field

本公开涉及车辆控制技术领域，特别涉及一种轨迹预测方法、装置及存储介质。The present disclosure relates to the technical field of vehicle control, and in particular to a trajectory prediction method, device and storage medium.

背景技术Background technique

目前，无人驾驶汽车的一大应用场景是混合交通流道路，即无人驾驶汽车和驾驶人驾驶汽车并存的道路上。在该场景中，无人驾驶汽车会受到来自驾驶人驾驶汽车的影响。因此，为了保证用户安全，无人驾驶汽车需要获取周围驾驶人驾驶汽车的轨迹，基于这个轨迹对无人驾驶汽车进行控制，从而保证无人驾驶汽车能够安全高效的抵达目的地。At present, a major application scenario of driverless cars is mixed traffic flow roads, that is, roads where driverless cars and driver-driven cars coexist. In this scenario, a self-driving car is influenced by a human driving a car. Therefore, in order to ensure the safety of users, unmanned vehicles need to obtain the trajectories of surrounding drivers, and control the unmanned vehicles based on this trajectory, so as to ensure that the unmanned vehicles can reach their destinations safely and efficiently.

发明内容Contents of the invention

本公开实施例提供了一种轨迹预测方法、装置及存储介质，可以准确预测车辆的运动轨迹。所述技术方案如下：Embodiments of the present disclosure provide a trajectory prediction method, device and storage medium, which can accurately predict the trajectory of a vehicle. Described technical scheme is as follows:

第一方面，提供了一种轨迹预测方法，所述方法包括：In a first aspect, a trajectory prediction method is provided, the method comprising:

获取目标车辆在当前时刻的多个第一驾驶行为特征值，所述目标车辆为本车周围的车辆，所述多个第一驾驶行为特征值与多个驾驶行为特征一一对应，所述多个驾驶行为特征为根据多个车辆的历史运动信息得到的与驾驶行为相关的特征，所述驾驶行为包括换道或者直行；Obtain a plurality of first driving behavior characteristic values of the target vehicle at the current moment, the target vehicle is a vehicle around the vehicle, the plurality of first driving behavior characteristic values are in one-to-one correspondence with a plurality of driving behavior characteristics, and the plurality of A driving behavior feature is a feature related to driving behavior obtained according to historical motion information of multiple vehicles, and the driving behavior includes changing lanes or going straight;

根据所述多个第一驾驶行为特征值，识别所述目标车辆在所述当前时刻是否有换道意图；Identifying whether the target vehicle has a lane-changing intention at the current moment according to the plurality of first driving behavior characteristic values;

当识别到所述目标车辆在所述当前时刻有换道意图时，获取所述目标车辆与所述本车之间的相对距离；When it is recognized that the target vehicle has an intention to change lanes at the current moment, acquiring the relative distance between the target vehicle and the own vehicle;

根据所述相对距离，通过预设的轨迹预测方程，预测所述目标车辆的运动轨迹。According to the relative distance, the trajectory of the target vehicle is predicted through a preset trajectory prediction equation.

在本公开实施例中，获取目标车辆在当前时刻的多个第一驾驶行为特征值，根据多个第一驾驶行为特征值，识别目标车辆在当前时刻是否有换道意图。由于多个驾驶行为特征为根据多个车辆的历史运动信息得到的与驾驶行为相关的特征。因此，提高了识别的准确性。并且，当识别出目标车辆在当前时刻有换道意图时，能够根据两车之间的相对距离，准确预测目标车辆的运动轨迹。In an embodiment of the present disclosure, a plurality of first driving behavior characteristic values of the target vehicle at the current moment are acquired, and according to the plurality of first driving behavior characteristic values, it is identified whether the target vehicle has an intention to change lanes at the current moment. Since the multiple driving behavior features are features related to driving behavior obtained according to historical motion information of multiple vehicles. Therefore, the accuracy of recognition is improved. Moreover, when it is recognized that the target vehicle intends to change lanes at the current moment, the trajectory of the target vehicle can be accurately predicted according to the relative distance between the two vehicles.

在一个可能的实现方式中，所述根据所述多个第一驾驶行为特征值，识别所述目标车辆在所述当前时刻是否有换道意图，包括：In a possible implementation manner, the identifying whether the target vehicle has an intention to change lanes at the current moment according to the plurality of first driving behavior characteristic values includes:

根据所述多个第一驾驶行为特征值，获取所述目标车辆在所述当前时刻的观测矩阵，所述观测矩阵用于指示所述目标车辆在不同驾驶行为的先验概率；Obtaining an observation matrix of the target vehicle at the current moment according to the plurality of first driving behavior eigenvalues, the observation matrix being used to indicate prior probabilities of the target vehicle in different driving behaviors;

获取所述目标车辆在所述当前时刻之前的上一时刻的第一概率转移矩阵，所述第一概率转移矩阵用于指示所述目标车辆在不同驾驶行为之间的转移概率；Acquiring a first probability transition matrix of the target vehicle at a previous moment before the current moment, where the first probability transition matrix is used to indicate transition probabilities of the target vehicle between different driving behaviors;

根据所述第一概率转移矩阵和所述观测矩阵，确定所述目标车辆在所述当前时刻的第二概率转移矩阵；determining a second probability transition matrix of the target vehicle at the current moment according to the first probability transition matrix and the observation matrix;

根据所述第二概率转移矩阵，确定所述目标车辆在所述当前时刻的换道概率，所述换道概率用于指示所述目标车辆进行换道的概率；determining the lane-changing probability of the target vehicle at the current moment according to the second probability transition matrix, where the lane-changing probability is used to indicate the probability of the target vehicle changing lanes;

当所述换道概率大于预设概率时，识别出所述目标车辆在当前时刻有换道意图。When the lane-changing probability is greater than the preset probability, it is recognized that the target vehicle has a lane-changing intention at the current moment.

在本公开实施例中，根据上一时刻的第一概率转移矩阵和当前时刻的观测矩阵，预测当前时刻的概率转移矩阵，根据当前时刻的概率转移矩阵，识别目标车辆是否有换道意图。由于通过基于上一时刻的第一概率转移矩阵，而上一时刻的第一概率转移矩阵也是通过上上时刻的概率转移矩阵迭代得到的，因此，本公开实施例中，是通过迭代多个历史时刻的概率转移得到当前时刻的第二概率转移矩阵，从而能够实现根据目标车辆的历史驾驶行为，识别出目标车辆的换道意图，进一步提高了意图识别的准确性。In the embodiment of the present disclosure, the probability transition matrix at the current moment is predicted according to the first probability transition matrix at the previous moment and the observation matrix at the current moment, and whether the target vehicle intends to change lanes is identified according to the probability transition matrix at the current moment. Since the first probability transition matrix based on the previous moment is obtained by iterating the probability transition matrix at the previous moment, the first probability transition matrix at the previous moment is obtained by iterating multiple historical The second probability transition matrix at the current moment is obtained by the probability transfer at the moment, so that the lane-changing intention of the target vehicle can be identified according to the historical driving behavior of the target vehicle, and the accuracy of intention recognition is further improved.

在一个可能的实现方式中，所述根据所述多个第一驾驶行为特征值，获取所述目标车辆在所述当前时刻的观测矩阵，包括：In a possible implementation manner, the acquiring the observation matrix of the target vehicle at the current moment according to the plurality of first driving behavior characteristic values includes:

根据所述多个第一驾驶行为特征值，确定所述目标车辆的驾驶行为特征向量；determining a driving behavior feature vector of the target vehicle according to the plurality of first driving behavior feature values;

获取所述目标车辆在当前时刻之前的预设时长内的平均驾驶行为特征向量，所述平均驾驶行为特征向量包括多个平均驾驶行为特征值，所述多个平均驾驶行为特征值与多个驾驶行为特征一一对应；Obtaining the average driving behavior feature vector of the target vehicle within a preset time period before the current moment, the average driving behavior feature vector including a plurality of average driving behavior feature values, the plurality of average driving behavior feature values and a plurality of driving behavior One-to-one correspondence of behavioral characteristics;

根据所述驾驶行为特征向量和所述平均驾驶行为特征向量，通过预设的高斯判别公式确定所述目标车辆在所述当前时刻的观测矩阵。According to the driving behavior feature vector and the average driving behavior feature vector, the observation matrix of the target vehicle at the current moment is determined through a preset Gaussian discriminant formula.

在本公开实施例中，通过迭代目标车辆在当前时刻之前的预设时长内的平均驾驶行为特征向量，得到当前时刻的观测矩阵。从而能够实现根据目标车辆的历史驾驶行为，识别出目标车辆的换道意图，进一步提高了意图识别的准确性。In the embodiment of the present disclosure, the observation matrix at the current moment is obtained by iterating the average driving behavior feature vector of the target vehicle within a preset time period before the current moment. Therefore, the lane-changing intention of the target vehicle can be recognized according to the historical driving behavior of the target vehicle, and the accuracy of intention recognition is further improved.

在一个可能的实现方式中，所述根据所述相对距离，通过预设的轨迹预测方程，预测所述目标车辆的运动轨迹，包括：In a possible implementation manner, predicting the trajectory of the target vehicle according to the relative distance through a preset trajectory prediction equation includes:

获取预设的轨迹预测方程中的横向偏移值和纵向偏移值；Obtain the horizontal offset value and vertical offset value in the preset trajectory prediction equation;

获取所述预设的轨迹预测方程中的第一形状参数值和第二形状参数值；Acquiring the first shape parameter value and the second shape parameter value in the preset trajectory prediction equation;

根据所述横向偏移值、所述纵向偏移值、所述第一形状参数值、第二形状参数值和所述相对距离，通过所述预设的轨迹预测方程，预测所述目标车辆的运动轨迹。According to the lateral offset value, the longitudinal offset value, the first shape parameter value, the second shape parameter value and the relative distance, through the preset trajectory prediction equation, predict the trajectory of the target vehicle motion track.

在本公开实施例中，结合横向偏移值和纵向偏移值，以及第一形状参数值和第二形状参数值，预测目标车辆的运动轨迹。由于横向偏移值和纵向偏移值分别用于指示轨迹预测方程中预测出的运动轨迹与目标车辆的实际运动轨迹之间的横向偏差和纵向偏差，第一形状参数值和第二形状参数值分别指示目标车辆的换道风格。由此可见，本公开实施例中，考虑到了横向偏移值和纵向偏移值以及换道风格，因此预测出的运动轨迹更接近目标车辆的实际运动轨迹，从而提高了预测出的运动轨迹的准确性。In the embodiment of the present disclosure, the trajectory of the target vehicle is predicted by combining the lateral offset value and the longitudinal offset value, as well as the first shape parameter value and the second shape parameter value. Since the lateral offset value and the longitudinal offset value are respectively used to indicate the lateral deviation and longitudinal deviation between the trajectory predicted in the trajectory prediction equation and the actual trajectory of the target vehicle, the first shape parameter value and the second shape parameter value respectively indicate the lane-changing style of the target vehicle. It can be seen that in the embodiments of the present disclosure, the lateral offset value, the longitudinal offset value and the lane change style are considered, so the predicted trajectory is closer to the actual trajectory of the target vehicle, thereby improving the accuracy of the predicted trajectory. accuracy.

在一个可能的实现方式中，所述获取所述预设的轨迹预测方程中的第一形状参数值和第二形状参数值，包括：In a possible implementation manner, the obtaining the first shape parameter value and the second shape parameter value in the preset trajectory prediction equation includes:

获取第一形状参数的第一概率密度分布和第二形状参数的第二概率密度分布，所述第一概率密度分布包括所述第一形状参数的参数值和概率密度分布的对应关系，所述第二概率密度分布包括所述第二形状参数的参数值和概率密度分布的对应关系；Acquiring a first probability density distribution of the first shape parameter and a second probability density distribution of the second shape parameter, the first probability density distribution includes a correspondence between the parameter value of the first shape parameter and the probability density distribution, the The second probability density distribution includes a corresponding relationship between the parameter value of the second shape parameter and the probability density distribution;

从所述第一概率密度分布中选择最大概率密度值对应的所述第一形状参数值，以及从所述第二概率密度分布中选择最大概率密度值对应的所述第二形状参数值。The first shape parameter value corresponding to the maximum probability density value is selected from the first probability density distribution, and the second shape parameter value corresponding to the maximum probability density value is selected from the second probability density distribution.

在本公开实施例中，分布根据第一形状参数值对应的第一概率密度分布和第二形状参数值对应的概率密度分布，确定出概率密度分布值最大的第一形状参数值和第二形状参数值，从而提高了确定出的第一形状参数值和第二形状参数值的准确性，进而进一步提高了轨迹预测的准确性。In the embodiment of the present disclosure, the distribution determines the first shape parameter value and the second shape with the largest probability density distribution value according to the first probability density distribution corresponding to the first shape parameter value and the probability density distribution corresponding to the second shape parameter value parameter values, thereby improving the accuracy of the determined first shape parameter value and the second shape parameter value, thereby further improving the accuracy of trajectory prediction.

在一个可能的实现方式中，所述获取第一形状参数的第一概率密度分布和第二形状参数的第二概率密度分布，包括：In a possible implementation, the acquiring the first probability density distribution of the first shape parameter and the second probability density distribution of the second shape parameter includes:

根据所述目标车辆在当前时刻的位置信息，获取与所述位置信息对应的所述第一概率密度分布和所述第二概率密度分布；或者，Acquiring the first probability density distribution and the second probability density distribution corresponding to the position information according to the position information of the target vehicle at the current moment; or,

根据所述目标车辆在当前时刻的环境信息，获取与所述环境信息对应的所述第一概率密度分布和所述第二概率密度分布。According to the environment information of the target vehicle at the current moment, the first probability density distribution and the second probability density distribution corresponding to the environment information are acquired.

在本公开实施例中，不同位置或者环境信息对应不同的换道风格。因此，获取与位置信息或者环境信息对应的第一概率密度分布和第二概率密度分布，进而提高了获取的第一概率密度分布和第二概率密度分布的准确性，进而进一步提高了后续轨迹预测的准确性。In the embodiments of the present disclosure, different locations or environmental information correspond to different lane changing styles. Therefore, the first probability density distribution and the second probability density distribution corresponding to the location information or the environment information are obtained, thereby improving the accuracy of the obtained first probability density distribution and the second probability density distribution, and further improving the follow-up trajectory prediction accuracy.

在一个可能的实现方式中，所述方法还包括：In a possible implementation, the method also includes:

获取所述本车与所述目标车辆在所述当前时刻的相对速度；Acquiring the relative speed between the vehicle and the target vehicle at the current moment;

根据所述目标车辆在所述当前时刻的换道概率，确定所述目标车辆在所述当前时刻的直行概率，所述直行概率用于指示所述目标车辆进行直行的概率；determining the straight-going probability of the target vehicle at the current moment according to the lane-changing probability of the target vehicle at the current moment, the straight-going probability being used to indicate the probability of the target vehicle going straight;

获取所述多个第一驾驶行为特征值中每个第一驾驶行为特征值在所述当前时刻的加权系数；Acquiring the weighting coefficient of each first driving behavior characteristic value in the plurality of first driving behavior characteristic values at the current moment;

根据所述相对速度、所述直行概率、所述多个第一驾驶行为特征值以及所述每个第一驾驶行为特征值在所述当前时刻的加权系数，通过预设的纵向运动方程，确定所述目标车辆在所述当前时刻的下一个时刻的运动速度。According to the relative speed, the probability of going straight, the plurality of first driving behavior characteristic values and the weighting coefficient of each first driving behavior characteristic value at the current moment, through a preset longitudinal motion equation, determine The moving speed of the target vehicle at the next moment of the current moment.

在本公开实施例中，不仅可以预测运动轨迹，还可以预测运动速度，从而能够实现时空轨迹的预测，进一步提高了准确性。In the embodiment of the present disclosure, not only the motion trajectory can be predicted, but also the motion speed can be predicted, so that the prediction of the space-time trajectory can be realized, and the accuracy can be further improved.

在一个可能的实现方式中，所述获取所述多个第一驾驶行为特征值中每个第一驾驶行为特征值在所述当前时刻的加权系数，包括：In a possible implementation manner, the obtaining the weighting coefficient of each first driving behavior characteristic value in the plurality of first driving behavior characteristic values at the current moment includes:

获取所述目标车辆在所述当前时刻之前的上一个时刻的多个第二驾驶行为特征值，以及所述多个第二驾驶行为特征值中每个第二驾驶行为特征值在所述上一个时刻的加权系数；Acquiring multiple second driving behavior characteristic values of the target vehicle at a previous moment before the current moment, and each second driving behavior characteristic value in the plurality of second driving behavior characteristic values at the previous The weighting coefficient of time;

根据所述每个第二驾驶行为特征值和所述每个第二驾驶行为特征值在所述上一个时刻的加权系数以及所述目标车辆在所述当前时刻的运动速度，通过预设算法，确定所述每个第一驾驶行为特征值在所述当前时刻的加权系数。According to each of the second driving behavior characteristic values and the weighting coefficient of each second driving behavior characteristic value at the last moment and the moving speed of the target vehicle at the current moment, through a preset algorithm, Determine the weighting coefficient of each first driving behavior characteristic value at the current moment.

在本公开实施例中，结合上一时刻的加权系数迭代得到当前时刻的加权系数，而上一时刻的加权系数，也是通过上上时刻的加权系数迭代得到的。因此，本公开实施例中，是通过迭代多个历史时刻的加权系数，得到当前时刻的加权系数，从而提高了确定出的加权系数的准确性，进而进一步提高了后续预测运动速度的准确性。In the embodiment of the present disclosure, the weighting coefficient at the current time is iteratively obtained in combination with the weighting coefficient at the previous time, and the weighting coefficient at the previous time is also obtained through iterating the weighting coefficient at the previous time. Therefore, in the embodiment of the present disclosure, the weighting coefficient at the current time is obtained by iterating the weighting coefficient at multiple historical moments, thereby improving the accuracy of the determined weighting coefficient, and further improving the accuracy of the subsequent predicted motion speed.

在一个可能的实现方式中，所述获取目标车辆在当前时刻的多个第一驾驶行为特征值，包括：In a possible implementation, the acquiring multiple first driving behavior characteristic values of the target vehicle at the current moment includes:

获取所述多个驾驶行为特征，以及，获取所述目标车辆在所述当前时刻的运动信息，所述运动信息包括运动速度、运动加速度和运动方向中的至少一个；Acquiring the plurality of driving behavior characteristics, and obtaining motion information of the target vehicle at the current moment, the motion information including at least one of motion speed, motion acceleration, and motion direction;

根据所述运动信息和所述多个驾驶行为特征，确定所述目标车辆在所述当前时刻的多个第一驾驶行为特征值。A plurality of first driving behavior characteristic values of the target vehicle at the current moment are determined according to the motion information and the plurality of driving behavior characteristics.

在一个可能的实现方式中，所述获取所述多个驾驶行为特征，包括：In a possible implementation, the acquiring the multiple driving behavior features includes:

向服务器发送获取请求，所述获取请求用于请求所述服务器发送所述多个驾驶行为特征；sending an acquisition request to a server, where the acquisition request is used to request the server to send the plurality of driving behavior characteristics;

接收所述服务器根据所述获取请求发送的所述多个驾驶行为特征，所述多个驾驶行为特征为所述服务器根据所述多个车辆的历史运动信息得到的与所述驾驶行为相关的特征。receiving the multiple driving behavior features sent by the server according to the acquisition request, the multiple driving behavior features being features related to the driving behavior obtained by the server according to historical motion information of the multiple vehicles .

在本公开实施例中，可以由服务器训练得到与驾驶行为相关的多个驾驶行为特征，本车直接从服务器中获取多个驾驶行为特征，从而提高了获取多个驾驶行为特征的效率。In the embodiment of the present disclosure, multiple driving behavior features related to driving behavior can be obtained through server training, and the vehicle directly acquires multiple driving behavior features from the server, thereby improving the efficiency of acquiring multiple driving behavior features.

在一个可能的实现方式中，所述获取所述多个驾驶行为特征，包括：In a possible implementation, the acquiring the multiple driving behavior features includes:

获取所述多个车辆的历史运动信息，所述历史运动信息包括运动速度、运动加速度和运动方向中的至少一个；Acquiring historical motion information of the plurality of vehicles, the historical motion information including at least one of motion speed, motion acceleration, and motion direction;

根据所述多个车辆的历史运动信息，通过预设训练模型，确定所述多个驾驶行为特征。The plurality of driving behavior characteristics are determined through a preset training model according to historical motion information of the plurality of vehicles.

在本公开实施例中，可以由本车训练与驾驶行为相关的多个驾驶行为特征，从而提高了准确性。In the embodiment of the present disclosure, multiple driving behavior features related to the driving behavior can be trained by the self-vehicle, thereby improving the accuracy.

在一个可能的实现方式中，所述根据所述多个车辆的历史运动信息，通过预设训练模型，确定所述多个驾驶行为特征，包括：In a possible implementation manner, the determining the multiple driving behavior characteristics through a preset training model according to the historical motion information of the multiple vehicles includes:

根据所述多个车辆中每个车辆的历史运动信息，分别确定所述每个车辆在历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，所述历史时刻为所述历史运动信息对应的时刻；According to the historical motion information of each vehicle in the plurality of vehicles, respectively determine the lateral velocity, linear velocity, lateral acceleration of each vehicle at historical moments, and the lane line lateral direction between each vehicle and the lane line where it is located Deviation, the historical moment is the moment corresponding to the historical motion information;

获取所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，所述统计信息包括均值、标准差、变异系数、均方根、香农熵、对数能量熵和正态熵中的至少一个；Acquiring the statistical information of each vehicle within at least one preset period before the historical moment, the statistical information including mean value, standard deviation, coefficient of variation, root mean square, Shannon entropy, logarithmic energy entropy and normal at least one of entropy;

根据所述每个车辆在所述历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，以及所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定所述多个驾驶行为特征。According to the lateral velocity, linear velocity, lateral acceleration of each vehicle at the historical moment and the lane line lateral deviation between each vehicle and the lane line where it is located, and each vehicle before the historical moment Statistical information within at least one preset period of time is used to determine the plurality of driving behavior characteristics through a preset training model.

在本公开实施例中，通过时间滑窗技术获取丰富的特征信息。相比业界常见做法(人工选取特征)，本发明通过获取丰富的特征信息，提高了检测的准确性和可靠性。业界常见算法的检测准确率大多在80％～90％，且没有对时效性进行评价。本发明不仅在准确性上达到了业界先进水平，在时效性上也达到了很高的水平(换道行为启动后1s内识别完成)。In the embodiments of the present disclosure, rich feature information is obtained through time sliding window technology. Compared with the common practice in the industry (manual feature selection), the present invention improves the accuracy and reliability of detection by obtaining rich feature information. The detection accuracy of common algorithms in the industry is mostly 80% to 90%, and the timeliness is not evaluated. The present invention not only reaches the advanced level in the industry in terms of accuracy, but also reaches a very high level in terms of timeliness (recognition is completed within 1 second after the behavior of changing lanes is started).

在一个可能的实现方式中，所述根据所述每个车辆在所述历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，以及所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定所述多个驾驶行为特征，包括：In a possible implementation manner, according to the lateral velocity, linear velocity, lateral acceleration of each vehicle at the historical moment, and the lane line lateral deviation between each vehicle and the lane line where it is located, and the Statistical information of each vehicle within at least one preset period of time before the historical moment, and determine the plurality of driving behavior characteristics through a preset training model, including:

根据所述每个车辆在历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，以及所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定多个候选驾驶行为特征；According to the lateral velocity, linear velocity, lateral acceleration of each vehicle at historical moments, and the lane line lateral deviation between each vehicle and its lane line, and at least Statistical information within a preset time period, through a preset training model, to determine multiple candidate driving behavior characteristics;

从所述多个候选驾驶行为特征中选择与所述驾驶行为相关的多个驾驶行为特征。A plurality of driving behavior features related to the driving behavior is selected from the plurality of candidate driving behavior features.

在本公开实施例中，通过时间滑窗技术获取丰富的特征信息，然后保留与换道行为相关性最高的特征，提高了获取的多个驾驶行为特征的准确性。In the embodiment of the present disclosure, rich feature information is obtained by using the time sliding window technology, and then the feature most relevant to the lane changing behavior is retained, which improves the accuracy of the acquired multiple driving behavior features.

在一个可能的实现方式中，所述从所述多个候选驾驶行为特征中选择与所述驾驶行为相关的多个驾驶行为特征，包括：In a possible implementation manner, the selecting a plurality of driving behavior characteristics related to the driving behavior from the plurality of candidate driving behavior characteristics includes:

通过预设相关度算法，分别确定所述多个候选驾驶行为特征中每个候选驾驶行为特征与所述驾驶行为之间的相关度；Determine the correlation between each candidate driving behavior feature and the driving behavior among the plurality of candidate driving behavior features through a preset correlation algorithm;

根据所述每个候选驾驶行为特征与所述换道行为之间的相关度，从所述多个候选驾驶行为特征中选择相关度满足预设条件的多个驾驶行为特征。According to the correlation between each of the candidate driving behavior features and the lane-changing behavior, a plurality of driving behavior features whose correlations satisfy a preset condition are selected from the plurality of candidate driving behavior features.

在本公开实施例中，通过时间滑窗技术获取丰富的特征信息并通过互信息筛选法、贪心前向搜索法对这些特征进行筛选，保留与换道行为相关性最高的特征。进一步提高了确定出的驾驶行为特征的准确性，进而提高了后续换道意图识别的准确性。In the embodiment of the present disclosure, rich feature information is obtained by time sliding window technology, and these features are screened by mutual information screening method and greedy forward search method, and the features most relevant to lane-changing behavior are retained. The accuracy of the determined driving behavior characteristics is further improved, thereby improving the accuracy of subsequent lane-changing intention recognition.

在一个可能的实现方式中，所述根据所述相对距离，通过预设的轨迹预测方程，预测所述目标车辆的运动轨迹之后，所述方法还包括：In a possible implementation manner, after predicting the trajectory of the target vehicle through a preset trajectory prediction equation according to the relative distance, the method further includes:

根据所述运动轨迹，确定与所述运动轨迹对应的驾驶策略；Determine a driving strategy corresponding to the movement trajectory according to the movement trajectory;

根据所述驾驶策略，控制所述本车行驶。According to the driving strategy, the vehicle is controlled to travel.

在本公开实施例中，预测出运动轨迹之后，基于该运动轨迹对应的驾驶策略，控制本车行驶，提高了本车的安全性。In the embodiment of the present disclosure, after the motion trajectory is predicted, the vehicle is controlled based on the driving strategy corresponding to the motion trajectory, which improves the safety of the vehicle.

第二方面，提供了一种轨迹预测方法，所述方法包括：In a second aspect, a trajectory prediction method is provided, the method comprising:

获取多个车辆的历史运动信息，所述历史运动信息包括运动速度、运动加速度和运动方向中的至少一个；Acquiring historical motion information of a plurality of vehicles, the historical motion information including at least one of motion speed, motion acceleration and motion direction;

根据所述多个车辆的历史运动信息，通过预设训练模型，确定多个驾驶行为特征，所述多个驾驶行为特征为与驾驶行为相关的特征，所述驾驶行为包括换道或者直行。According to the historical motion information of the plurality of vehicles, a plurality of driving behavior characteristics are determined through a preset training model, the plurality of driving behavior characteristics are characteristics related to driving behavior, and the driving behavior includes changing lanes or going straight.

在本公开实施例中，基于多个车辆的历史运动信息，通过预设训练模型，训练得到与驾驶行为相关的多个驾驶行为特征，从而相较于业界常用做法(人工选取特征)，本公开实施例相较于业界常用做法(人工选取特征)，本公开实施例能够获取丰富的驾驶行为特征，提高了获取驾驶行为特征的准确性。In the embodiment of the present disclosure, based on the historical motion information of multiple vehicles, through preset training models, multiple driving behavior features related to driving behavior are trained. Embodiments Compared with the common practice in the industry (manually selecting features), the embodiments of the present disclosure can acquire rich driving behavior features and improve the accuracy of acquiring driving behavior features.

在一个可能的实现方式中，所述根据所述多个车辆的历史运动信息，通过预设训练模型，确定多个驾驶行为特征，包括：In a possible implementation manner, the determining a plurality of driving behavior characteristics through a preset training model according to the historical movement information of the plurality of vehicles includes:

根据所述多个车辆中每个车辆的历史运动信息，分别确定所述每个车辆在历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，所述历史时刻为所述历史运动信息对应的时刻；According to the historical motion information of each vehicle in the plurality of vehicles, respectively determine the lateral velocity, linear velocity, lateral acceleration of each vehicle at historical moments, and the lane line lateral direction between each vehicle and the lane line where it is located Deviation, the historical moment is the moment corresponding to the historical motion information;

获取所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，所述统计信息包括均值、标准差、变异系数、均方根、香农熵、对数能量熵和正态熵中的至少一个；Acquiring the statistical information of each vehicle within at least one preset period before the historical moment, the statistical information including mean value, standard deviation, coefficient of variation, root mean square, Shannon entropy, logarithmic energy entropy and normal at least one of entropy;

根据所述每个车辆在所述历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，以及所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定所述多个驾驶行为特征。According to the lateral velocity, linear velocity, lateral acceleration of each vehicle at the historical moment and the lane line lateral deviation between each vehicle and the lane line where it is located, and each vehicle before the historical moment Statistical information within at least one preset period of time is used to determine the plurality of driving behavior characteristics through a preset training model.

在本公开实施例中，通过时间滑窗技术获取丰富的特征信息。相比业界常见做法(人工选取特征)，本发明通过获取丰富的特征信息，提高了检测的准确性和可靠性。业界常见算法的检测准确率大多在80％～90％，且没有对时效性进行评价。本发明不仅在准确性上达到了业界先进水平，在时效性上也达到了很高的水平(换道行为启动后1s内识别完成)。In the embodiments of the present disclosure, rich feature information is obtained through time sliding window technology. Compared with the common practice in the industry (manual feature selection), the present invention improves the accuracy and reliability of detection by obtaining rich feature information. The detection accuracy of common algorithms in the industry is mostly 80% to 90%, and the timeliness is not evaluated. The present invention not only reaches the advanced level in the industry in terms of accuracy, but also reaches a very high level in terms of timeliness (recognition is completed within 1 second after the behavior of changing lanes is started).

在一个可能的实现方式中，所述根据所述每个车辆在所述历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，以及所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定所述多个驾驶行为特征，包括：In a possible implementation manner, according to the lateral velocity, linear velocity, lateral acceleration of each vehicle at the historical moment, and the lane line lateral deviation between each vehicle and the lane line where it is located, and the Statistical information of each vehicle within at least one preset period of time before the historical moment, and determine the plurality of driving behavior characteristics through a preset training model, including:

根据所述每个车辆在历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，以及所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定多个候选驾驶行为特征；According to the lateral velocity, linear velocity, lateral acceleration of each vehicle at historical moments, and the lane line lateral deviation between each vehicle and its lane line, and at least Statistical information within a preset time period, through a preset training model, to determine multiple candidate driving behavior characteristics;

从所述多个候选驾驶行为特征中选择与所述驾驶行为相关的多个驾驶行为特征。A plurality of driving behavior features related to the driving behavior is selected from the plurality of candidate driving behavior features.

在本公开实施例中，通过时间滑窗技术获取丰富的特征信息，然后保留与换道行为相关性最高的特征，提高了获取的多个驾驶行为特征的准确性。In the embodiment of the present disclosure, rich feature information is obtained by using the time sliding window technology, and then the feature most relevant to the lane changing behavior is retained, which improves the accuracy of the acquired multiple driving behavior features.

在一个可能的实现方式中，所述从所述多个候选驾驶行为特征中选择与所述驾驶行为相关的多个驾驶行为特征，包括：In a possible implementation manner, the selecting a plurality of driving behavior characteristics related to the driving behavior from the plurality of candidate driving behavior characteristics includes:

通过预设相关度算法，分别确定所述多个候选驾驶行为特征中每个候选驾驶行为特征与所述驾驶行为之间的相关度；Determine the correlation between each candidate driving behavior feature and the driving behavior among the plurality of candidate driving behavior features through a preset correlation algorithm;

根据所述每个候选驾驶行为特征与所述换道行为之间的相关度，从所述多个候选驾驶行为特征中选择相关度满足预设条件的多个驾驶行为特征。According to the correlation between each of the candidate driving behavior features and the lane-changing behavior, a plurality of driving behavior features whose correlations satisfy a preset condition are selected from the plurality of candidate driving behavior features.

在本公开实施例中，通过时间滑窗技术获取丰富的特征信息并通过互信息筛选法、贪心前向搜索法对这些特征进行筛选，保留与换道行为相关性最高的特征。进一步提高了确定出的驾驶行为特征的准确性，进而提高了本车基于多个驾驶行为特征，进行换道意图识别的准确性。In the embodiment of the present disclosure, rich feature information is obtained by time sliding window technology, and these features are screened by mutual information screening method and greedy forward search method, and the features most relevant to lane-changing behavior are retained. The accuracy of the determined driving behavior characteristics is further improved, thereby improving the accuracy of lane-changing intention recognition of the vehicle based on multiple driving behavior characteristics.

在一个可能的实现方式中，所述根据所述多个车辆的历史运动信息，通过预设训练模型，确定多个驾驶行为特征之后，所述方法还包括：In a possible implementation, after determining a plurality of driving behavior characteristics according to the historical motion information of the plurality of vehicles through a preset training model, the method further includes:

接收车辆发送的获取请求，所述获取请求用于请求所述服务器发送所述多个驾驶行为特征；receiving an acquisition request sent by the vehicle, where the acquisition request is used to request the server to send the plurality of driving behavior characteristics;

根据所述获取请求，向所述车辆发送所述多个驾驶行为特征。Sending the plurality of driving behavior features to the vehicle according to the acquisition request.

在本公开实施例中，车辆从服务器中获取多个驾驶行为特征，从而基于该多个驾驶行为特征，进行后续轨迹预测，提高了获取多个驾驶行为特征的效率，并提高了后续轨迹预测的准确性。In the embodiment of the present disclosure, the vehicle obtains a plurality of driving behavior features from the server, thereby performing subsequent trajectory prediction based on the multiple driving behavior characteristics, which improves the efficiency of obtaining multiple driving behavior characteristics, and improves the accuracy of subsequent trajectory prediction. accuracy.

第三方面，提供了一种轨迹预测装置，所述装置包括：In a third aspect, a trajectory prediction device is provided, the device comprising:

第一获取单元，用于获取目标车辆在当前时刻的多个第一驾驶行为特征值，所述目标车辆为本车周围的车辆，所述多个第一驾驶行为特征值与多个驾驶行为特征一一对应，所述多个驾驶行为特征为根据多个车辆的历史运动信息得到的与驾驶行为相关的特征，所述驾驶行为包括换道或者直行；The first acquisition unit is used to acquire a plurality of first driving behavior characteristic values of the target vehicle at the current moment, the target vehicle is a vehicle around the vehicle, the plurality of first driving behavior characteristic values and the plurality of driving behavior characteristics In one-to-one correspondence, the multiple driving behavior features are features related to driving behavior obtained according to historical motion information of multiple vehicles, and the driving behavior includes changing lanes or going straight;

识别单元，用于根据所述多个第一驾驶行为特征值，识别所述目标车辆在所述当前时刻是否有换道意图；An identification unit, configured to identify whether the target vehicle has an intention to change lanes at the current moment according to the plurality of first driving behavior characteristic values;

第二获取单元，用于当识别到所述目标车辆在所述当前时刻有换道意图时，获取所述目标车辆与所述本车之间的相对距离；A second acquiring unit, configured to acquire the relative distance between the target vehicle and the own vehicle when it is recognized that the target vehicle has an intention to change lanes at the current moment;

预测单元，用于根据所述相对距离，通过预设的轨迹预测方程，预测所述目标车辆的运动轨迹。The prediction unit is configured to predict the trajectory of the target vehicle according to the relative distance and through a preset trajectory prediction equation.

在一个可能的实现方式中，所述识别单元，还用于根据所述多个第一驾驶行为特征值，获取所述目标车辆在所述当前时刻的观测矩阵，所述观测矩阵用于指示所述目标车辆在不同驾驶行为的先验概率；获取所述目标车辆在所述当前时刻之前的上一时刻的第一概率转移矩阵，所述第一概率转移矩阵用于指示所述目标车辆在不同驾驶行为之间的转移概率；根据所述第一概率转移矩阵和所述观测矩阵，确定所述目标车辆在所述当前时刻的第二概率转移矩阵；根据所述第二概率转移矩阵，确定所述目标车辆在所述当前时刻的换道概率，所述换道概率用于指示所述目标车辆进行换道的概率；当所述换道概率大于预设概率时，识别出所述目标车辆在当前时刻有换道意图。In a possible implementation manner, the identification unit is further configured to acquire an observation matrix of the target vehicle at the current moment according to the plurality of first driving behavior characteristic values, and the observation matrix is used to indicate the the prior probability of the target vehicle in different driving behaviors; obtain the first probability transition matrix of the target vehicle at the previous moment before the current moment, and the first probability transition matrix is used to indicate that the target vehicle is in different The transition probability between driving behaviors; according to the first probability transition matrix and the observation matrix, determine the second probability transition matrix of the target vehicle at the current moment; according to the second probability transition matrix, determine the The lane change probability of the target vehicle at the current moment, the lane change probability is used to indicate the probability of the target vehicle changing lanes; when the lane change probability is greater than the preset probability, it is recognized that the target vehicle is in the There is an intention to change lanes at this moment.

在一个可能的实现方式中，所述识别单元，还用于根据所述多个第一驾驶行为特征值，确定所述目标车辆的驾驶行为特征向量；获取所述目标车辆在当前时刻之前的预设时长内的平均驾驶行为特征向量，所述平均驾驶行为特征向量包括多个平均驾驶行为特征值，所述多个平均驾驶行为特征值与多个驾驶行为特征一一对应；根据所述驾驶行为特征向量和所述平均驾驶行为特征向量，通过预设的高斯判别公式确定所述目标车辆在所述当前时刻的观测矩阵。In a possible implementation manner, the identification unit is further configured to determine the driving behavior feature vector of the target vehicle according to the plurality of first driving behavior feature values; Set the average driving behavior feature vector within the duration, the average driving behavior feature vector includes a plurality of average driving behavior feature values, the plurality of average driving behavior feature values correspond to a plurality of driving behavior features; according to the driving behavior The eigenvector and the average driving behavior eigenvector determine the observation matrix of the target vehicle at the current moment through a preset Gaussian discriminant formula.

在一个可能的实现方式中，所述预测单元，还用于获取预设的轨迹预测方程中的横向偏移值和纵向偏移值；获取所述预设的轨迹预测方程中的第一形状参数值和第二形状参数值；根据所述横向偏移值、所述纵向偏移值、所述第一形状参数值、第二形状参数值和所述相对距离，通过所述预设的轨迹预测方程，预测所述目标车辆的运动轨迹。In a possible implementation manner, the predicting unit is further configured to obtain a lateral offset value and a vertical offset value in a preset trajectory prediction equation; obtain a first shape parameter in the preset trajectory prediction equation value and the second shape parameter value; according to the horizontal offset value, the vertical offset value, the first shape parameter value, the second shape parameter value and the relative distance, the preset trajectory prediction Equation to predict the trajectory of the target vehicle.

在一个可能的实现方式中，所述预测单元，还用于获取第一形状参数的第一概率密度分布和第二形状参数的第二概率密度分布，所述第一概率密度分布包括所述第一形状参数的参数值和概率密度分布的对应关系，所述第二概率密度分布包括所述第二形状参数的参数值和概率密度分布的对应关系；从所述第一概率密度分布中选择最大概率密度值对应的所述第一形状参数值，以及从所述第二概率密度分布中选择最大概率密度值对应的所述第二形状参数值。In a possible implementation manner, the prediction unit is further configured to obtain a first probability density distribution of the first shape parameter and a second probability density distribution of the second shape parameter, the first probability density distribution includes the first A corresponding relationship between a parameter value of a shape parameter and a probability density distribution, and the second probability density distribution includes a corresponding relationship between a parameter value of the second shape parameter and a probability density distribution; selecting a maximum from the first probability density distribution The first shape parameter value corresponding to a probability density value, and the second shape parameter value corresponding to a maximum probability density value selected from the second probability density distribution.

在一个可能的实现方式中，所述预测单元，还用于根据所述目标车辆在当前时刻的位置信息，获取与所述位置信息对应的所述第一概率密度分布和所述第二概率密度分布；或者，In a possible implementation manner, the prediction unit is further configured to acquire the first probability density distribution and the second probability density corresponding to the position information according to the position information of the target vehicle at the current moment distribution; or,

所述预测单元，还用于根据所述目标车辆在当前时刻的环境信息，获取与所述环境信息对应的所述第一概率密度分布和所述第二概率密度分布。The predicting unit is further configured to acquire the first probability density distribution and the second probability density distribution corresponding to the environment information according to the environment information of the target vehicle at the current moment.

在一个可能的实现方式中，所述装置还包括：In a possible implementation, the device further includes:

第三获取单元，用于获取所述本车与所述目标车辆在所述当前时刻的相对速度；a third acquisition unit, configured to acquire the relative speed between the host vehicle and the target vehicle at the current moment;

第一确定单元，用于根据所述目标车辆在所述当前时刻的换道概率，确定所述目标车辆在所述当前时刻的直行概率，所述直行概率用于指示所述目标车辆进行直行的概率；A first determining unit, configured to determine the straight-going probability of the target vehicle at the current moment according to the lane-changing probability of the target vehicle at the current moment, and the straight-going probability is used to instruct the target vehicle to go straight probability;

第四获取单元，用于获取所述多个第一驾驶行为特征值中每个第一驾驶行为特征值在所述当前时刻的加权系数；A fourth obtaining unit, configured to obtain a weighting coefficient of each first driving behavior characteristic value in the plurality of first driving behavior characteristic values at the current moment;

第二确定单元，用于根据所述相对速度、所述直行概率、所述多个第一驾驶行为特征值以及所述每个第一驾驶行为特征值在所述当前时刻的加权系数，通过预设的纵向运动方程，确定所述目标车辆在所述当前时刻的下一个时刻的运动速度。The second determination unit is configured to, according to the relative speed, the probability of going straight, the plurality of first driving behavior characteristic values, and the weighting coefficient of each first driving behavior characteristic value at the current moment, by predicting Set the longitudinal motion equation to determine the moving speed of the target vehicle at the next moment of the current moment.

在一个可能的实现方式中，所述第四获取单元，还用于获取所述目标车辆在所述当前时刻之前的上一个时刻的多个第二驾驶行为特征值，以及所述多个第二驾驶行为特征值中每个第二驾驶行为特征值在所述上一个时刻的加权系数；根据所述每个第二驾驶行为特征值和所述每个第二驾驶行为特征值在所述上一个时刻的加权系数以及所述目标车辆在所述当前时刻的运动速度，通过预设算法，确定所述每个第一驾驶行为特征值在所述当前时刻的加权系数。In a possible implementation manner, the fourth acquiring unit is further configured to acquire a plurality of second driving behavior characteristic values of the target vehicle at a previous moment before the current moment, and the plurality of second The weighting coefficient of each second driving behavior characteristic value in the driving behavior characteristic value at the last moment; according to each second driving behavior characteristic value and each second driving behavior characteristic value at the last The weighting coefficient at the time and the moving speed of the target vehicle at the current time are used to determine the weighting coefficient of each first driving behavior characteristic value at the current time through a preset algorithm.

在一个可能的实现方式中，所述第一获取单元，还用于获取所述多个驾驶行为特征，以及，获取所述目标车辆在所述当前时刻的运动信息，所述运动信息包括运动速度、运动加速度和运动方向中的至少一个；根据所述运动信息和所述多个驾驶行为特征，确定所述目标车辆在所述当前时刻的多个第一驾驶行为特征值。In a possible implementation manner, the first acquisition unit is further configured to acquire the plurality of driving behavior characteristics, and acquire motion information of the target vehicle at the current moment, the motion information includes motion speed At least one of motion acceleration and motion direction; according to the motion information and the multiple driving behavior features, determine a plurality of first driving behavior characteristic values of the target vehicle at the current moment.

在一个可能的实现方式中，所述第一获取单元，还用于向服务器发送获取请求，所述获取请求用于请求所述服务器发送所述多个驾驶行为特征；接收所述服务器根据所述获取请求发送的所述多个驾驶行为特征，所述多个驾驶行为特征为所述服务器根据所述多个车辆的历史运动信息得到的与所述驾驶行为相关的特征。In a possible implementation manner, the first obtaining unit is further configured to send an obtaining request to a server, where the obtaining request is used to request the server to send the plurality of driving behavior characteristics; Obtaining the multiple driving behavior features sent by the request, where the multiple driving behavior features are features related to the driving behavior obtained by the server according to historical motion information of the multiple vehicles.

在一个可能的实现方式中，所述第一获取单元，还用于获取所述多个车辆的历史运动信息，所述历史运动信息包括运动速度、运动加速度和运动方向中的至少一个；根据所述多个车辆的历史运动信息，通过预设训练模型，确定所述多个驾驶行为特征。In a possible implementation manner, the first acquiring unit is further configured to acquire historical motion information of the plurality of vehicles, where the historical motion information includes at least one of motion speed, motion acceleration, and motion direction; according to the The historical motion information of the plurality of vehicles is used to determine the plurality of driving behavior characteristics through a preset training model.

在一个可能的实现方式中，所述第一获取单元，还用于根据所述多个车辆中每个车辆的历史运动信息，分别确定所述每个车辆在历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，所述历史时刻为所述历史运动信息对应的时刻；获取所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，所述统计信息包括均值、标准差、变异系数、均方根、香农熵、对数能量熵和正态熵中的至少一个；根据所述每个车辆在所述历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，以及所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定所述多个驾驶行为特征。In a possible implementation manner, the first acquisition unit is further configured to respectively determine the lateral velocity, linear velocity, Lateral acceleration and the lane line lateral deviation between each vehicle and the lane line where it is located, the historical moment is the moment corresponding to the historical motion information; obtain at least one forecast of each vehicle before the historical moment Statistical information within a given time period, the statistical information including at least one of mean value, standard deviation, coefficient of variation, root mean square, Shannon entropy, logarithmic energy entropy and normal entropy; according to each vehicle in the history The lateral velocity, linear velocity, lateral acceleration and the lateral deviation of the lane line between each vehicle and the lane line at the moment, and the statistical information of each vehicle within at least one preset time period before the historical moment , determining the plurality of driving behavior characteristics by preset training models.

在一个可能的实现方式中，所述第一获取单元，还用于根据所述每个车辆在历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，以及所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定多个候选驾驶行为特征；从所述多个候选驾驶行为特征中选择与所述驾驶行为相关的多个驾驶行为特征。In a possible implementation manner, the first acquisition unit is further configured to: line lateral deviation, and the statistical information of each vehicle within at least one preset time period before the historical moment, and determine a plurality of candidate driving behavior characteristics through a preset training model; from the plurality of candidate driving behavior characteristics Select a plurality of driving behavior characteristics related to the driving behavior.

在一个可能的实现方式中，所述第一获取单元，还用于通过预设相关度算法，分别确定所述多个候选驾驶行为特征中每个候选驾驶行为特征与所述驾驶行为之间的相关度；根据所述每个候选驾驶行为特征与所述换道行为之间的相关度，从所述多个候选驾驶行为特征中选择相关度满足预设条件的多个驾驶行为特征。In a possible implementation manner, the first acquisition unit is further configured to respectively determine the relationship between each candidate driving behavior feature among the plurality of candidate driving behavior features and the driving behavior by using a preset correlation algorithm. Correlation: according to the correlation between each candidate driving behavior feature and the lane-changing behavior, select a plurality of driving behavior features whose correlation meets a preset condition from the plurality of candidate driving behavior features.

在一个可能的实现方式中，所述装置还包括：In a possible implementation, the device further includes:

控制单元，用于根据所述运动轨迹，确定与所述运动轨迹对应的驾驶策略；根据所述驾驶策略，控制所述本车行驶。The control unit is configured to determine a driving strategy corresponding to the movement trajectory according to the movement trajectory; and control the driving of the vehicle according to the driving strategy.

第四方面，提供了了一种轨迹预测装置，所述装置包括：In a fourth aspect, a trajectory prediction device is provided, the device comprising:

第五获取单元，用于获取多个车辆的历史运动信息，所述历史运动信息包括运动速度、运动加速度和运动方向中的至少一个；A fifth acquisition unit, configured to acquire historical movement information of a plurality of vehicles, the historical movement information including at least one of movement speed, movement acceleration and movement direction;

第三确定单元，用于根据所述多个车辆的历史运动信息，通过预设训练模型，确定多个驾驶行为特征，所述多个驾驶行为特征为与驾驶行为相关的特征，所述驾驶行为包括换道或者直行。The third determining unit is configured to determine a plurality of driving behavior characteristics through a preset training model according to historical motion information of the plurality of vehicles, the plurality of driving behavior characteristics are characteristics related to driving behavior, and the driving behavior Including changing lanes or going straight.

在一个可能的实现方式中，所述第三确定单元，还用于根据所述多个车辆中每个车辆的历史运动信息，分别确定所述每个车辆在历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，所述历史时刻为所述历史运动信息对应的时刻；获取所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，所述统计信息包括均值、标准差、变异系数、均方根、香农熵、对数能量熵和正态熵中的至少一个；根据所述每个车辆在所述历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，以及所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定所述多个驾驶行为特征。In a possible implementation manner, the third determination unit is further configured to respectively determine the lateral velocity, linear velocity, Lateral acceleration and the lane line lateral deviation between each vehicle and the lane line where it is located, the historical moment is the moment corresponding to the historical motion information; obtain at least one forecast of each vehicle before the historical moment Statistical information within a given time period, the statistical information including at least one of mean value, standard deviation, coefficient of variation, root mean square, Shannon entropy, logarithmic energy entropy and normal entropy; according to each vehicle in the history The lateral velocity, linear velocity, lateral acceleration and the lateral deviation of the lane line between each vehicle and the lane line at the moment, and the statistical information of each vehicle within at least one preset time period before the historical moment , determining the plurality of driving behavior characteristics by preset training models.

在一个可能的实现方式中，所述第三确定单元，还用于根据所述每个车辆在历史时刻的横向速度、线速度、横向加速度和所述每个车辆与所在车道线之间的车道线横向偏差，以及所述每个车辆在所述历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定多个候选驾驶行为特征；从所述多个候选驾驶行为特征中选择与所述驾驶行为相关的多个驾驶行为特征。In a possible implementation manner, the third determination unit is further configured to: line lateral deviation, and the statistical information of each vehicle within at least one preset time period before the historical moment, and determine a plurality of candidate driving behavior characteristics through a preset training model; from the plurality of candidate driving behavior characteristics Select a plurality of driving behavior characteristics related to the driving behavior.

在一个可能的实现方式中，所述第三确定单元，还用于通过预设相关度算法，分别确定所述多个候选驾驶行为特征中每个候选驾驶行为特征与所述驾驶行为之间的相关度；根据所述每个候选驾驶行为特征与所述换道行为之间的相关度，从所述多个候选驾驶行为特征中选择相关度满足预设条件的多个驾驶行为特征。In a possible implementation manner, the third determining unit is further configured to respectively determine the relationship between each candidate driving behavior feature among the plurality of candidate driving behavior features and the driving behavior by using a preset correlation algorithm. Correlation: according to the correlation between each candidate driving behavior feature and the lane-changing behavior, select a plurality of driving behavior features whose correlation meets a preset condition from the plurality of candidate driving behavior features.

在一个可能的实现方式中，所述装置还包括：In a possible implementation, the device further includes:

接收单元，用于接收车辆发送的获取请求，所述获取请求用于请求所述服务器发送所述多个驾驶行为特征；a receiving unit, configured to receive an acquisition request sent by the vehicle, where the acquisition request is used to request the server to send the plurality of driving behavior characteristics;

发送单元，用于根据所述获取请求，向所述车辆发送所述多个驾驶行为特征。A sending unit, configured to send the plurality of driving behavior features to the vehicle according to the acquisition request.

第五方面，提供了一种设备，所述设备包括：处理器、存储器、通信接口及总线；In a fifth aspect, a device is provided, and the device includes: a processor, a memory, a communication interface, and a bus;

其中，所述存储器、所述处理器及所述通信接口通过所述总线连接，所述存储器上存储有可编程指令，所述处理器调用所述存储器上存储的可编程指令用于执行第一方面中任一项所述的方法。Wherein, the memory, the processor and the communication interface are connected through the bus, the memory stores programmable instructions, and the processor invokes the programmable instructions stored in the memory to execute the first The method of any one of the aspects.

第六方面，提供了一种设备，所述设备包括：处理器、存储器、通信接口及总线；In a sixth aspect, a device is provided, and the device includes: a processor, a memory, a communication interface, and a bus;

其中，所述存储器、所述处理器及所述通信接口通过所述总线连接，所述存储器上存储有可编程指令，所述处理器调用所述存储器上存储的可编程指令用于执行第二方面中任一项所述的方法。Wherein, the memory, the processor and the communication interface are connected through the bus, the memory stores programmable instructions, and the processor invokes the programmable instructions stored in the memory to execute the second The method of any one of the aspects.

第七方面，提供了一种计算机可读存储介质，其特征在于，所述存储介质包括指令，当其在计算机上运行时，使得所述计算机执行第一方面中任一项所述的方法。In a seventh aspect, there is provided a computer-readable storage medium, wherein the storage medium includes instructions, which when run on a computer, cause the computer to execute the method described in any one of the first aspect.

第八方面，提供了一种计算机可读存储介质，所述存储介质包括指令，当其在计算机上运行时，使得所述计算机执行第二方面中任一项所述的方法。In an eighth aspect, a computer-readable storage medium is provided, and the storage medium includes instructions, which, when run on a computer, cause the computer to execute the method described in any one of the second aspect.

附图说明Description of drawings

图1是本公开实施例提供的一种系统架构的示意图；FIG. 1 is a schematic diagram of a system architecture provided by an embodiment of the present disclosure;

图2-1是本公开实施例提供的一种轨迹预测的方法流程图；Figure 2-1 is a flow chart of a trajectory prediction method provided by an embodiment of the present disclosure;

图2-2是本公开实施例提供的一种参数a的分布示意图；Figure 2-2 is a schematic diagram of the distribution of a parameter a provided by an embodiment of the present disclosure;

图2-3是本公开实施例提供的一种参数b的分布示意图；Figure 2-3 is a schematic diagram of the distribution of a parameter b provided by an embodiment of the present disclosure;

图2-4是本公开实施例提供的一种换道意图识别时效性的示意图；2-4 are schematic diagrams of the timeliness of lane-changing intention recognition provided by an embodiment of the present disclosure;

图2-5是本公开实施例提供的一种轨迹预测的准确性的示意图；2-5 are schematic diagrams of the accuracy of a trajectory prediction provided by an embodiment of the present disclosure;

图3-1是本公开实施例提供的一种轨迹预测装置结构示意图；Fig. 3-1 is a schematic structural diagram of a trajectory prediction device provided by an embodiment of the present disclosure;

图3-2是本公开实施例提供的另一种轨迹预测装置结构示意图；Fig. 3-2 is a schematic structural diagram of another trajectory prediction device provided by an embodiment of the present disclosure;

图3-3是本公开实施例提供的另一种轨迹预测装置结构示意图；Fig. 3-3 is a schematic structural diagram of another trajectory prediction device provided by an embodiment of the present disclosure;

图4-1是本公开实施例提供的一种轨迹预测装置结构示意图；Fig. 4-1 is a schematic structural diagram of a trajectory prediction device provided by an embodiment of the present disclosure;

图4-2是本公开实施例提供的另一种轨迹预测装置结构示意图；Fig. 4-2 is a schematic structural diagram of another trajectory prediction device provided by an embodiment of the present disclosure;

图5是本公开实施例提供的终端结构示意图；FIG. 5 is a schematic structural diagram of a terminal provided by an embodiment of the present disclosure;

图6是本公开实施例提供的轨迹预测装置结构示意图。Fig. 6 is a schematic structural diagram of a trajectory prediction device provided by an embodiment of the present disclosure.

具体实施方式Detailed ways

为使本公开的目的、技术方案和优点更加清楚，下面将结合附图对本公开实施方式作进一步地详细描述。In order to make the purpose, technical solution and advantages of the present disclosure clearer, the implementation manners of the present disclosure will be further described in detail below in conjunction with the accompanying drawings.

上述所有可选技术方案，可以采用任意结合形成本公开的可选实施例，在此不再一一赘述。All the above optional technical solutions may be combined in any way to form optional embodiments of the present disclosure, which will not be repeated here.

本公开实施例提供了一种实施环境，该实施环境中包括本车和目标车辆。本车为任一行驶的车辆。标车辆为本车周围的车辆。本车用于预测目标车辆的运动轨迹。其中，本车可以为无人驾驶汽车，也可以为驾驶员驾驶汽车。并且，本车可以为新能源汽车，也可以为燃油汽车。其中，新能源汽车可以为纯电动汽车或者混动动力汽车。在本发明实施例中，对本车不作具体限定。同样，该目标车辆可以为无人驾驶汽车，也可以为驾驶员驾驶汽车。该目标车辆可以为新能源汽车，也可以为燃油汽车。在本公开实施例中，对目标车辆同样不作具体限定。车辆的周围是指该车辆为中心，预设半径范围内组成的区域。预设半径可以根据需要进行设置并更改，在本公开实施例中，对预设半径不作具体限定。并且，预设半径可以根据基于不同的道路属性进行设置；例如，高速公路对应的预设半径可以为100米或者200米；普通道路对应的预设半径可以为2米或者3米等。需要说明的是，本公开实施例中本车主要用于识别出目标车辆有换道意图时，才进行运动轨迹的预测。只有目标车辆在本车的车前时，目标车辆的运动轨迹才会影响到本车的行驶。因此，目标车辆可以为本车前方的车辆。The embodiment of the present disclosure provides an implementation environment, which includes an own vehicle and a target vehicle. The car is any running vehicle. The marked vehicles are the vehicles around the vehicle. The vehicle is used to predict the trajectory of the target vehicle. Wherein, the vehicle may be an unmanned vehicle, or may be a driver driving a vehicle. Moreover, the vehicle can be a new energy vehicle or a fuel vehicle. Among them, the new energy vehicle may be a pure electric vehicle or a hybrid vehicle. In the embodiment of the present invention, the vehicle is not specifically limited. Likewise, the target vehicle can be a driverless car or a driver-driven car. The target vehicle can be a new energy vehicle or a fuel vehicle. In the embodiments of the present disclosure, the target vehicle is also not specifically limited. The surroundings of the vehicle refer to the area composed of the vehicle as the center and within the preset radius range. The preset radius can be set and changed as required, and in the embodiment of the present disclosure, the preset radius is not specifically limited. Moreover, the preset radius can be set based on different road attributes; for example, the preset radius corresponding to the expressway can be 100 meters or 200 meters; the preset radius corresponding to the ordinary road can be 2 meters or 3 meters, etc. It should be noted that in the embodiment of the present disclosure, the own vehicle is mainly used to predict the trajectory of the target vehicle when it is recognized that the target vehicle has an intention to change lanes. Only when the target vehicle is in front of the vehicle, the trajectory of the target vehicle will affect the driving of the vehicle. Therefore, the target vehicle may be a vehicle ahead of the own vehicle.

在本公开实施例中，本车获取多个驾驶行为特征，根据多个驾驶行为特征，获取目标车辆在当前时刻的多个第一驾驶行为特征值；根据多个第一驾驶行为特征值，识别目标车辆在当前时刻是否有换道意图；当识别到目标车辆在当前时刻有换道意图时，预测目标车辆的运动轨迹。因此，在进行轨迹预测之前，本车或者服务器训练多个驾驶行为特征。当由服务器训练多个驾驶行为特征时，参见图1，轨迹预测的系统架构中包括本车和服务器。服务器包括交通数据采集单元、驾驶行为统计单元和特征筛选提取单元。其中，交通数据采集单元用于采集多个车辆的运动信息。该交通数据采集单元可以为摄像头等监控设备。该交通数据采集单元部署于十字路口、高速公路等位置。驾驶行为统计单元，用于根据多个车辆的运动信息，得到多个候选驾驶行为特征。特征筛选提取单元，用于从多个候选驾驶行为特征中选择与换道行为相关的多个驾驶行为特征。In the embodiment of the present disclosure, the vehicle obtains a plurality of driving behavior characteristics, and according to the plurality of driving behavior characteristics, obtains a plurality of first driving behavior characteristic values of the target vehicle at the current moment; according to the plurality of first driving behavior characteristic values, identifies Whether the target vehicle has the intention to change lanes at the current moment; when it is recognized that the target vehicle has the intention to change lanes at the current moment, the trajectory of the target vehicle is predicted. Therefore, before trajectory prediction, the vehicle or the server trains multiple driving behavior features. When multiple driving behavior features are trained by the server, see Figure 1, the trajectory prediction system architecture includes the vehicle and the server. The server includes a traffic data collection unit, a driving behavior statistics unit and a feature screening and extraction unit. Wherein, the traffic data collection unit is used to collect movement information of multiple vehicles. The traffic data collection unit may be a monitoring device such as a camera. The traffic data acquisition unit is deployed at intersections, expressways and other locations. The driving behavior statistics unit is used to obtain multiple candidate driving behavior features according to the motion information of multiple vehicles. The feature screening and extraction unit is used to select multiple driving behavior features related to lane changing behavior from multiple candidate driving behavior features.

本车包括车载感知单元、他车换道行为特征向量生成单元、他车换道意图识别单元、他车换道轨迹预测单元、驾驶行为决策单元和车辆控制执行单元。其中，车载感知单元，用于采集目标车辆在当前时刻的运动信息。该车载感知单元可以为摄像头、毫米波或者激光雷达等设备。该车载感知单元还可以为车联网通信设备等。他车换道行为特征向量生成单元，用于从服务器中获取多个驾驶行为特征，以及从车载感知单元获取目标车辆的运动信息；基于多个驾驶行为特征和目标车辆在当前时刻的运动信息，确定目标车辆在当前时刻的多个第一驾驶行为特征值。他车换道意图识别单元，用于确定根据多个第一驾驶行为特征值，识别目标车辆在当前时刻是否有换道意图。他车换道轨迹预测单元，用于在识别出目标车辆在当前时刻有换道意图时，预测目标车辆的运动轨迹。驾驶行为决策单元，用于基于该运动轨迹，确定与该运动轨迹对应的驾驶策略。车辆控制执行单元，用于基于该驾驶策略，控制该车辆行驶。需要说明的是，该本车可以应用在智能提醒装置中，确定出目标车辆的驾驶轨迹时，显示该驾驶轨迹，从而驾驶员根据目标车辆的驾驶轨迹，控制车辆行驶。该本车还可以应用在自动驾驶系统或者高级驾驶辅助系统中，本车确定出该驾驶轨迹之后，基于运动轨迹对应的驾驶策略控制该车辆行驶。The vehicle includes a vehicle-mounted perception unit, another vehicle’s lane-changing behavior feature vector generation unit, another vehicle’s lane-changing intention recognition unit, another vehicle’s lane-changing trajectory prediction unit, a driving behavior decision-making unit, and a vehicle control execution unit. Wherein, the on-vehicle perception unit is used to collect the movement information of the target vehicle at the current moment. The on-vehicle perception unit may be a device such as a camera, a millimeter wave, or a laser radar. The on-vehicle perception unit may also be a vehicle networking communication device and the like. The lane-changing behavior feature vector generation unit of other vehicles is used to obtain multiple driving behavior features from the server, and to obtain the motion information of the target vehicle from the on-board perception unit; based on the multiple driving behavior features and the motion information of the target vehicle at the current moment, A plurality of first driving behavior characteristic values of the target vehicle at the current moment are determined. The lane-changing intention recognition unit of other vehicles is configured to determine whether the target vehicle has lane-changing intentions at the current moment according to a plurality of first driving behavior characteristic values. The other vehicle lane-changing trajectory prediction unit is used to predict the trajectory of the target vehicle when it is recognized that the target vehicle has a lane-changing intention at the current moment. The driving behavior decision-making unit is configured to determine a driving strategy corresponding to the motion trajectory based on the motion trajectory. The vehicle control execution unit is configured to control the vehicle to run based on the driving strategy. It should be noted that the self-vehicle can be applied in an intelligent reminder device, and when the driving track of the target vehicle is determined, the driving track is displayed, so that the driver can control the driving of the vehicle according to the driving track of the target vehicle. The self-vehicle can also be applied in an automatic driving system or an advanced driving assistance system. After the self-vehicle determines the driving trajectory, it controls the driving of the vehicle based on the driving strategy corresponding to the motion trajectory.

本公开实施例提供了一种轨迹预测方法，该方法应用本车中，本车为任一行驶的车辆。参见图2-1，该方法包括：An embodiment of the present disclosure provides a trajectory prediction method, the method is applied to a self-vehicle, and the self-vehicle is any driving vehicle. Referring to Figure 2-1, the method includes:

步骤201：本车获取目标车辆在当前时刻的多个第一驾驶行为特征值。Step 201: The vehicle acquires a plurality of first driving behavior characteristic values of the target vehicle at the current moment.

目标车辆为本车周围的车辆，多个第一驾驶行为特征值与多个驾驶行为特征一一对应，多个驾驶行为特征为根据多个车辆的历史运动信息得到的与驾驶行为相关的特征。驾驶行为包括换道或者直行。本车在行驶过程中，实时检测本车的周围是否有车辆，当检测到目标车辆时，通过以下步骤(1)和(2)获取目标车辆在当前时刻的多个第一驾驶行为特征值。The target vehicle is the vehicles around the own vehicle, and the multiple first driving behavior feature values correspond to the multiple driving behavior features one by one, and the multiple driving behavior features are features related to driving behavior obtained according to the historical motion information of multiple vehicles. Driving behavior includes changing lanes or going straight. When the vehicle is running, it detects whether there is a vehicle around the vehicle in real time. When the target vehicle is detected, a plurality of first driving behavior characteristic values of the target vehicle at the current moment are obtained through the following steps (1) and (2).

(1)：本车获取多个驾驶行为特征。(1): The vehicle acquires multiple driving behavior features.

在本步骤中，可以由本车训练多个驾驶行为特征，也即以下第一种实现方式。也可以由服务器训练多个驾驶行为特征，本车从服务器中获取多个驾驶行为特征，也即以下第二种实现方式。In this step, multiple driving behavior features can be trained by the own vehicle, that is, the following first implementation manner. It is also possible to train multiple driving behavior features by the server, and the vehicle obtains multiple driving behavior features from the server, that is, the following second implementation.

对于第一种实现方式，本步骤可以通过以下步骤(1-1)至(1-2)实现，包括：For the first implementation, this step can be achieved through the following steps (1-1) to (1-2), including:

(1-1)：本车获取多个车辆的历史运动信息。(1-1): The vehicle acquires historical movement information of multiple vehicles.

其中，多个车辆为用于训练驾驶行为特征的样本车辆。历史运动信息包括运动速度、运动加速度和运动方向中的至少一个。Among them, multiple vehicles are sample vehicles for training driving behavior features. The historical motion information includes at least one of motion speed, motion acceleration and motion direction.

在一个可能的实现方式中，本车可以从多个车辆中获取多个车辆的历史运动信息。相应的，对于任一车辆，本车获取该车辆的历史运动信息的步骤可以为：本车向该车辆发送第一获取请求，第一获取请求用于请求获取该车辆的历史运动信息。该车辆接收本车发送的该第一获取请求，根据第一获取请求向本车发送该车辆的历史运动信息。历史运动信息中至少包括运动速度、运动加速度和运动方向中的至少一个，历史运动信息中还可以包括采集该历史运动信息的历史时刻和/或车辆所在的位置信息。In a possible implementation manner, the own vehicle may acquire historical movement information of multiple vehicles from multiple vehicles. Correspondingly, for any vehicle, the step of obtaining the historical movement information of the vehicle by the own vehicle may be: the own vehicle sends a first acquisition request to the vehicle, and the first acquisition request is used to request acquisition of the historical movement information of the vehicle. The vehicle receives the first acquisition request sent by the own vehicle, and sends the historical movement information of the vehicle to the own vehicle according to the first acquisition request. The historical motion information includes at least one of the motion speed, motion acceleration and motion direction, and the historical motion information may also include the historical time when the historical motion information was collected and/or the location information of the vehicle.

本车向该车辆发送第一获取请求之前，本车与该车辆建立信息传输通道。相应的，本车与该车辆通过该信息传输通道进行信息传输。其中，该信息传输通道可以为车联网(Vehicles of Internet，V2I)传输通道或者近距离通信(例如，蓝牙)传输通道等。Before the vehicle sends the first acquisition request to the vehicle, the vehicle establishes an information transmission channel with the vehicle. Correspondingly, the vehicle and the vehicle perform information transmission through the information transmission channel. Wherein, the information transmission channel may be a Vehicles of Internet (Vehicles of Internet, V2I) transmission channel or a short distance communication (for example, Bluetooth) transmission channel or the like.

在另一个可能的实现方式中，本车可以从监控设备中获取多个车辆的历史运动信息。相应的，本车获取多个车辆的历史运动信息的步骤可以为：本车向监控设备发送第二获取请求，第二获取请求携带多个车辆中每个车辆的车辆标识。监控设备接收本车发送的第二获取请求，根据每个车辆的车辆标识，获取每个车辆的历史运动信息，向本车发送每个车辆的历史运动信息。本车接收监控设备发送的每个车辆的历史运动信息。In another possible implementation manner, the vehicle may acquire historical movement information of multiple vehicles from the monitoring device. Correspondingly, the step for the own vehicle to obtain the historical motion information of the multiple vehicles may be: the own vehicle sends a second acquisition request to the monitoring device, and the second acquisition request carries the vehicle identification of each of the multiple vehicles. The monitoring device receives the second acquisition request sent by the own vehicle, acquires the historical movement information of each vehicle according to the vehicle identification of each vehicle, and sends the historical movement information of each vehicle to the own vehicle. The vehicle receives the historical movement information of each vehicle sent by the monitoring equipment.

监控设备为用于采集运动信息的传感器，例如，摄像头或者雷达等。为了提高精度，该雷达可以为毫米波雷达或者激光雷达等。车辆标识可以为车辆的车牌号码等。The monitoring device is a sensor for collecting motion information, for example, a camera or a radar. In order to improve the accuracy, the radar may be a millimeter-wave radar or a laser radar. The vehicle identification may be the license plate number of the vehicle or the like.

(1-2)：本车根据多个车辆的历史运动信息，通过预设训练模型，确定多个驾驶行为特征。(1-2): According to the historical motion information of multiple vehicles, the car determines multiple driving behavior characteristics through a preset training model.

预设训练模型可以为任一训练驾驶行为特征的训练模型。在本公开实施例中，对预设训练模型不作具体限定；例如，预设训练模型可以为隐马尔可夫模型。相应的，本步骤可以通过以下步骤(1-2-1)至(1-2-3)实现，包括：The preset training model can be any training model for training driving behavior characteristics. In the embodiment of the present disclosure, the preset training model is not specifically limited; for example, the preset training model may be a hidden Markov model. Correspondingly, this step can be realized through the following steps (1-2-1) to (1-2-3), including:

(1-2-1)：本车根据多个车辆中每个车辆的历史运动信息，分别确定每个车辆在历史时刻的横向速度、线速度、横向加速度和每个车辆与所在车道线之间的车道线横向偏差，历史时刻为历史运动信息对应的时刻。(1-2-1): According to the historical movement information of each vehicle in multiple vehicles, the vehicle determines the lateral velocity, linear velocity, lateral acceleration and the distance between each vehicle and the lane line of each vehicle at the historical moment. The lateral deviation of the lane line, the historical moment is the moment corresponding to the historical motion information.

每个车辆的运动信息中包括运动速度、运动加速度和运动方向。每个车辆的运动信息中还包括采集运动信息的历史时刻以及车辆所在的位置信息。相应的，对于多个车辆中每个车辆，本车确定该车辆的横向速度、线速度和横向加速度的步骤可以为：本车根据该车辆在历史时刻的运动速度和运动方向，确定该车辆在历史时刻的横向速度和线速度；根据该车辆在历史时刻的位置信息，确定该车辆与所在车道线之间的车道线横向偏差。The motion information of each vehicle includes motion speed, motion acceleration and motion direction. The motion information of each vehicle also includes the historical moment of collecting the motion information and the location information of the vehicle. Correspondingly, for each of the plurality of vehicles, the step of determining the lateral velocity, linear velocity and lateral acceleration of the vehicle by the vehicle may be as follows: the vehicle determines that the vehicle is in the The lateral velocity and linear velocity at the historical moment; according to the position information of the vehicle at the historical moment, determine the lateral deviation of the lane line between the vehicle and the lane line where it is located.

在本步骤中，本车获取到每个车辆在历史时刻的横向速度、线速度、横向加速度和每个车辆与所在车道线之间的车道线横向偏差之后，本车可以直接通过预设训练模型，确定多个驾驶行为特征。为了丰富获取的特征，本车获取到每个车辆在历史时刻的横向速度、线速度、横向加速度和每个车辆与所在车道线之间的车道线横向偏差之后，可以通过时间滑窗技术获取每个车辆在历史时刻之前的至少一个预设时长内的统计信息，基于该统计信息确定丰富的驾驶行为特征。In this step, after the vehicle obtains the lateral velocity, linear velocity, lateral acceleration of each vehicle at historical moments, and the lane line lateral deviation between each vehicle and the lane line, the vehicle can directly pass the preset training model , to determine a number of driving behavior characteristics. In order to enrich the acquired features, after the vehicle obtains the lateral velocity, linear velocity, lateral acceleration of each vehicle at historical moments, and the lane line lateral deviation between each vehicle and the lane line it is in, it can obtain each vehicle through time sliding window technology. Statistical information of a vehicle in at least one preset period of time before the historical moment, based on the statistical information to determine rich driving behavior characteristics.

(1-2-2)：本车获取每个车辆在历史时刻之前的至少一个预设时长内的统计信息，该统计信息包括均值、标准差、变异系数、均方根、香农熵、对数能量熵和正态熵中的至少一个。(1-2-2): The vehicle obtains the statistical information of each vehicle within at least one preset period of time before the historical moment, the statistical information includes mean value, standard deviation, coefficient of variation, root mean square, Shannon entropy, logarithm At least one of energy entropy and normal entropy.

对于至少一个预设时长中每个预设时长，本车获取该车辆在该预设时长内的运动信息，根据该预设时长内的运动信息，确定该车辆在该预设时长内的统计信息。需要说明的是，统计信息可以根据需要进行设置并更改，以上仅是举例并不对统计信息构成限定。至少一个预设时长中每个预设时长不同，且预设时长的数量以及每个预设时长都可以根据需要进行设置并更改，在本公开实施例中，对此不作具体限定。例如，3个预设时长，3个预设时长分别为0.5s，1s和2s。相应的，本车可以通过以下公式一，获取每个车辆在历史时刻之前的至少一个预设时长内的统计信息。For each preset duration in at least one preset duration, the vehicle obtains the motion information of the vehicle within the preset duration, and determines the statistical information of the vehicle within the preset duration according to the motion information within the preset duration . It should be noted that the statistical information can be set and changed as required, and the above is just an example and does not constitute a limitation to the statistical information. Each of the at least one preset duration is different, and the number of preset durations and each preset duration can be set and changed as required, which is not specifically limited in this embodiment of the present disclosure. For example, 3 preset durations, the 3 preset durations are 0.5s, 1s and 2s respectively. Correspondingly, the vehicle can use the following formula 1 to obtain statistical information of each vehicle within at least one preset time period before the historical moment.

其中，|车道线横线偏差|为车道线横向偏差的绝对值。变异系数为标准差与均值的比值。本车通过以下公式二确定香农熵：Among them, |lane line deviation| is the absolute value of the lane line deviation. The coefficient of variation is the ratio of the standard deviation to the mean. The car determines the Shannon entropy through the following formula 2:

公式二：Formula two:

其中，H₁(x)为x的香农熵，x_i为输入特征参数的时间序列值，并且设置0log0＝0。Wherein, H₁ (x) is the Shannon entropy of x,_xi is the time series value of the input feature parameter, and 0log0=0 is set.

本车通过以下公式三确定对数能量熵：The vehicle determines the logarithmic energy entropy through the following formula three:

公式三：Formula three:

其中，H₂(x)为x的对数能量熵，x_i为输入特征参数的时间序列值。Among them, H₂ (x) is the logarithmic energy entropy of x, and_xi is the time series value of the input feature parameter.

本车通过以下公式四确定正态熵：The vehicle determines the normal entropy through the following formula 4:

公式四：Formula four:

其中，H₃(x)为x的正态熵，p为幂指数。Wherein, H₃ (x) is the normal entropy of x, and p is the power exponent.

需要说明的是，以上只是对驾驶行为特征进行举例说明，并不构成对驾驶行为特征的具体限定。在实际应用中可以根据需要进行设置并更改，例如，选择均值、标准差、变异系数、均方根、香农熵、对数能量熵和正态熵中的部分参数。再如，选择车道线横线偏差、横向速度、横向速度、线速度和横向加速度中的部分速度。当然，后续也可以在增加几个参数，这些都不构成对驾驶行为特征的具体限定。It should be noted that the above is only an example of the driving behavior characteristics, and does not constitute a specific limitation on the driving behavior characteristics. In practical applications, it can be set and changed according to needs, for example, select some parameters in mean value, standard deviation, coefficient of variation, root mean square, Shannon entropy, logarithmic energy entropy and normal entropy. For another example, select a part of the lane deviation, lateral velocity, lateral velocity, linear velocity, and lateral acceleration. Of course, several parameters can also be added in the future, and these do not constitute a specific limitation on the characteristics of driving behavior.

在本公开实施例中，通过引入加权系数等统计学特征，并与时间滑窗技术相结合，本公开实施例可以获取丰富的驾驶行为特征。驾驶行为特征数量可达到几十个，远远超出业界常见的特征数量，丰富的驾驶行为特征大大增强了对换道行为的表征能力。相比采用单一时刻的速度或者加速度作为驾驶行为特征，本公开实施例极大地提高了特征的信息量，提高了后续对换道意图的识别准确性和时效性。In the embodiments of the present disclosure, by introducing statistical features such as weighting coefficients and combining them with the time sliding window technology, the embodiments of the present disclosure can obtain rich driving behavior features. The number of driving behavior features can reach dozens, far exceeding the number of common features in the industry. The rich driving behavior features greatly enhance the ability to characterize lane-changing behavior. Compared with using the speed or acceleration at a single moment as the driving behavior feature, the embodiments of the present disclosure greatly increase the amount of information of the feature, and improve the accuracy and timeliness of subsequent recognition of lane-changing intentions.

(1-2-3)：本车根据每个车辆在历史时刻的横向速度、线速度、横向加速度和每个车辆与所在车道线之间的车道线横向偏差，以及每个车辆在历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定多个驾驶行为特征。(1-2-3): According to the lateral velocity, linear velocity, lateral acceleration of each vehicle at the historical moment and the lateral deviation of the lane line between each vehicle and the lane line, and each vehicle before the historical moment Statistical information within at least one preset period of time, and a preset training model is used to determine a plurality of driving behavior characteristics.

在一个可能的实现方式中，本车通过以上步骤(1-2-3)确定出的多个驾驶行为特征即为最终的驾驶行为特征。在另一个可能的实现方式中，本车通过以上步骤(1-2-3)确定出的多个驾驶行为特征为候选驾驶行为特征，本车再从多个候选驾驶行为特征中选择出与驾驶行为相关的驾驶行为特征。相应的，步骤(1-2-3)可以为：In a possible implementation manner, the multiple driving behavior characteristics determined by the vehicle through the above steps (1-2-3) are the final driving behavior characteristics. In another possible implementation, the multiple driving behavior features determined by the vehicle through the above steps (1-2-3) are candidate driving behavior features, and then the vehicle selects the driving behavior features from the multiple candidate driving behavior features. Behavior-related driving behavior characteristics. Correspondingly, step (1-2-3) can be:

本车根据每个车辆在历史时刻的横向速度、线速度、横向加速度和每个车辆与所在车道线之间的车道线横向偏差，以及每个车辆在历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定多个候选驾驶行为特征；从多个候选驾驶行为特征中选择与驾驶行为相关的多个驾驶行为特征。The vehicle is based on the lateral velocity, linear velocity, lateral acceleration of each vehicle at the historical moment and the lane line lateral deviation between each vehicle and the lane line where it is located, as well as each vehicle's time in at least one preset time period before the historical moment. Statistical information, through preset training models, to determine multiple candidate driving behavior features; select multiple driving behavior features related to driving behavior from the multiple candidate driving behavior features.

在一个可能的实现方式中，本车从多个候选驾驶行为特征中选择与驾驶行为相关的多个驾驶行为特征的步骤可以通过以下步骤(a)和(b)实现，包括：In a possible implementation, the step of the vehicle selecting a plurality of driving behavior characteristics related to driving behavior from a plurality of candidate driving behavior characteristics can be realized through the following steps (a) and (b), including:

(a)：本车通过预设相关度算法，分别确定多个候选驾驶行为特征中每个候选驾驶行为特征与驾驶行为之间的相关度。(a): The car uses a preset correlation algorithm to determine the correlation between each candidate driving behavior feature and the driving behavior among multiple candidate driving behavior features.

预设相关度算法可以根据需要进行设置并更改，在本公开实施例中，对预设相关度算法不作具体限定。例如，预设相关度算法可以为贪心算法。对于多个候选驾驶行为特征中每个候选驾驶行为特征，本车通过以下公式五确定该驾驶行为特征与驾驶行为之间的相关度：The preset correlation degree algorithm can be set and changed as required, and in the embodiment of the present disclosure, the preset correlation degree algorithm is not specifically limited. For example, the preset correlation algorithm may be a greedy algorithm. For each candidate driving behavior feature in multiple candidate driving behavior features, the car determines the correlation between the driving behavior feature and the driving behavior through the following formula five:

公式五：Formula five:

f_i为该驾驶行为特征，b_j为驾驶行为，包括直行、左换道和右换道。MI(f_i,b_j)为该驾驶行为特征与该换道行为之间的相关度。x为特征在区间(fimin,fimax)积分时的变量。p(x,b_j)为当驾驶行为特征值为x时，执行换道行为b_j的概率。p(x)为驾驶行为特征值为x的概率，p(b_j)为换道行为是b_j的概率。f_i is the characteristic of the driving behavior, b_j is the driving behavior, including going straight, changing lanes left and changing lanes right. MI(f_i ,b_j ) is the correlation between the driving behavior feature and the lane changing behavior. x is the variable when the feature is integrated in the interval (fimin, fimax). p(x,b_j ) is the probability of performing lane-changing behavior b_j when the characteristic value of driving behavior is x. p(x) is the probability that the characteristic value of the driving behavior is x, and p(b_j ) is the probability that the lane-changing behavior is b_j .

(b)：本车根据每个候选驾驶行为特征与驾驶行为之间的相关度，从多个候选驾驶行为特征中选择相关度满足预设条件的多个驾驶行为特征。(b): According to the correlation between each candidate driving behavior feature and the driving behavior, the vehicle selects multiple driving behavior features whose correlations meet the preset conditions from multiple candidate driving behavior features.

预设条件可以根据需要进行设置并更改，在本公开实施例中，对预设条件不作具体限定。例如，预设条件可以为相关度最高或者相关度超过预设阈值。当预设条件为相关度最高时，本步骤可以为：本车根据每个候选驾驶行为特征与驾驶行为之间的相关度，从多个候选驾驶行为特征中选择相关度最高的指定数目个驾驶行为特征。指定数目可以根据需要进行设置并更改，在本公开实施例中，对指定数目不作具体限定。例如，指定数据可以为16或者20等。当预设条件为相关度超过预设阈值时，本步骤可以为：本车根据每个候选驾驶行为特征与驾驶行为之间的相关度，从多个候选驾驶行为特征中选择相关度超过预设阈值的多个驾驶行为特征。预设阈值可以根据需要进行设置并更改，在本公开实施例中，对预设阈值不作具体限定；例如，预设阈值可以为0.8或者0.7等。The preset conditions may be set and changed as required, and in the embodiments of the present disclosure, the preset conditions are not specifically limited. For example, the preset condition may be that the degree of correlation is the highest or the degree of correlation exceeds a preset threshold. When the default condition is the highest correlation, this step can be: the vehicle selects the specified number of driving behaviors with the highest correlation from multiple candidate driving behavior characteristics according to the correlation between each candidate driving behavior characteristic and the driving behavior. Behavioral characteristics. The specified number can be set and changed as required, and in the embodiment of the present disclosure, the specified number is not specifically limited. For example, the designated data can be 16 or 20 and so on. When the preset condition is that the correlation exceeds the preset threshold, this step can be as follows: the vehicle selects from a plurality of candidate driving behavior characteristics according to the correlation between each candidate driving behavior characteristic and the correlation degree exceeds the preset threshold. Thresholds for multiple driving behavior characteristics. The preset threshold may be set and changed as required, and in the embodiment of the present disclosure, the preset threshold is not specifically limited; for example, the preset threshold may be 0.8 or 0.7.

在另一个可能的实现方式中，本车通过贪心算法，从多个候选驾驶行为特征中选择与驾驶行为相关的多个驾驶行为特征。相应的，本车从多个候选驾驶行为特征中选择与驾驶行为相关的多个驾驶行为特征的步骤可以为：In another possible implementation, the vehicle selects multiple driving behavior features related to the driving behavior from multiple candidate driving behavior features through a greedy algorithm. Correspondingly, the steps for the vehicle to select multiple driving behavior features related to driving behavior from multiple candidate driving behavior features can be as follows:

本车定义多个空集，将多个空集作为抑制欠拟合的贪心算法的特征集，并将多个候选驾驶行为特征随机划分为训练集和测试集。训练集中包括至少一个驾驶行为特征，测试集中包括至少一个候选驾驶行为特征。本车基于训练集、测试集和贪心算法的多个特征集，从多个候选驾驶行为特征中选择与换道行为相关度最高的多个驾驶行为特征。The vehicle defines multiple empty sets, and uses the multiple empty sets as the feature set of the greedy algorithm that suppresses underfitting, and randomly divides multiple candidate driving behavior features into a training set and a test set. The training set includes at least one driving behavior feature, and the test set includes at least one candidate driving behavior feature. Based on multiple feature sets of training set, test set and greedy algorithm, the car selects multiple driving behavior features with the highest correlation with lane-changing behavior from multiple candidate driving behavior features.

在本公开实施例中，对贪心算法的特征集的数量、测试集中包括的特征数量以及训练集中包括的特征数量都不作具体限定。例如，贪心算法的特征集的数量为5，则贪心算法的特征集分别为F₁、F₂、F₃、F₄和F₅，并将多个候选驾驶行为特征的70％的候选驾驶行为特征组成训练集和30％的候选驾驶行为特征组成测试集。相应的，本车基于训练集、测试集和贪心算法的多个特征集，从多个候选驾驶行为特征中选择与换道行为相关度最高的多个驾驶行为特征的步骤可以为：In the embodiments of the present disclosure, the number of feature sets of the greedy algorithm, the number of features included in the test set, and the number of features included in the training set are not specifically limited. For example, if the number of feature sets of the greedy algorithm is 5, then the feature sets of the greedy algorithm are F₁ , F₂ , F₃ , F₄ and F₅ respectively, and 70% of the candidate driving behavior The features constitute the training set and 30% of the candidate driving behavior features constitute the test set. Correspondingly, based on multiple feature sets of the training set, test set and greedy algorithm, the steps of selecting multiple driving behavior features with the highest correlation with lane-changing behavior from multiple candidate driving behavior features can be as follows:

首先，本车在多个候选驾驶行为特征中，任意选取5个候选驾驶行为特征分别并入5个候选特征集，即利用该5个候选特征集训练模型，得出训练误差本车将前述5个特征更换为另外5个候选驾驶行为特征重新训练得出新的训练误差重复该步骤直至覆盖所有的候选驾驶行为特征。本车选取最小训练误差对应的候选驾驶行为特征并确定相应的First, among multiple candidate driving behavior features, the vehicle randomly selects 5 candidate driving behavior features Merge into 5 candidate feature sets respectively, namely Use the 5 candidate feature sets to train the model and get the training error The car will be the aforementioned 5 features Replaced by another 5 candidate driving behavior characteristics Retrain to get new training error Repeat this step until all candidate driving behavior features are covered. The vehicle selects the minimum training error Corresponding candidate driving behavior characteristics and determine the corresponding

其次，本车在Γ/F_i⁽¹⁾(i＝1,…5)中任意选取新的候选驾驶行为特征并放入候选特征集重复前述步骤并选取最小训练误差对应的候选特征加入相应的特征集。重复上述步骤，直至第M步收敛，即新选入的候选驾驶行为特征使得五个特征集对应的平均训练误差不再减少。此时，在对应的五个特征集F_i^(M)(i＝1,…5)中选取训练误差最小的特征集中的特征作为与换道行为相关度最高的多个驾驶行为特征。Secondly, the vehicle arbitrarily selects new candidate driving behavior features in Γ/F_i⁽¹⁾ (i=1,...5) and put into the candidate feature set Repeat the preceding steps and select the candidate features corresponding to the minimum training error to add to the corresponding feature set. Repeat the above steps until the Mth step converges, that is, the newly selected candidate driving behavior features make the average training error corresponding to the five feature sets no longer decrease. At this time, among the corresponding five feature sets F_i^(M) (i=1,...5), the features in the feature set with the smallest training error are selected as the driving behavior features with the highest correlation with the lane changing behavior.

在本公开实施例中，通过时间滑窗技术获取丰富的特征信息并通过互信息筛选法、贪心前向搜索法对这些特征进行筛选，保留与换道行为相关性最高的特征。相比业界常见做法(人工选取特征)，本发明通过获取丰富的特征信息，提高了检测的准确性和可靠性。业界常见算法的检测准确率大多在80％～90％，且没有对时效性进行评价。本发明不仅在准确性上达到了业界先进水平，在时效性上也达到了很高的水平(换到行为启动后1s内识别完成)。并且，经实验测试，不采用时间滑窗技术提取特征值，检测结果的准确度约为70％，大大低于本发明的方案。同时，本发明通过信息筛选算法有效剔除了不相关信息对检测结果的影响，提高了检测效率。In the embodiment of the present disclosure, rich feature information is obtained by time sliding window technology, and these features are screened by mutual information screening method and greedy forward search method, and the features most relevant to lane-changing behavior are retained. Compared with the common practice in the industry (manual feature selection), the present invention improves the accuracy and reliability of detection by obtaining rich feature information. The detection accuracy of common algorithms in the industry is mostly 80% to 90%, and the timeliness is not evaluated. The present invention not only reaches the advanced level in the industry in terms of accuracy, but also reaches a very high level in terms of timeliness (recognition is completed within 1 second after the behavior is started). Moreover, through experimental testing, the accuracy of the detection result is about 70% without using the time sliding window technology to extract the feature value, which is much lower than the solution of the present invention. At the same time, the invention effectively eliminates the influence of irrelevant information on the detection result through the information screening algorithm, thereby improving the detection efficiency.

需要说明的是，本车可以在预测目标车辆的运动轨迹时通过以上步骤获取多个驾驶行为特征。本车也可以在预测目标车辆的运动轨迹之前，通过以上步骤获取多个驾驶行为特征，并存储多个驾驶行为特征。本车在预测目标车辆的运动轨迹时，本车直接获取已存储的多个驾驶行为特征。It should be noted that the own vehicle can obtain multiple driving behavior features through the above steps when predicting the trajectory of the target vehicle. The vehicle may also obtain multiple driving behavior features through the above steps before predicting the trajectory of the target vehicle, and store the multiple driving behavior features. When the vehicle is predicting the trajectory of the target vehicle, the vehicle directly obtains multiple stored driving behavior features.

对于第二种实现方式，本步骤可以通过以下步骤(1-a)至(1-c)实现，包括：For the second implementation, this step can be achieved through the following steps (1-a) to (1-c), including:

(1-a)：本车向服务器发送第三获取请求，第三获取请求用于请求服务器发送多个驾驶行为特征。(1-a): The vehicle sends a third acquisition request to the server, and the third acquisition request is used to request the server to send multiple driving behavior characteristics.

在本步骤之前，服务器需要训练得到多个驾驶行为特征。服务器训练得到多个驾驶行为特征的具体过程和本车训练得到多个驾驶行为特征的过程相同，在此不再赘述。Before this step, the server needs to be trained to obtain multiple driving behavior features. The specific process of the server training to obtain multiple driving behavior features is the same as the process of the vehicle training to obtain multiple driving behavior features, and will not be repeated here.

(1-b)：服务器接收本车发送的第三获取请求，根据第三获取请求向本车发送多个驾驶行为特征。(1-b): The server receives the third acquisition request sent by the vehicle, and sends multiple driving behavior characteristics to the vehicle according to the third acquisition request.

(1-c)：本车接收服务器根据第三获取请求发送的多个驾驶行为特征，多个驾驶行为特征为服务器根据多个车辆的历史运动信息得到的与驾驶行为相关的特征。(1-c): The vehicle receives multiple driving behavior features sent by the server according to the third acquisition request, and the multiple driving behavior features are features related to driving behavior obtained by the server based on historical motion information of multiple vehicles.

需要说明的是，本车或者服务器训练得到多个驾驶行为特征之后，本车或者服务器还可以每个预设周期重新训练并更新多个驾驶行为特征。预设周期可以根据需要进行设置并更改，在本公开实施例中，对预设周期不作具体限定。例如，预设周期可以为1天或者1个月等。It should be noted that after the vehicle or the server obtains multiple driving behavior features through training, the vehicle or the server can also retrain and update the multiple driving behavior features every preset period. The preset period may be set and changed as required, and in the embodiment of the present disclosure, the preset period is not specifically limited. For example, the preset period may be 1 day or 1 month.

在本公开实施例中，通过互信息筛选法和贪心前向搜索法筛选与换道行为相关性最大的多个驾驶行为特征，并将这些驾驶行为特征的统计、提取、筛选等过程放到服务器中进行离线处理，因此对驾驶行为特征的种类、数量都没有约束，可以通过在线更新的方式对在线预测模型参数进行更新。当传感器类型增加或统计数据集发生变化时，均可增加或改变特征的类型重新利用本发明进行处理。因此，本发明解决了常见算法中特征合理性不明确、模型不可扩展的问题。In the embodiment of the present disclosure, a plurality of driving behavior features most relevant to lane-changing behavior are screened through the mutual information screening method and the greedy forward search method, and the statistics, extraction, screening and other processes of these driving behavior features are placed on the server Therefore, there are no restrictions on the type and quantity of driving behavior characteristics, and the parameters of the online prediction model can be updated through online updating. When the sensor type is increased or the statistical data set is changed, the type of feature can be added or changed, and the present invention can be used for processing again. Therefore, the present invention solves the problems of unclear feature rationality and unexpandable models in common algorithms.

(2)：本车获取目标车辆在当前时刻的运动信息，该运动信息包括运动速度、运动加速度和运动方向中的至少一个。(2): The own vehicle obtains the movement information of the target vehicle at the current moment, and the movement information includes at least one of movement speed, movement acceleration and movement direction.

在一个可能的实现方式中，本车可以通过传感器采集目标车辆在当前时刻的运动信息。该传感器可以为摄像头或者雷达等。在另一个可能的实现方式中，本车通过车联网通信技术获取目标车辆在当前时刻的运动信息。相应的，本车获取目标车辆在当前时刻的运动信息的步骤可以为：本车向目标车辆发送第四获取请求，第四获取请求用于获取目标车辆在当前时刻的运动信息。目标车辆根据第四获取请求，向本车发送当前时刻的运动信息。本车接收目标车辆发送的该运动信息。在另一个可能的实现方式中，本车可以通过监控设备获取目标车辆在当前时刻的运动信息。相应的，本车获取目标车辆在当前时刻的运动信息的步骤可以为：本车向监控设备发送第五获取请求，第五获取请求携带目标车辆的车辆标识，接收监控设备根据第五获取请求发送的该目标车辆在当前时刻的运动信息。In a possible implementation manner, the own vehicle may collect movement information of the target vehicle at the current moment through a sensor. The sensor may be a camera or a radar. In another possible implementation, the own vehicle acquires the movement information of the target vehicle at the current moment through the Internet of Vehicles communication technology. Correspondingly, the step for the own vehicle to acquire the movement information of the target vehicle at the current moment may be: the own vehicle sends a fourth acquisition request to the target vehicle, and the fourth acquisition request is used to acquire the movement information of the target vehicle at the current moment. The target vehicle sends the current movement information to the own vehicle according to the fourth acquisition request. The own vehicle receives the movement information sent by the target vehicle. In another possible implementation manner, the ego vehicle may acquire the movement information of the target vehicle at the current moment through a monitoring device. Correspondingly, the step for the own vehicle to obtain the motion information of the target vehicle at the current moment may be: the own vehicle sends a fifth acquisition request to the monitoring device, the fifth acquisition request carries the vehicle identification of the target vehicle, and the receiving monitoring device sends The movement information of the target vehicle at the current moment.

(3)：本车根据该运动信息和多个驾驶行为特征，确定目标车辆在当前时刻的多个第一驾驶行为特征值。(3): The vehicle determines multiple first driving behavior feature values of the target vehicle at the current moment according to the motion information and multiple driving behavior features.

本车将该目标车辆在当前时刻的运动信息代入多个驾驶行为特征中，得到多个第一驾驶行为特征值。例如，驾驶行为特征为横向速度，则本车根据该目标车辆在当前时刻的运动信息，确定该目标车辆的横向速度值，将该横向速度值作为横向速度对应的第一驾驶行为特征值。The own vehicle substitutes the motion information of the target vehicle at the current moment into a plurality of driving behavior features to obtain a plurality of first driving behavior feature values. For example, if the driving behavior characteristic is lateral speed, then the own vehicle determines the lateral speed value of the target vehicle according to the motion information of the target vehicle at the current moment, and uses the lateral speed value as the first driving behavior characteristic value corresponding to the lateral speed.

步骤202：本车根据多个第一驾驶行为特征值，识别目标车辆在当前时刻是否有换道意图。Step 202: The own vehicle identifies whether the target vehicle intends to change lanes at the current moment according to a plurality of first driving behavior characteristic values.

在本步骤中，本车将目标车辆在不同时刻的换道意图构成隐马尔可夫过程，但是目标车辆在不同时刻的换道意图对本车来说是未知的。因此，本车对目标车辆的行为分类问题就转换为隐马尔可夫模型中的估计问题。本车定义该目标车辆i在时刻k的驾驶行为特征为B_i(k)。并且，定义b₁＝直行，b₂＝左换道，b₃＝右换道。其中，b₁是稳定的驾驶行为特征(包含停车情况)，b₂和b₃都是用于指示换道的驾驶行为特征。由于目标车辆在行驶过程中，车辆的驾驶行为可能发生改变。因此，每个驾驶行为特征之间可以互相转换，例如，一个典型的驾驶过程为b₁-＞b₂-＞b₃，则该驾驶过程指示的驾驶行为特征为直行-左换道-右换道。因此，本车定义P(B_i(t)＝b_j|0:t)为目标车辆i从时刻0到时刻t的运动信息，其在时刻t意欲实施的行为b_j的概率。因此，P(B_i(t)|0:t)和P(B_i(t+1)＝b_j|0:t)的关系可以用一个Markov(马尔可夫)矩阵A＝P(B_i(t+1)|B_i(t)来表示。相应的，本车根据贝叶斯网络前向迭代算法，在当前时刻t有如下公式六：In this step, the self-vehicle constructs a hidden Markov process with the lane-changing intentions of the target vehicle at different moments, but the lane-changing intentions of the target vehicle at different moments are unknown to the self-vehicle. Therefore, the behavior classification problem of the vehicle to the target vehicle is transformed into an estimation problem in the hidden Markov model. The own vehicle defines the driving behavior characteristic of the target vehicle i at time k as B_i (k). Also, it is defined that b₁ =going straight, b₂ =changing left, and b₃ =changing right. Among them, b₁ is a stable driving behavior feature (including parking situation), b₂ and b₃ are driving behavior features used to indicate lane change. As the target vehicle is moving, the driving behavior of the vehicle may change. Therefore, each driving behavior feature can be converted to each other. For example, a typical driving process is b₁ -> b₂ -> b₃ , then the driving behavior feature indicated by the driving process is going straight - changing lanes left - changing right road. Therefore, the self-vehicle defines P(B_i (t)=b_j |0:t) as the target vehicle i's motion information from time 0 to time t, and the probability of the behavior b_j it intends to implement at time t. Therefore, the relationship between P(B_i (t)|0:t) and P(B_i (t+1)=b_j |0:t) can be expressed by a Markov matrix A=P(B_i (t+1)|B_i (t). Correspondingly, according to the Bayesian network forward iterative algorithm, the vehicle has the following formula 6 at the current moment t:

公式六：P(B_i(k)|0:t)∝P(B_i(k)，y(0),…,y(t))∝P(y(k)|B_i(k)*P(B_i(k)|0:t-1)Formula 6: P(B_i (k)|0:t)∝P(B_i (k), y(0),…,y(t))∝P(y(k)|B_i (k)* P(B_i (k)|0:t-1)

其中，P(B_i(k)|0:t)为目标车辆在当前时刻的第一概率转移矩阵，P(y(k)|B_i(k)为目标车辆在当前时刻的观测矩阵，P(B_i(k)|0:t-1)为目标车辆在当前时刻的上一时刻的第一概率转移矩阵。相应的，本步骤可以通过以下步骤(1)至(5)实现，包括：Among them, P(B_i (k)|0:t) is the first probability transition matrix of the target vehicle at the current moment, P(y(k)|B_i (k) is the observation matrix of the target vehicle at the current moment, P (B_i (k)|0:t-1) is the first probability transition matrix of the target vehicle at the previous moment at the current moment. Correspondingly, this step can be realized through the following steps (1) to (5), including:

(1)：本车根据多个第一驾驶行为特征值，获取目标车辆在当前时刻的观测矩阵，观测矩阵用于指示目标车辆在不同驾驶行为的先验概率。(1): The vehicle obtains the observation matrix of the target vehicle at the current moment according to multiple first driving behavior eigenvalues, and the observation matrix is used to indicate the prior probability of the target vehicle in different driving behaviors.

本步骤可以通过以下步骤(1-1)至(1-3)实现，包括：This step can be achieved through the following steps (1-1) to (1-3), including:

(1-1)：本车根据多个第一驾驶行为特征值，确定目标车辆的驾驶行为特征向量。(1-1): The vehicle determines the driving behavior feature vector of the target vehicle according to a plurality of first driving behavior feature values.

本车直接将多个第一驾驶行为特征值组成驾驶行为特征向量。The vehicle directly forms a plurality of first driving behavior feature values into a driving behavior feature vector.

(1-2)：本车获取目标车辆在当前时刻之前的预设时长内的平均驾驶行为特征向量，平均驾驶行为特征向量包括多个平均驾驶行为特征值，多个平均驾驶行为特征值与多个驾驶行为特征一一对应。(1-2): The vehicle obtains the average driving behavior feature vector of the target vehicle within the preset time period before the current moment. The average driving behavior feature vector includes multiple average driving behavior feature values, multiple average driving behavior feature values and multiple One-to-one correspondence of driving behavior characteristics.

由于本车在行驶过程中实时获取目标车辆在当前时刻的多个第一驾驶行为特征值，基于多个第一驾驶行为特征值，识别目标车辆在当前时刻是否有换道意图。因此，本车中存储有当前时刻之前的多个时刻的第二驾驶行为特征值。相应的，本步骤可以为：本车获取目标车辆在当前时刻之前的预设时长内的每个时刻的多个第二驾驶行为特征值，根据每个时刻的多个第二驾驶行为特征值，确定多个平均驾驶行为特征值，将多个平均驾驶行为特征值组成平均驾驶行为向量。Since the vehicle acquires multiple first driving behavior characteristic values of the target vehicle at the current moment in real time during driving, based on the multiple first driving behavior characteristic values, it is identified whether the target vehicle has an intention to change lanes at the current moment. Therefore, the second driving behavior characteristic values at multiple times before the current time are stored in the own vehicle. Correspondingly, this step may be as follows: the vehicle acquires multiple second driving behavior characteristic values of the target vehicle at each moment within the preset time before the current moment, and according to the plurality of second driving behavior characteristic values at each moment, A plurality of average driving behavior characteristic values are determined, and the plurality of average driving behavior characteristic values are formed into an average driving behavior vector.

预设时长可以根据需要进行设置并更改，在本公开实施例中，对预设时长不作具体限定；例如，预设时长可以为10s或者30s等。The preset duration can be set and changed as required, and in the embodiment of the present disclosure, the preset duration is not specifically limited; for example, the preset duration can be 10s or 30s.

(1-3)：本车根据该驾驶行为特征向量和该平均驾驶行为特征向量，通过预设的高斯判别公式确定目标车辆在当前时刻的观测矩阵。(1-3): According to the driving behavior eigenvector and the average driving behavior eigenvector, the vehicle determines the observation matrix of the target vehicle at the current moment through the preset Gaussian discriminant formula.

本车根据该驾驶行为特征向量和该平均驾驶行为特征向量，通过以下公式七，确定目标车辆在当前时刻的观测矩阵。According to the driving behavior eigenvector and the average driving behavior eigenvector, the vehicle determines the observation matrix of the target vehicle at the current moment through the following formula 7.

公式七：Formula seven:

其中，P(y(k)|B_i(k))为观测矩阵，F(k)为该驾驶行为特征向量，μ为该平均驾驶行为特征向量，n为多个驾驶行为特征的数量，∑为协方差矩阵，∑_ii＝Var(f_i)，∑_ij＝cov(f_i,f_j)，|∑|为∑的行列式。Among them, P(y(k)|B_i (k)) is the observation matrix, F(k) is the driving behavior feature vector, μ is the average driving behavior feature vector, n is the number of multiple driving behavior features, ∑ is the covariance matrix, ∑_ii =Var(f_i ), ∑_ij =cov(f_i , f_j ), and |∑| is the determinant of ∑.

(2)：本车获取目标车辆在当前时刻之前的上一时刻的第一概率转移矩阵，第一概率转移矩阵用于指示目标车辆在不同驾驶行为之间的转移概率。(2): The vehicle acquires the first probability transition matrix of the target vehicle at the previous moment before the current moment, and the first probability transition matrix is used to indicate the transition probability of the target vehicle between different driving behaviors.

由于本车在行驶过程中实时获取目标车辆在当前时刻的多个第一驾驶行为特征值，基于多个第一驾驶行为特征值，识别目标车辆在当前时刻是否有换道意图。因此，本车中存储有当前时刻之前的上一时刻的第一概率转移矩阵。在本步骤中，本车直接获取已存储的在当前时刻之前的上一时刻的第一概率转移矩阵。Since the vehicle acquires multiple first driving behavior characteristic values of the target vehicle at the current moment in real time during driving, based on the multiple first driving behavior characteristic values, it is identified whether the target vehicle has an intention to change lanes at the current moment. Therefore, the first probability transition matrix at the previous moment before the current moment is stored in the own vehicle. In this step, the own vehicle directly obtains the stored first probability transition matrix at the previous moment before the current moment.

(3)：本车根据第一概率转移矩阵和该观测矩阵，确定目标车辆在当前时刻的第二概率转移矩阵。(3): The vehicle determines the second probability transition matrix of the target vehicle at the current moment according to the first probability transition matrix and the observation matrix.

本车将第一概率转移矩阵与该观测矩阵相乘，得到目标车辆在当前时刻的第二概率转移矩阵。本车确定出目标车辆在当前时刻的第二概率转移矩阵之后，存储第二概率转移矩阵，以便于后续基于第二概率转移矩阵，确定目标车辆在当前时刻的下一时刻的第三概率转移矩阵。The own vehicle multiplies the first probability transition matrix with the observation matrix to obtain the second probability transition matrix of the target vehicle at the current moment. After the vehicle determines the second probability transition matrix of the target vehicle at the current moment, it stores the second probability transition matrix, so as to determine the third probability transition matrix of the target vehicle at the next moment of the current moment based on the second probability transition matrix .

(4)：本车根据第二概率转移矩阵，确定目标车辆在当前时刻的换道概率，换道概率用于指示目标车辆进行换道的概率。(4): The vehicle determines the lane-changing probability of the target vehicle at the current moment according to the second probability transition matrix, and the lane-changing probability is used to indicate the probability of the target vehicle changing lanes.

本车根据第二概率转移矩阵，通过以下公式八，确定目标车辆在当前时刻的换道概率。According to the second probability transition matrix, the own vehicle determines the lane-changing probability of the target vehicle at the current moment through the following formula 8.

公式八：Formula eight:

其中，A_pq表示s概率转移矩阵的第p行第q列。I(·)是指示函数，用于统计括号中事件发生的次数。Among them, A_pq represents the pth row and qth column of the s probability transition matrix. I(·) is an indicator function used to count the number of occurrences of events in parentheses.

(5)：当该换道概率大于预设概率时，本车识别出目标车辆在当前时刻有换道意图。(5): When the lane-changing probability is greater than the preset probability, the vehicle recognizes that the target vehicle has a lane-changing intention at the current moment.

当该换道概率不大于预设概率时，本车识别出目标车辆在当前时刻没有换道意图，继续执行以上步骤201和202，直到识别出目标车辆有换道意图时，执行步骤203。其中，预设概率可以根据需要进行设置并更改，在本公开实施例中，对预设概率不作具体限定。例如，预设概率为0.8或者0.7等。When the lane-changing probability is not greater than the preset probability, the vehicle recognizes that the target vehicle has no lane-changing intention at the current moment, and continues to perform the above steps 201 and 202 until it recognizes that the target vehicle has a lane-changing intention, then executes step 203 . Wherein, the preset probability may be set and changed as required, and in the embodiment of the present disclosure, the preset probability is not specifically limited. For example, the preset probability is 0.8 or 0.7 and so on.

需要说明的是，当本车识别到目标车辆在当前时刻有换道意图时，本车先通过以下步骤203-206预测目标车辆在当前时刻的下一个时刻的运动速度。然后通过以下步骤207和208预测目标车辆的运动轨迹。当然，本车识别到目标车辆在当前时刻有换道意图时，也可以直接通过以下步骤207和208预测目标车辆的运动轨迹。It should be noted that when the self-vehicle recognizes that the target vehicle intends to change lanes at the current moment, the self-vehicle first predicts the moving speed of the target vehicle at the next moment at the current moment through the following steps 203-206. Then the trajectory of the target vehicle is predicted through the following steps 207 and 208 . Of course, when the own vehicle recognizes that the target vehicle intends to change lanes at the current moment, it can also directly predict the trajectory of the target vehicle through the following steps 207 and 208 .

步骤203：当识别到目标车辆在当前时刻有换道意图时，本车获取本车与目标车辆在当前时刻的相对速度。Step 203: When it is recognized that the target vehicle intends to change lanes at the current moment, the own vehicle obtains the relative speed between the own vehicle and the target vehicle at the current moment.

在一个可能的实现方式中，本车可以通过速度传感器获取本车与目标车辆在当前时刻的相对速度。在另一个可能的实现方式中，本车获取本车在当前时刻的第一运动速度，以及获取目标车辆在当前时刻的第二运动速度，根据第一运动速度和第二运动速度，确定本车与目标车辆在当前时刻的相对速度。In a possible implementation manner, the own vehicle can acquire the relative speed between the own vehicle and the target vehicle at the current moment through a speed sensor. In another possible implementation, the own vehicle obtains the first moving speed of the own vehicle at the current moment, and obtains the second moving speed of the target vehicle at the current moment, and determines the speed of the own vehicle according to the first moving speed and the second moving speed The relative speed of the target vehicle at the current moment.

步骤204：本车根据目标车辆在当前时刻的换道概率，确定目标车辆在当前时刻的直行概率，直行概率用于指示目标车辆进行直行的概率。Step 204: The own vehicle determines the straight-going probability of the target vehicle at the current moment according to the lane-changing probability of the target vehicle at the current moment, and the straight-going probability is used to indicate the probability of the target vehicle going straight.

本车在步骤202中的步骤(4)中已经确定出目标车辆在当前时刻的换道概率，在本步骤中直接确定1与该换道概率的差值，将该差值确定为该目标车辆在当前时刻的直行概率。The vehicle has determined the lane-changing probability of the target vehicle at the current moment in step (4) in step 202, and directly determines the difference between 1 and the lane-changing probability in this step, and determines the difference as the target vehicle Probability of going straight at the current moment.

步骤205：本车获取多个第一驾驶行为特征值中每个第一驾驶行为特征值在当前时刻的加权系数。Step 205: The vehicle acquires the weighting coefficient of each first driving behavior characteristic value at the current moment among the plurality of first driving behavior characteristic values.

本步骤可以通过以下步骤(1)和(2)实现，包括：This step can be achieved through the following steps (1) and (2), including:

(1)：本车获取目标车辆在当前时刻之前的上一个时刻的多个第二驾驶行为特征值，以及多个第二驾驶行为特征值中每个第二驾驶行为特征值在上一个时刻的加权系数。(1): The vehicle acquires a plurality of second driving behavior characteristic values of the target vehicle at the previous moment before the current moment, and a plurality of second driving behavior characteristic values of each second driving behavior characteristic value at the previous moment weighting factor.

由于本车在行驶过程中实时获取目标车辆在当前时刻的多个第一驾驶行为特征值，基于多个第一驾驶行为特征值，识别目标车辆在当前时刻是否有换道意图。因此，本车中存储有当前时刻之前的多个时刻的第二驾驶行为特征值。相应的，本车获取已存储的目标车辆在当前时刻之前的上一个时刻的多个第二驾驶行为特征值。并且，本车中也记录上一个时刻的每个第二驾驶行为特征值的加权系数。相应的，本车获取已存储的多个第二驾驶行为特征值中每个第二驾驶行为特征值在上一个时刻的加权系数。Since the vehicle acquires multiple first driving behavior characteristic values of the target vehicle at the current moment in real time during driving, based on the multiple first driving behavior characteristic values, it is identified whether the target vehicle has an intention to change lanes at the current moment. Therefore, the second driving behavior characteristic values at multiple times before the current time are stored in the own vehicle. Correspondingly, the self-vehicle acquires a plurality of second driving behavior characteristic values of the stored target vehicle at a previous moment before the current moment. Moreover, the weighting coefficient of each second driving behavior characteristic value at the previous moment is also recorded in the own vehicle. Correspondingly, the vehicle acquires the weighting coefficient of each second driving behavior characteristic value at the previous moment in the stored multiple second driving behavior characteristic values.

需要说明的是，如果当前时刻为进行换道意图识别的第一时刻时，本车中可能不存储目标车辆在当前时刻之前的上一个时刻的多个第二驾驶行为特征值以及多个第二驾驶行为特征值中每个第二驾驶行为特征值在上一个时刻的加权系数。此时本车可以设置目标车辆在当前时刻之前的上一个时刻的多个第二驾驶行为特征值可以为本车设置的任意值，且多个第二驾驶行为特征值中每个第二驾驶行为特征值在上一个时刻的加权系数也可以为本车设置的任意值。It should be noted that if the current moment is the first moment for lane-changing intention recognition, the vehicle may not store multiple second driving behavior characteristic values and multiple second driving behavior characteristic values of the target vehicle at the previous moment before the current moment. The weighting coefficient of each second driving behavior characteristic value in the driving behavior characteristic value at the previous moment. At this time, the vehicle can set a plurality of second driving behavior characteristic values of the target vehicle at the previous moment before the current moment, which can be any value set by the vehicle, and each second driving behavior in the plurality of second driving behavior characteristic values The weighting coefficient of the feature value at the previous moment can also be any value set by the vehicle.

(2)：本车根据每个第二驾驶行为特征值和每个第二驾驶行为特征值在上一个时刻的加权系数以及目标车辆在当前时刻的运动速度，通过预设算法，确定每个第一驾驶行为特征值在当前时刻的加权系数。(2): According to each second driving behavior characteristic value and the weighting coefficient of each second driving behavior characteristic value at the previous moment and the moving speed of the target vehicle at the current moment, the vehicle determines each second driving behavior characteristic value through a preset algorithm. A weighting coefficient of the driving behavior characteristic value at the current moment.

预设算法可以根据需要进行设置并更改，在本公开实施例中，对预设算法不作具体限定。例如，预设算法为在线参数自适应法。相应的，本步骤可以为：本车定义每个第一驾驶行为特征值在当前时刻的加权系统与每个第一驾驶行为特征值的关系如下公式九所示：The preset algorithm can be set and changed as required, and in the embodiment of the present disclosure, the preset algorithm is not specifically limited. For example, the preset algorithm is an online parameter adaptive method. Correspondingly, this step can be: the vehicle defines the relationship between the weighting system of each first driving behavior characteristic value at the current moment and each first driving behavior characteristic value as shown in formula 9 below:

公式九：Formula nine:

其中，β＝[q₁,…,q_N]^Τ，q₁,…,q_N为时变系数。利用的历史数据和递归最小乘法可以确定β的参数。定义目标车辆在时刻k-1时的β的估计为β(k-1)＝[f₁,…f_N]^Τ，目标车辆在时刻k时的β(k)的参数可以通过以下公式十计算得到。Among them, β=[q₁ ,…,q_N ]^Τ , q₁ ,…,q_N are time-varying coefficients. use The historical data and recursive least multiplication can determine the parameters of β. Define the estimate of β of the target vehicle at time k-1 as β(k-1)=[f₁ ,...f_N ]^Τ , the parameter of β(k) of the target vehicle at time k can be calculated by the following formula ten get.

公式十：Formula ten:

其中，λ为学习因子，且学习因子的取值可以根据需要进行设置并更改，在本公开实施例中，对学习因子的取值不作具体限定。x_i(k)为目标车辆在当前时刻的运动速度，g(x_i(k))为目标车辆在当前时刻的纵向运动方程，且g(x_i(k))可以通过以下公式十一确定。Wherein, λ is a learning factor, and the value of the learning factor can be set and changed as required, and in the embodiment of the present disclosure, the value of the learning factor is not specifically limited. x_i (k) is the moving speed of the target vehicle at the current moment, g(_xi (k)) is the longitudinal motion equation of the target vehicle at the current moment, and g(_xi (k)) can be determined by the following formula 11 .

公式十一：Formula Eleven:

其中，w₁、w₂、w₃、w₄分别为第一系数、第二系数、第三系数和第四系数。v_i(k)为目标车辆在时刻k的纵向速度，Δx_i(k)为目标车辆与前车在时刻k的相对速度、Δv_i(k-τ)为目标车辆与前车在时刻k-τ的相对速度，τ为时间间隔。需要说明的是，第一系数、第二系数、第三系数、第四系数和时间间隔都可以根据需要进行设置并更改，在本公开实施例中，对第一系数、第二系数、第三系数、第四系数和时间间隔不作具体限定。Wherein, w₁ , w₂ , w₃ , and w₄ are the first coefficient, the second coefficient, the third coefficient, and the fourth coefficient, respectively. v_i (k) is the longitudinal velocity of the target vehicle at time k,_Δxi (k) is the relative velocity of the target vehicle and the preceding vehicle at time k, Δv_i (k-τ) is the The relative velocity of τ, where τ is the time interval. It should be noted that the first coefficient, the second coefficient, the third coefficient, the fourth coefficient and the time interval can all be set and changed as required. In the embodiment of the present disclosure, the first coefficient, the second coefficient, the third coefficient The coefficient, the fourth coefficient and the time interval are not specifically limited.

步骤206：本车根据该相对速度、该直行概率、多个第一驾驶行为特征值以及每个第一驾驶行为特征值在当前时刻的加权系数，通过预设的纵向运动方程，确定目标车辆在当前时刻的下一个时刻的运动速度。Step 206: According to the relative speed, the probability of going straight, multiple first driving behavior characteristic values and the weighting coefficient of each first driving behavior characteristic value at the current moment, the target vehicle is determined by the preset longitudinal motion equation. The movement speed of the next moment at the current moment.

本车根据该相对速度、该直行概率、多个第一驾驶行为特征值以及每个第一驾驶行为特征值在当前时刻的加权系数，通过预设的纵向运动方程，通过以下公式十二，确定目标车辆在当前时刻的下一个时刻的运动速度According to the relative speed, the probability of going straight, a plurality of first driving behavior characteristic values and the weighting coefficient of each first driving behavior characteristic value at the current moment, through the preset longitudinal motion equation, through the following formula 12, determine The movement speed of the target vehicle at the next moment at the current moment

公式十二：Formula twelve:

其中，x_i为目标车辆在时刻i的运动速度，g(x_i)为系统原有的车辆纵向运动方程，B₁为目标车辆直行的概率，q_j为时变系数，f_j(x_i)为多个第一驾驶行为特征值。Among them, x_i is the speed of the target vehicle at time i, g(_xi ) is the original vehicle longitudinal motion equation of the system, B₁ is the probability of the target vehicle going straight, q_j is the time-varying coefficient, f_j (_xi ) is a plurality of first driving behavior feature values.

以上步骤203-206是预测目标车辆在当前时刻的下一个时刻的运动速度。本车预测出目标车辆的运动速度之后，还可以通过以下步骤207和208预测目标车辆的运动轨迹。The above steps 203-206 are to predict the moving speed of the target vehicle at the next moment at the current moment. After the own vehicle predicts the moving speed of the target vehicle, it can also predict the moving trajectory of the target vehicle through the following steps 207 and 208 .

步骤207：本车获取目标车辆与本车之间的相对距离。Step 207: The own vehicle acquires the relative distance between the target vehicle and the own vehicle.

在一个可能的实现方式中，本车可以通过距离传感器获取本车与目标车辆在当前时刻的相对距离。在另一个可能的实现方式中，本车获取本车在当前时刻的第一位置信息，以及获取目标车辆在当前时刻的第二位置信息，根据第一位置信息和第二位置信息，确定本车与目标车辆在当前时刻的相对距离。In a possible implementation manner, the own vehicle can acquire the relative distance between the own vehicle and the target vehicle at the current moment through a distance sensor. In another possible implementation, the own vehicle acquires the first position information of the own vehicle at the current moment, and obtains the second position information of the target vehicle at the current moment, and determines the position of the own vehicle according to the first position information and the second position information. The relative distance to the target vehicle at the current moment.

步骤208：本车根据该相对距离，通过预设的轨迹预测方程，预测目标车辆的运动轨迹。Step 208: According to the relative distance, the host vehicle predicts the trajectory of the target vehicle through a preset trajectory prediction equation.

本步骤可以通过以下步骤(1)至(3)实现，包括：This step can be achieved through the following steps (1) to (3), including:

(1)：本车获取预设的轨迹预测方程中的横向偏移值和纵向偏移值。(1): The vehicle obtains the lateral offset value and longitudinal offset value in the preset trajectory prediction equation.

横向偏移值用于指示预测出的运动轨迹与的实际运动轨迹的横向偏差值，纵向偏差值用于指示预测出的运动轨迹与实际运动轨迹的纵向偏差值。在一个可能的实现方式中，本车可以根据多个车辆的历史运动信息，训练预设的轨迹预测方程中的横向偏差值和纵向偏差值。在另一个可能的实现方式中，服务器根据多个车辆的历史运动信息，训练预设的轨迹预测方程中的横向偏差值和纵向偏差值，本车从服务器中获取该横向偏差值和纵向偏差值。在另一个可能的实现方式中，本车可以任意设置横向偏移值和纵向偏移值。The lateral offset value is used to indicate the lateral deviation value between the predicted motion trajectory and the actual motion trajectory, and the vertical deviation value is used to indicate the vertical deviation value between the predicted motion trajectory and the actual motion trajectory. In a possible implementation manner, the own vehicle can train the lateral deviation value and the longitudinal deviation value in the preset trajectory prediction equation according to the historical motion information of multiple vehicles. In another possible implementation, the server trains the lateral deviation value and longitudinal deviation value in the preset trajectory prediction equation according to the historical movement information of multiple vehicles, and the vehicle obtains the lateral deviation value and longitudinal deviation value from the server . In another possible implementation manner, the vehicle can arbitrarily set the lateral offset value and the longitudinal offset value.

(2)：本车获取预设的轨迹预测方程中的第一形状参数值和第二形状参数值。(2): The vehicle acquires the first shape parameter value and the second shape parameter value in the preset trajectory prediction equation.

第一形状参数值和第二形状参数值表征换道轨迹的形状，也即表征不同的换道风格。在一个可能的实现方式中，本车可以根据多个车辆的历史运动信息，训练第一形状参数值和第二形状参数值。在另一个可能的实现方式中，服务器根据多个车辆的历史运动信息，训练第一形状参数值和第二形状参数值，本车从服务器中获取该第一形状参数值和第二形状参数值。在另一个可能的实现方式中，本车可以任意设置第一形状参数值和第二形状参数值。The first shape parameter value and the second shape parameter value represent the shape of the lane-changing trajectory, that is, represent different lane-changing styles. In a possible implementation manner, the own vehicle may train the first shape parameter value and the second shape parameter value according to historical motion information of multiple vehicles. In another possible implementation, the server trains the first shape parameter value and the second shape parameter value according to the historical motion information of multiple vehicles, and the vehicle acquires the first shape parameter value and the second shape parameter value from the server . In another possible implementation manner, the vehicle may arbitrarily set the first shape parameter value and the second shape parameter value.

在本公开实施例中，本车基于不同驾驶人的历史大数据分析，可以得出参数第一形状参数值a和第二形状参数值b的分布。其中，a的分布服从高斯分布，也即a～N(μ_N,σ_N)，如图2-2所示。b的分布服从朗道分布，也即b～Landau(μ_L,c_L)，如图2-3所示。In the embodiment of the present disclosure, based on the historical big data analysis of different drivers, the vehicle can obtain the distribution of the first shape parameter value a and the second shape parameter value b. Among them, the distribution of a obeys the Gaussian distribution, that is, a～N(μ_N ,σ_N ), as shown in Figure 2-2. The distribution of b obeys the Landau distribution, that is, b～Landau(μ_L , c_L ), as shown in Figure 2-3.

其中，μ_N,σ_N和μ_L,c_L分别为高斯分布的参数和朗道分布的参数。μ_N,σ_N和μ_L,c_L的参数值都可以根据需要进行设置并更改，在本公开实施例中，对μ_N,σ_N和μ_L,c_L的参数值不作具体限定。在一个可能的实现方式中，本车可以利用贝叶斯参数学习方法更新a和b的分布。本车更新a和b的分布。相应的，本步骤可以通过以下步骤(2-1)和(2-2)实现，包括：Among them, μ_N , σ_N and μ_L , c_L are parameters of Gaussian distribution and Landau distribution respectively. The parameter values of μ_N , σ_N and μ_L , c_L can be set and changed as required, and in the embodiments of the present disclosure, the parameter values of μ_N , σ_N and μ_L , c_L are not specifically limited. In a possible implementation, the vehicle can use a Bayesian parameter learning method to update the distributions of a and b. The vehicle updates the distribution of a and b. Correspondingly, this step can be realized through the following steps (2-1) and (2-2), including:

(2-1)：本车获取第一形状参数的第一概率密度分布和第二形状参数的第二概率密度分布，第一概率密度分布包括第一形状参数的参数值和概率密度分布的对应关系，第二概率密度分布包括第二形状参数的参数值和概率密度分布的对应关系。(2-1): The vehicle obtains the first probability density distribution of the first shape parameter and the second probability density distribution of the second shape parameter, and the first probability density distribution includes the correspondence between the parameter value of the first shape parameter and the probability density distribution The second probability density distribution includes a corresponding relationship between the parameter value of the second shape parameter and the probability density distribution.

第一形状参数的第一概率密度分布和第二形状参数的第二概率密度分布为基于多个车辆的历史运动信息训练得到的。在一个可能的实现方式中，不同的位置对应不同的轨迹形状参数的参数值。相应的，本步骤可以为：本车根据目标车辆在当前时刻的位置信息，获取与位置信息对应的第一概率密度分布和第二概率密度分布。在另一个可能的实现方式中，不同的环境信息对应不同的换道轨迹形状的参数值。相应的，本步骤可以为：本车根据目标车辆在当前时刻的环境信息，获取与环境信息对应的第一概率密度分布和第二概率密度分布。其中，环境信息可以为温度、天气等信息。The first probability density distribution of the first shape parameter and the second probability density distribution of the second shape parameter are obtained through training based on historical motion information of a plurality of vehicles. In a possible implementation manner, different positions correspond to different parameter values of the trajectory shape parameter. Correspondingly, this step may be: the own vehicle acquires the first probability density distribution and the second probability density distribution corresponding to the position information according to the position information of the target vehicle at the current moment. In another possible implementation manner, different environmental information corresponds to different parameter values of the shape of the lane-changing trajectory. Correspondingly, this step may be: the own vehicle acquires the first probability density distribution and the second probability density distribution corresponding to the environment information according to the environment information of the target vehicle at the current moment. Wherein, the environmental information may be information such as temperature and weather.

(2-2)：本车从第一概率密度分布中选择最大概率密度值对应的第一形状参数值，以及从第二概率密度分布中选择最大概率密度值对应的第二形状参数值。(2-2): The vehicle selects the first shape parameter value corresponding to the maximum probability density value from the first probability density distribution, and selects the second shape parameter value corresponding to the maximum probability density value from the second probability density distribution.

本车假设第一形状参数值和第二形状参数值独立，则有如下公式十三：The car assumes that the value of the first shape parameter and the value of the second shape parameter are independent, then the following formula 13:

公式十三：、P(θ|o_0：k)∝P(θ|o_1：k-1)*P(ok|θ)Formula 13:, P(θ|o_0:k )∝P(θ|o_1:k-1 )*P(ok|θ)

其中，P(θ)为先验概率，θ＝(a,b)为模型参数的向量，P(o_k|θ)为每个参数对应的似然度，o_k为时刻k的观测状态，sse_a和sse_b分别为对应参数的标准差。在经过参数学习更新后分布中，选取概率密度最大的参数值作为换道轨迹的参数。相应的，第一形状参数值a和第二形状参数值b的参数值如下公式十四所示：Among them, P(θ) is the prior probability, θ=(a,b) is the vector of model parameters, P(o_k |θ) is the likelihood corresponding to each parameter, o_k is the observation state at time k, sse_a and sse_b are the standard deviations of the corresponding parameters, respectively. In the updated distribution after parameter learning, the parameter value with the largest probability density is selected as the parameter of the lane-changing trajectory. Correspondingly, the parameter values of the first shape parameter value a and the second shape parameter value b are shown in the following formula fourteen:

公式十四：a(k)＝argmax(P(a))Formula 14: a(k)=argmax(P(a))

b(k)＝argmax(P(b))b(k)=argmax(P(b))

(3)：本车根据横向偏移值、纵向偏移值、第一形状参数值、第二形状参数值和相对距离，通过预设的轨迹预测方程，预测目标车辆的运动轨迹。(3): According to the lateral offset value, longitudinal offset value, first shape parameter value, second shape parameter value and relative distance, the vehicle predicts the trajectory of the target vehicle through the preset trajectory prediction equation.

本车根据横向偏移值、纵向偏移值、第一形状参数值、第二形状参数值和相对距离，通过预设的轨迹预测方程，通过以下公式十五，预测目标车辆的运动轨迹。According to the lateral offset value, longitudinal offset value, first shape parameter value, second shape parameter value and relative distance, the vehicle predicts the trajectory of the target vehicle through the preset trajectory prediction equation and the following formula 15.

公式十五：Formula fifteen:

其中，L(x)为该相对距离中的纵向距离，x为该相对距离中的横向距离。a和b为第一形状参数值和第二形状参数值；c和d分别为横向偏移值和纵向偏移值。Wherein, L(x) is the longitudinal distance in the relative distance, and x is the transverse distance in the relative distance. a and b are the first shape parameter value and the second shape parameter value; c and d are the horizontal offset value and the vertical offset value respectively.

在本公开实施例中，在换道轨迹预测过程中，本发明首先通过统计真实交通流数据，获得换道轨迹模型的参数分布，而非单纯地对轨迹进行拟合。以这些参数的分布作为先验知识，对于某一次具体的换道行为，通过贝叶斯学习法更新参数，使得轨迹预测更加准确。相比业界常用的轨迹拟合预测，本发明既考虑了换道行为轨迹大数据统计结果的通用性，又考虑了当前换道过程中的特殊性，很好地兼顾了通用性和精确性。并且，本车预测出目标车辆的运动轨迹和运动速度之后，本车可以在运动轨迹中标注运动速度，从而得到时空轨迹运动轨迹。由于结合了运动速度，因此提高了预测出的运动轨迹的准确性。In the embodiments of the present disclosure, in the process of predicting the lane-changing trajectory, the present invention first obtains the parameter distribution of the lane-changing trajectory model by counting real traffic flow data, rather than simply fitting the trajectory. Using the distribution of these parameters as prior knowledge, for a specific lane-changing behavior, the parameters are updated by Bayesian learning method to make trajectory prediction more accurate. Compared with the trajectory fitting prediction commonly used in the industry, the present invention not only considers the versatility of the big data statistical results of the lane-changing behavior trajectory, but also considers the particularity of the current lane-changing process, taking into account both versatility and accuracy. Moreover, after the vehicle predicts the trajectory and velocity of the target vehicle, the vehicle can mark the velocity in the trajectory to obtain the trajectory of the space-time trajectory. The accuracy of the predicted motion trajectory is improved due to the incorporation of the motion velocity.

需要说明的是，该方法可以应用在无人驾驶车辆或者驾驶员驾驶车辆中。并且，本车预测出目标车辆的运动轨迹之后，可以执行以下步骤209和210实现对本车进行控制。当该方法应用在驾驶员驾驶车辆中时，本车预测出目标车辆的运动轨迹之后，也可以不执行步骤209和210，而是直接显示目标车辆的运动轨迹，以使驾驶员根据该运动轨迹控制本车行驶。It should be noted that this method can be applied to unmanned vehicles or vehicles driven by drivers. Moreover, after the own vehicle predicts the trajectory of the target vehicle, the following steps 209 and 210 may be executed to control the own vehicle. When this method is applied to a vehicle driven by a driver, after the vehicle predicts the trajectory of the target vehicle, it may not perform steps 209 and 210, but directly displays the trajectory of the target vehicle, so that the driver can follow the trajectory of the target vehicle. Control the driving of the vehicle.

步骤209：本车根据该运动轨迹，确定与该运动轨迹对应的驾驶策略，根据该驾驶策略，控制本车行驶。Step 209: The vehicle determines a driving strategy corresponding to the movement trajectory according to the movement trajectory, and controls the driving of the vehicle according to the driving strategy.

在一个可能的实现方式中，本车中存储不同的运动轨迹和驾驶策略的对应关系。相应的，本车基于该运动轨迹，确定与该运动轨迹对应的驾驶策略的步骤可以为：本车根据该运动轨迹，从运动轨迹和驾驶策略的对应关系中获取该运动轨迹对应的驾驶策略。In a possible implementation, the corresponding relationship between different motion trajectories and driving strategies is stored in the vehicle. Correspondingly, the step of determining the driving strategy corresponding to the movement trajectory based on the movement trajectory of the vehicle may be as follows: the vehicle obtains the driving strategy corresponding to the movement trajectory from the corresponding relationship between the movement trajectory and the driving strategy according to the movement trajectory.

在另一个可能的实现方式中，本车中存储不同的驾驶类型对应的驾驶策略的对应关系。相应的，本车基于该运动轨迹，确定与该运动轨迹对应的驾驶策略的步骤可以为：本车根据该运动轨迹，确定该运动轨迹对应的驾驶类型；根据该驾驶类型，从驾驶类型和驾驶策略的对应关系中获取该运动轨迹对应的驾驶策略。In another possible implementation manner, the corresponding relationship of driving strategies corresponding to different driving types is stored in the own vehicle. Correspondingly, the step of determining the driving strategy corresponding to the movement trajectory of the vehicle based on the movement trajectory may be as follows: the vehicle determines the driving type corresponding to the movement trajectory according to the movement trajectory; according to the driving type, from the driving type and driving The driving strategy corresponding to the motion trajectory is obtained from the corresponding relationship of the strategy.

需要说明的是，当该方法应用在驾驶员驾驶车辆中时，本车确定出与该运动轨迹对应的驾驶策略之后，直接基于该驾驶车辆，控制该车辆行驶。本车也可以显示该驾驶策略，以使驾驶员根据该驾驶策略控制该车辆行驶。It should be noted that when the method is applied to a vehicle driven by a driver, after the own vehicle determines the driving strategy corresponding to the motion trajectory, it directly controls the driving of the vehicle based on the driving vehicle. The vehicle can also display the driving strategy, so that the driver can control the driving of the vehicle according to the driving strategy.

在本公开实施例中，由于脱离了车辆运动模型，从而不受车辆运动模型的局限，提高了应用的广泛性。并且，由于本车基于目标车辆的运动信息，预测出了目标车辆的运动速度，从而提高了预测出的目标车辆的运动速度的准确性，进而提高了后续预测轨迹的准确性。In the embodiment of the present disclosure, since the vehicle motion model is broken away, it is not limited by the vehicle motion model, which improves the versatility of application. Moreover, since the self-vehicle predicts the moving speed of the target vehicle based on the moving information of the target vehicle, the accuracy of the predicted moving speed of the target vehicle is improved, thereby improving the accuracy of the subsequent predicted trajectory.

在本公开实施例中，获取目标车辆在当前时刻的多个第一驾驶行为特征值，根据多个第一驾驶行为特征值，识别目标车辆在当前时刻是否有换道意图。由于多个驾驶行为特征为根据多个车辆的历史运动信息得到的与驾驶行为相关的特征。因此，提高了识别的准确性。并且，当识别出目标车辆在当前时刻有换道意图时，能够根据两车之间的相对距离，准确预测目标车辆的运动轨迹。In an embodiment of the present disclosure, a plurality of first driving behavior characteristic values of the target vehicle at the current moment are acquired, and according to the plurality of first driving behavior characteristic values, it is identified whether the target vehicle has an intention to change lanes at the current moment. Since the multiple driving behavior features are features related to driving behavior obtained according to historical motion information of multiple vehicles. Therefore, the accuracy of recognition is improved. Moreover, when it is recognized that the target vehicle intends to change lanes at the current moment, the trajectory of the target vehicle can be accurately predicted according to the relative distance between the two vehicles.

在本公开实施例中通过上述换道意图识别的方法进行了模拟测试。使用驾驶模拟器模拟城市直行车道开展驾驶实验，得到左右换到有效数据各34组，总计59461个有效数据点，包括各时刻车辆的坐标、速度、转向灯等相关信息。采用本发明提出的算法，换道行为的识别率达到了84.08％，综合识别率达到89.35％。换道准确性如下表1所示：In the embodiment of the present disclosure, a simulation test is carried out through the above-mentioned method for lane-changing intention recognition. Using the driving simulator to simulate the city straight lane to carry out driving experiments, 34 sets of valid data for left and right shifts were obtained, with a total of 59,461 valid data points, including vehicle coordinates, speed, turn signals and other relevant information at each time. By adopting the algorithm proposed by the invention, the recognition rate of lane-changing behavior reaches 84.08%, and the comprehensive recognition rate reaches 89.35%. The lane change accuracy is shown in Table 1 below:

表1Table 1

本公开实施例还提高了换道意图识别的时效性。如图2-4所示。其中X轴沿车道线方向，Y轴垂直于车道线方向。T1表示换道行为开始到越过车道线的时间，T2表示本发明识别到换道行为的时间，ΔT表示二者时间差，ΔT越小，说明越能够及时识别换道行为，时效性越高。根据对以上测试数据统计，结果如下表2所示：The embodiments of the present disclosure also improve the timeliness of lane-changing intention identification. As shown in Figure 2-4. The X-axis is along the direction of the lane line, and the Y-axis is perpendicular to the direction of the lane line. T1 represents the time from the start of the lane change behavior to crossing the lane line, T2 represents the time when the present invention recognizes the lane change behavior, and ΔT represents the time difference between the two, and the smaller the ΔT, the more timely the lane change behavior can be identified and the higher the timeliness. According to the above test data statistics, the results are shown in Table 2 below:

从以上数据可以看到，由于采用了较为丰富的驾驶行为特征识别换道行为，本发明平均在换道行为启动0.62s即可识别换到行为，说明本发明对换道行为较为敏感，具有很高的时效性。From the above data, it can be seen that due to the use of relatively abundant driving behavior features to identify lane-changing behaviors, the present invention can recognize the lane-changing behavior after the start of the lane-changing behavior on average 0.62s, which shows that the present invention is more sensitive to lane-changing behaviors and has great advantages. High timeliness.

本公开实施例还提高了轨迹预测的精度，利用本公开实施例提供的轨迹预测对以上测试数据进行处理并预测，结果如图2-5所示。其中，x距离表示横向误差，y距离表示纵向误差。轨迹预测横向误差在1s时不高于0.25m，2s时不高于0.5m，纵向误差1s时不高于0.5m，2s时不高于1.6s，欧氏距离误差1s时不高于0.6m，2s时不高于1.8m。对于换道过程而言，更关注的是横向误差，对于平均车长4～5m、车宽2.5～3m的轿车而言，本发明对他车换道行为的横向预测具有很高的精度。The embodiment of the present disclosure also improves the accuracy of trajectory prediction. The above test data is processed and predicted by using the trajectory prediction provided by the embodiment of the present disclosure, and the results are shown in FIGS. 2-5 . Among them, the x distance represents the lateral error, and the y distance represents the longitudinal error. Trajectory prediction lateral error is not higher than 0.25m in 1s, not higher than 0.5m in 2s, longitudinal error is not higher than 0.5m in 1s, not higher than 1.6s in 2s, Euclidean distance error is not higher than 0.6m in 1s , not higher than 1.8m in 2s. For the lane changing process, more attention is paid to the lateral error. For a car with an average vehicle length of 4-5m and a vehicle width of 2.5-3m, the present invention has high accuracy in lateral prediction of other vehicles' lane-changing behavior.

本公开实施例提供了一种轨迹预测装置300，该装置应用在本车，用于执行上述轨迹预测方法中的本车执行的操作。参见图3-1，该装置包括：An embodiment of the present disclosure provides a trajectory prediction device 300, which is applied to an own vehicle and used to perform operations performed by the own vehicle in the above trajectory prediction method. See Figure 3-1, the device includes:

第一获取单元301，用于获取目标车辆在当前时刻的多个第一驾驶行为特征值，该目标车辆为本车周围的车辆，该多个第一驾驶行为特征值与多个驾驶行为特征一一对应，该多个驾驶行为特征为根据多个车辆的历史运动信息得到的与驾驶行为相关的特征，该驾驶行为包括换道或者直行；The first acquiring unit 301 is configured to acquire a plurality of first driving behavior characteristic values of the target vehicle at the current moment, the target vehicle is a vehicle around the vehicle, and the plurality of first driving behavior characteristic values are the same as the plurality of driving behavior characteristics One-to-one correspondence, the multiple driving behavior features are features related to driving behavior obtained according to the historical motion information of multiple vehicles, and the driving behavior includes changing lanes or going straight;

识别单元302，用于根据该多个第一驾驶行为特征值，识别该目标车辆在该当前时刻是否有换道意图；An identification unit 302, configured to identify whether the target vehicle has a lane-changing intention at the current moment according to the plurality of first driving behavior characteristic values;

第二获取单元303，用于当识别到该目标车辆在该当前时刻有换道意图时，获取该目标车辆与该本车之间的相对距离；The second acquisition unit 303 is configured to acquire the relative distance between the target vehicle and the vehicle when it is recognized that the target vehicle has an intention to change lanes at the current moment;

预测单元304，用于根据该相对距离，通过预设的轨迹预测方程，预测该目标车辆的运动轨迹。The prediction unit 304 is configured to predict the trajectory of the target vehicle according to the relative distance and through a preset trajectory prediction equation.

在一个可能的实现方式中，该识别单元302，还用于根据该多个第一驾驶行为特征值，获取该目标车辆在该当前时刻的观测矩阵，该观测矩阵用于指示该目标车辆在不同驾驶行为的先验概率；获取该目标车辆在该当前时刻之前的上一时刻的第一概率转移矩阵，该第一概率转移矩阵用于指示该目标车辆在不同驾驶行为之间的转移概率；根据该第一概率转移矩阵和该观测矩阵，确定该目标车辆在该当前时刻的第二概率转移矩阵；根据该第二概率转移矩阵，确定该目标车辆在该当前时刻的换道概率，该换道概率用于指示该目标车辆进行换道的概率；当该换道概率大于预设概率时，识别出该目标车辆在当前时刻有换道意图。In a possible implementation manner, the identification unit 302 is further configured to acquire an observation matrix of the target vehicle at the current moment according to the plurality of first driving behavior characteristic values, and the observation matrix is used to indicate that the target vehicle is in different The prior probability of driving behavior; obtaining the first probability transition matrix of the target vehicle at the previous moment before the current moment, the first probability transition matrix is used to indicate the transition probability of the target vehicle between different driving behaviors; according to The first probability transition matrix and the observation matrix determine the second probability transition matrix of the target vehicle at the current moment; according to the second probability transition matrix, determine the lane change probability of the target vehicle at the current moment, the lane change The probability is used to indicate the probability of the target vehicle changing lanes; when the lane changing probability is greater than the preset probability, it is recognized that the target vehicle has a lane changing intention at the current moment.

在一个可能的实现方式中，该识别单元302，还用于根据该多个第一驾驶行为特征值，确定该目标车辆的驾驶行为特征向量；获取该目标车辆在当前时刻之前的预设时长内的平均驾驶行为特征向量，该平均驾驶行为特征向量包括多个平均驾驶行为特征值，该多个平均驾驶行为特征值与多个驾驶行为特征一一对应；根据该驾驶行为特征向量和该平均驾驶行为特征向量，通过预设的高斯判别公式确定该目标车辆在该当前时刻的观测矩阵。In a possible implementation, the identification unit 302 is further configured to determine the driving behavior feature vector of the target vehicle according to the plurality of first driving behavior feature values; The average driving behavior feature vector, the average driving behavior feature vector includes a plurality of average driving behavior feature values, the plurality of average driving behavior feature values are in one-to-one correspondence with a plurality of driving behavior features; according to the driving behavior feature vector and the average driving behavior Behavior feature vector, the observation matrix of the target vehicle at the current moment is determined through the preset Gaussian discriminant formula.

在一个可能的实现方式中，该预测单元304，还用于获取预设的轨迹预测方程中的横向偏移值和纵向偏移值；获取该预设的轨迹预测方程中的第一形状参数值和第二形状参数值；根据该横向偏移值、该纵向偏移值、该第一形状参数值、第二形状参数值和该相对距离，通过该预设的轨迹预测方程，预测该目标车辆的运动轨迹。In a possible implementation manner, the predicting unit 304 is also used to obtain the lateral offset value and the vertical offset value in the preset trajectory prediction equation; obtain the first shape parameter value in the preset trajectory prediction equation and the second shape parameter value; according to the lateral offset value, the longitudinal offset value, the first shape parameter value, the second shape parameter value and the relative distance, the target vehicle is predicted through the preset trajectory prediction equation motion track.

在一个可能的实现方式中，该预测单元304，还用于获取第一形状参数的第一概率密度分布和第二形状参数的第二概率密度分布，该第一概率密度分布包括该第一形状参数的参数值和概率密度分布的对应关系，该第二概率密度分布包括该第二形状参数的参数值和概率密度分布的对应关系；从该第一概率密度分布中选择最大概率密度值对应的该第一形状参数值，以及从该第二概率密度分布中选择最大概率密度值对应的该第二形状参数值。In a possible implementation manner, the predicting unit 304 is further configured to obtain a first probability density distribution of the first shape parameter and a second probability density distribution of the second shape parameter, the first probability density distribution including the first shape The corresponding relationship between the parameter value of the parameter and the probability density distribution, the second probability density distribution includes the corresponding relationship between the parameter value of the second shape parameter and the probability density distribution; select the corresponding maximum probability density value from the first probability density distribution The first shape parameter value, and the second shape parameter value corresponding to the maximum probability density value selected from the second probability density distribution.

在一个可能的实现方式中，该预测单元304，还用于根据该目标车辆在当前时刻的位置信息，获取与该位置信息对应的该第一概率密度分布和该第二概率密度分布；或者，In a possible implementation manner, the prediction unit 304 is further configured to acquire the first probability density distribution and the second probability density distribution corresponding to the position information according to the position information of the target vehicle at the current moment; or,

该预测单元304，还用于根据该目标车辆在当前时刻的环境信息，获取与该环境信息对应的该第一概率密度分布和该第二概率密度分布。The prediction unit 304 is further configured to acquire the first probability density distribution and the second probability density distribution corresponding to the environment information according to the environment information of the target vehicle at the current moment.

在一个可能的实现方式中，参见图3-2，该装置还包括：In a possible implementation, referring to FIG. 3-2, the device further includes:

第三获取单元305，用于获取该本车与该目标车辆在该当前时刻的相对速度；A third acquisition unit 305, configured to acquire the relative speed between the own vehicle and the target vehicle at the current moment;

第一确定单元306，用于根据该目标车辆在该当前时刻的换道概率，确定该目标车辆在该当前时刻的直行概率，该直行概率用于指示该目标车辆进行直行的概率；The first determining unit 306 is configured to determine the straight-going probability of the target vehicle at the current moment according to the lane-changing probability of the target vehicle at the current moment, and the straight-going probability is used to indicate the probability of the target vehicle going straight;

第四获取单元307，用于获取该多个第一驾驶行为特征值中每个第一驾驶行为特征值在该当前时刻的加权系数；A fourth obtaining unit 307, configured to obtain a weighting coefficient of each first driving behavior characteristic value in the plurality of first driving behavior characteristic values at the current moment;

第二确定单元308，用于根据该相对速度、该直行概率、该多个第一驾驶行为特征值以及该每个第一驾驶行为特征值在该当前时刻的加权系数，通过预设的纵向运动方程，确定该目标车辆在该当前时刻的下一个时刻的运动速度。The second determining unit 308 is configured to use a preset longitudinal motion according to the relative speed, the probability of going straight, the plurality of first driving behavior characteristic values and the weighting coefficient of each first driving behavior characteristic value at the current moment. Equation to determine the moving speed of the target vehicle at the next moment of the current moment.

在一个可能的实现方式中，该第四获取单元307，还用于获取该目标车辆在该当前时刻之前的上一个时刻的多个第二驾驶行为特征值，以及该多个第二驾驶行为特征值中每个第二驾驶行为特征值在该上一个时刻的加权系数；根据该每个第二驾驶行为特征值和该每个第二驾驶行为特征值在该上一个时刻的加权系数以及该目标车辆在该当前时刻的运动速度，通过预设算法，确定该每个第一驾驶行为特征值在该当前时刻的加权系数。In a possible implementation, the fourth acquiring unit 307 is also configured to acquire a plurality of second driving behavior characteristic values of the target vehicle at a previous moment before the current moment, and the plurality of second driving behavior characteristic values The weighting coefficient of each second driving behavior characteristic value in the value at the previous moment; according to each second driving behavior characteristic value and the weighting coefficient of each second driving behavior characteristic value at the previous moment and the target The moving speed of the vehicle at the current moment determines the weighting coefficient of each first driving behavior characteristic value at the current moment through a preset algorithm.

在一个可能的实现方式中，该第一获取单元301，还用于获取该多个驾驶行为特征，以及，获取该目标车辆在该当前时刻的运动信息，该运动信息包括运动速度、运动加速度和运动方向中的至少一个；根据该运动信息和该多个驾驶行为特征，确定该目标车辆在该当前时刻的多个第一驾驶行为特征值。In a possible implementation manner, the first acquiring unit 301 is also configured to acquire the plurality of driving behavior characteristics, and acquire motion information of the target vehicle at the current moment, the motion information includes motion speed, motion acceleration and At least one of the direction of movement; according to the movement information and the plurality of driving behavior characteristics, determine a plurality of first driving behavior characteristic values of the target vehicle at the current moment.

在一个可能的实现方式中，该第一获取单元301，还用于向服务器发送获取请求，该获取请求用于请求该服务器发送该多个驾驶行为特征；接收该服务器根据该获取请求发送的该多个驾驶行为特征，该多个驾驶行为特征为该服务器根据该多个车辆的历史运动信息得到的与该驾驶行为相关的特征。In a possible implementation, the first obtaining unit 301 is further configured to send an obtaining request to the server, where the obtaining request is used to request the server to send the plurality of driving behavior characteristics; A plurality of driving behavior characteristics, the plurality of driving behavior characteristics are characteristics related to the driving behavior obtained by the server according to the historical movement information of the plurality of vehicles.

在一个可能的实现方式中，该第一获取单元301，还用于获取该多个车辆的历史运动信息，该历史运动信息包括运动速度、运动加速度和运动方向中的至少一个；根据该多个车辆的历史运动信息，通过预设训练模型，确定该多个驾驶行为特征。In a possible implementation manner, the first acquiring unit 301 is also configured to acquire historical motion information of the multiple vehicles, the historical motion information includes at least one of motion speed, motion acceleration, and motion direction; according to the multiple The historical motion information of the vehicle is used to determine the multiple driving behavior characteristics through a preset training model.

在一个可能的实现方式中，该第一获取单元301，还用于根据该多个车辆中每个车辆的历史运动信息，分别确定该每个车辆在历史时刻的横向速度、线速度、横向加速度和该每个车辆与所在车道线之间的车道线横向偏差，该历史时刻为该历史运动信息对应的时刻；获取该每个车辆在该历史时刻之前的至少一个预设时长内的统计信息，该统计信息包括均值、标准差、变异系数、均方根、香农熵、对数能量熵和正态熵中的至少一个；根据该每个车辆在该历史时刻的横向速度、线速度、横向加速度和该每个车辆与所在车道线之间的车道线横向偏差，以及该每个车辆在该历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定该多个驾驶行为特征。In a possible implementation, the first acquisition unit 301 is further configured to determine the lateral velocity, linear velocity, and lateral acceleration of each vehicle at historical moments according to the historical motion information of each vehicle in the plurality of vehicles. and the lane line lateral deviation between each vehicle and its lane line, the historical moment is the moment corresponding to the historical motion information; obtain statistical information of each vehicle within at least one preset time period before the historical moment, The statistical information includes at least one of mean value, standard deviation, coefficient of variation, root mean square, Shannon entropy, logarithmic energy entropy and normal entropy; according to the lateral velocity, linear velocity, lateral acceleration of each vehicle at the historical moment and the lane line lateral deviation between each vehicle and the lane line where it is located, and the statistical information of each vehicle within at least one preset time period before the historical moment, and determine the plurality of driving behaviors through a preset training model feature.

在一个可能的实现方式中，该第一获取单元301，还用于根据该每个车辆在历史时刻的横向速度、线速度、横向加速度和该每个车辆与所在车道线之间的车道线横向偏差，以及该每个车辆在该历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定多个候选驾驶行为特征；从该多个候选驾驶行为特征中选择与该驾驶行为相关的多个驾驶行为特征。In a possible implementation manner, the first acquisition unit 301 is further configured to: Deviation, and the statistical information of each vehicle in at least one preset time period before the historical moment, determine a plurality of candidate driving behavior characteristics through a preset training model; select from the plurality of candidate driving behavior characteristics and the driving behavior Behavior-related multiple driving behavior characteristics.

在一个可能的实现方式中，该第一获取单元301，还用于通过预设相关度算法，分别确定该多个候选驾驶行为特征中每个候选驾驶行为特征与该驾驶行为之间的相关度；根据该每个候选驾驶行为特征与该换道行为之间的相关度，从该多个候选驾驶行为特征中选择相关度满足预设条件的多个驾驶行为特征。In a possible implementation manner, the first acquisition unit 301 is further configured to respectively determine the correlation between each candidate driving behavior feature and the driving behavior in the plurality of candidate driving behavior features through a preset correlation algorithm ; According to the correlation between each candidate driving behavior feature and the lane-changing behavior, select a plurality of driving behavior features whose correlation meets a preset condition from the plurality of candidate driving behavior features.

在一个可能的实现方式中，参见图3-3，该装置还包括：In a possible implementation, referring to FIG. 3-3, the device further includes:

控制单元309，用于根据该运动轨迹，确定与该运动轨迹对应的驾驶策略；根据该驾驶策略，控制该本车行驶。The control unit 309 is configured to determine a driving strategy corresponding to the movement trajectory according to the movement trajectory; and control the driving of the own vehicle according to the driving strategy.

在本公开实施例中，获取目标车辆在当前时刻的多个第一驾驶行为特征值，根据多个第一驾驶行为特征值，识别目标车辆在当前时刻是否有换道意图。由于多个驾驶行为特征为根据多个车辆的历史运动信息得到的与驾驶行为相关的特征。因此，提高了识别的准确性。并且，当识别出目标车辆在当前时刻有换道意图时，能够根据两车之间的相对距离，准确预测目标车辆的运动轨迹。In an embodiment of the present disclosure, a plurality of first driving behavior characteristic values of the target vehicle at the current moment are acquired, and according to the plurality of first driving behavior characteristic values, it is identified whether the target vehicle has an intention to change lanes at the current moment. Since the multiple driving behavior features are features related to driving behavior obtained according to historical motion information of multiple vehicles. Therefore, the accuracy of recognition is improved. Moreover, when it is recognized that the target vehicle intends to change lanes at the current moment, the trajectory of the target vehicle can be accurately predicted according to the relative distance between the two vehicles.

本公开实施例提供了一种轨迹预测装置400，该装置应用在服务器，用于执行上述轨迹预测方法中的服务器执行的操作。参见图4-1，该装置包括：An embodiment of the present disclosure provides a trajectory prediction apparatus 400, which is applied to a server and configured to execute the operations performed by the server in the above trajectory prediction method. See Figure 4-1, the device includes:

第五获取单元401，用于获取多个车辆的历史运动信息，该历史运动信息包括运动速度、运动加速度和运动方向中的至少一个；The fifth acquisition unit 401 is configured to acquire historical movement information of multiple vehicles, the historical movement information including at least one of movement speed, movement acceleration and movement direction;

第三确定单元402，用于根据该多个车辆的历史运动信息，通过预设训练模型，确定多个驾驶行为特征，该多个驾驶行为特征为与驾驶行为相关的特征，该驾驶行为包括换道或者直行。The third determining unit 402 is configured to determine a plurality of driving behavior characteristics through a preset training model according to the historical motion information of the plurality of vehicles, the plurality of driving behavior characteristics are characteristics related to driving behavior, and the driving behavior includes changing road or go straight.

在一个可能的实现方式中，该第三确定单元402，还用于根据该多个车辆中每个车辆的历史运动信息，分别确定该每个车辆在历史时刻的横向速度、线速度、横向加速度和该每个车辆与所在车道线之间的车道线横向偏差，该历史时刻为该历史运动信息对应的时刻；获取该每个车辆在该历史时刻之前的至少一个预设时长内的统计信息，该统计信息包括均值、标准差、变异系数、均方根、香农熵、对数能量熵和正态熵中的至少一个；根据该每个车辆在该历史时刻的横向速度、线速度、横向加速度和该每个车辆与所在车道线之间的车道线横向偏差，以及该每个车辆在该历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定该多个驾驶行为特征。In a possible implementation manner, the third determination unit 402 is further configured to determine the lateral velocity, linear velocity, and lateral acceleration of each vehicle at historical moments according to the historical motion information of each vehicle in the plurality of vehicles. and the lane line lateral deviation between each vehicle and its lane line, the historical moment is the moment corresponding to the historical motion information; obtain statistical information of each vehicle within at least one preset time period before the historical moment, The statistical information includes at least one of mean value, standard deviation, coefficient of variation, root mean square, Shannon entropy, logarithmic energy entropy and normal entropy; according to the lateral velocity, linear velocity, lateral acceleration of each vehicle at the historical moment and the lane line lateral deviation between each vehicle and the lane line where it is located, and the statistical information of each vehicle within at least one preset time period before the historical moment, and determine the plurality of driving behaviors through a preset training model feature.

在一个可能的实现方式中，该第三确定单元402，还用于根据该每个车辆在历史时刻的横向速度、线速度、横向加速度和该每个车辆与所在车道线之间的车道线横向偏差，以及该每个车辆在该历史时刻之前的至少一个预设时长内的统计信息，通过预设训练模型，确定多个候选驾驶行为特征；从该多个候选驾驶行为特征中选择与该驾驶行为相关的多个驾驶行为特征。In a possible implementation, the third determination unit 402 is further configured to: Deviation, and the statistical information of each vehicle in at least one preset time period before the historical moment, determine a plurality of candidate driving behavior characteristics through a preset training model; select from the plurality of candidate driving behavior characteristics and the driving behavior Behavior-related multiple driving behavior characteristics.

在一个可能的实现方式中，该第三确定单元402，还用于通过预设相关度算法，分别确定该多个候选驾驶行为特征中每个候选驾驶行为特征与该驾驶行为之间的相关度；根据该每个候选驾驶行为特征与该换道行为之间的相关度，从该多个候选驾驶行为特征中选择相关度满足预设条件的多个驾驶行为特征。In a possible implementation manner, the third determination unit 402 is further configured to respectively determine the correlation between each candidate driving behavior feature and the driving behavior in the plurality of candidate driving behavior features through a preset correlation algorithm ; According to the correlation between each candidate driving behavior feature and the lane-changing behavior, select a plurality of driving behavior features whose correlation meets a preset condition from the plurality of candidate driving behavior features.

在一个可能的实现方式中，参见图4-2，该装置还包括：In a possible implementation, referring to FIG. 4-2, the device further includes:

接收单元403，用于接收车辆发送的获取请求，该获取请求用于请求该服务器发送该多个驾驶行为特征；The receiving unit 403 is configured to receive an acquisition request sent by the vehicle, and the acquisition request is used to request the server to send the plurality of driving behavior characteristics;

发送单元404，用于根据该获取请求，向该车辆发送该多个驾驶行为特征。The sending unit 404 is configured to send the plurality of driving behavior features to the vehicle according to the acquisition request.

在本公开实施例中，将这些驾驶行为特征的统计、提取、筛选等过程放到服务器中进行离线处理，因此对驾驶行为特征的种类、数量都没有约束，可以通过在线更新的方式对在线预测模型参数进行更新。当传感器类型增加或统计数据集发生变化时，均可增加或改变特征的类型重新利用本发明进行处理。因此，本发明解决了常见算法中特征合理性不明确、模型不可扩展的问题。In the embodiment of the present disclosure, the processes of statistics, extraction, and screening of these driving behavior characteristics are placed in the server for offline processing, so there is no restriction on the type and quantity of driving behavior characteristics, and the online prediction can be made through online updating. The model parameters are updated. When the sensor type is increased or the statistical data set is changed, the type of feature can be added or changed, and the present invention can be used for processing again. Therefore, the present invention solves the problems of unclear feature rationality and unexpandable models in common algorithms.

需要说明的是：上述实施例提供的轨迹预测装置在轨迹预测时，仅以上述各功能模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能模块完成，即将设备的内部结构划分成不同的功能模块，以完成以上描述的全部或者部分功能。另外，上述实施例提供的轨迹预测装置与轨迹预测方法实施例属于同一构思，其具体实现过程详见方法实施例，这里不再赘述。It should be noted that: the trajectory prediction device provided by the above embodiment only uses the division of the above-mentioned functional modules as an example when predicting the trajectory. In practical applications, the above-mentioned function allocation can be completed by different functional modules according to needs. The internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the trajectory prediction device and the trajectory prediction method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process thereof is detailed in the method embodiments, and will not be repeated here.

本实施例提供了一种终端，该终端可以用于执行上述各个实施例中提供的轨迹预测方法。参见图5，该终端500包括：This embodiment provides a terminal, which can be used to execute the trajectory prediction methods provided in the foregoing embodiments. Referring to Figure 5, the terminal 500 includes:

终端500可以包括射频(Radio Frequency，RF)电路510、包括有一个或一个以上计算机可读存储介质的存储器520、输入单元530、显示单元540、传感器550、音频电路560、无线保真(Wireless Fidelity，WiFi)模块570、包括有一个或者一个以上处理核心的处理器580、以及电源590等部件。本领域技术人员可以理解，图5中示出的终端结构并不构成对终端的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。其中：The terminal 500 may include a radio frequency (Radio Frequency, RF) circuit 510, a memory 520 including one or more computer-readable storage media, an input unit 530, a display unit 540, a sensor 550, an audio circuit 560, a wireless fidelity (Wireless Fidelity , WiFi) module 570, a processor 580 including one or more processing cores, and a power supply 590 and other components. Those skilled in the art can understand that the terminal structure shown in FIG. 5 does not constitute a limitation on the terminal, and may include more or less components than those shown in the figure, or combine some components, or arrange different components. in:

RF电路510可用于收发信息或通话过程中，信号的接收和发送，特别地，将基站的下行信息接收后，交由一个或者一个以上处理器580处理；另外，将涉及上行的数据发送给基站。通常，RF电路510包括但不限于天线、至少一个放大器、调谐器、一个或多个振荡器、用户身份模块(SIM)卡、收发信机、耦合器、低噪声放大器(Low Noise Amplifier，LNA)、双工器等。此外，RF电路510还可以通过无线通信与网络和其他设备通信。所述无线通信可以使用任一通信标准或协议，包括但不限于全球移动通讯系统(Global System of Mobilecommunication，GSM)、通用分组无线服务(General Packet Radio Service，GPRS)、码分多址(Code Division Multiple Access，CDMA)、宽带码分多址(Wideband Code DivisionMultiple Access,WCDMA)、长期演进(Long Term Evolution,LTE)、电子邮件、短消息服务(Short Messaging Service，SMS)等。The RF circuit 510 can be used for sending and receiving information or receiving and sending signals during a call. In particular, after receiving the downlink information of the base station, it is handed over to one or more processors 580 for processing; in addition, the data related to the uplink is sent to the base station . Generally, the RF circuit 510 includes, but is not limited to, an antenna, at least one amplifier, a tuner, one or more oscillators, a subscriber identity module (SIM) card, a transceiver, a coupler, a low noise amplifier (Low Noise Amplifier, LNA) , duplexer, etc. In addition, RF circuitry 510 may also communicate with networks and other devices via wireless communications. The wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile Communication (Global System of Mobilecommunication, GSM), General Packet Radio Service (General Packet Radio Service, GPRS), Code Division Multiple Access (Code Division Multiple Access, CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), e-mail, Short Messaging Service (Short Messaging Service, SMS), etc.

存储器520可用于存储软件程序以及模块，处理器580通过运行存储在存储器520的软件程序以及模块，从而执行各种功能应用以及数据处理。存储器520可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需的应用程序(比如声音播放功能、图像播放功能等)等；存储数据区可存储根据终端500的使用所创建的数据(比如音频数据、电话本等)等。此外，存储器520可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他易失性固态存储器件。相应地，存储器520还可以包括存储器控制器，以提供处理器580和输入单元530对存储器520的访问。The memory 520 can be used to store software programs and modules, and the processor 580 executes various functional applications and data processing by running the software programs and modules stored in the memory 520 . The memory 520 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playback function, an image playback function, etc.); Data created using the terminal 500 (such as audio data, phone book, etc.) and the like. In addition, the memory 520 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage devices. Correspondingly, the memory 520 may further include a memory controller to provide access to the memory 520 by the processor 580 and the input unit 530 .

输入单元530可用于接收输入的数字或字符信息，以及产生与用户设置以及功能控制有关的键盘、鼠标、操作杆、光学或者轨迹球信号输入。具体地，输入单元530可包括触敏表面531以及其他输入设备532。触敏表面531，也称为触摸显示屏或者触控板，可收集用户在其上或附近的触摸操作(比如用户使用手指、触笔等任何适合的物体或附件在触敏表面531上或在触敏表面531附近的操作)，并根据预先设定的程式驱动相应的连接装置。可选的，触敏表面531可包括触摸检测装置和触摸控制器两个部分。其中，触摸检测装置检测用户的触摸方位，并检测触摸操作带来的信号，将信号传送给触摸控制器；触摸控制器从触摸检测装置上接收触摸信息，并将它转换成触点坐标，再送给处理器580，并能接收处理器580发来的命令并加以执行。此外，可以采用电阻式、电容式、红外线以及表面声波等多种类型实现触敏表面531。除了触敏表面531，输入单元530还可以包括其他输入设备532。具体地，其他输入设备532可以包括但不限于物理键盘、功能键(比如音量控制按键、开关按键等)、轨迹球、鼠标、操作杆等中的一种或多种。The input unit 530 can be used to receive input numbers or character information, and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control. Specifically, the input unit 530 may include a touch-sensitive surface 531 and other input devices 532 . The touch-sensitive surface 531, also referred to as a touch screen or a touchpad, can collect user touch operations on or near it (for example, the user uses any suitable object or accessory such as a finger, a stylus, etc. on the touch-sensitive surface 531 or on The operation near the touch-sensitive surface 531), and drive the corresponding connection device according to the preset program. Optionally, the touch-sensitive surface 531 may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch orientation, and detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into contact coordinates, and sends it to the to the processor 580, and can receive and execute commands sent by the processor 580. In addition, the touch-sensitive surface 531 can be implemented in various types such as resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch-sensitive surface 531 , the input unit 530 may also include other input devices 532 . Specifically, other input devices 532 may include but not limited to one or more of a physical keyboard, function keys (such as volume control keys, switch keys, etc.), trackball, mouse, joystick, and the like.

显示单元540可用于显示由用户输入的信息或提供给用户的信息以及终端500的各种图形用户接口，这些图形用户接口可以由图形、文本、图标、视频和其任意组合来构成。显示单元540可包括显示面板451，可选的，可以采用液晶显示器(Liquid CrystalDisplay，LCD)、有机发光二极管(Organic Light-Emitting Diode,OLED)等形式来配置显示面板451。进一步的，触敏表面531可覆盖显示面板451，当触敏表面531检测到在其上或附近的触摸操作后，传送给处理器580以确定触摸事件的类型，随后处理器580根据触摸事件的类型在显示面板451上提供相应的视觉输出。虽然在图5中，触敏表面531与显示面板451是作为两个独立的部件来实现输入和输入功能，但是在某些实施例中，可以将触敏表面531与显示面板451集成而实现输入和输出功能。The display unit 540 can be used to display information input by or provided to the user and various graphical user interfaces of the terminal 500. These graphical user interfaces can be composed of graphics, text, icons, videos and any combination thereof. The display unit 540 may include a display panel 451. Optionally, the display panel 451 may be configured in the form of a liquid crystal display (Liquid Crystal Display, LCD), an organic light-emitting diode (Organic Light-Emitting Diode, OLED), or the like. Further, the touch-sensitive surface 531 may cover the display panel 451, and when the touch-sensitive surface 531 detects a touch operation on or near it, the touch operation is sent to the processor 580 to determine the type of the touch event, and then the processor 580 determines the type of the touch event according to the type of the touch event. The type provides a corresponding visual output on the display panel 451 . Although in FIG. 5, the touch-sensitive surface 531 and the display panel 451 are used as two independent components to realize input and input functions, in some embodiments, the touch-sensitive surface 531 and the display panel 451 can be integrated to realize input. and output functions.

终端500还可包括至少一种传感器550，比如光传感器、运动传感器以及其他传感器。具体地，光传感器可包括环境光传感器及接近传感器，其中，环境光传感器可根据环境光线的明暗来调节显示面板451的亮度，接近传感器可在终端500移动到耳边时，关闭显示面板451和/或背光。作为运动传感器的一种，重力加速度传感器可检测各个方向上(一般为三轴)加速度的大小，静止时可检测出重力的大小及方向，可用于识别手机姿态的应用(比如横竖屏切换、相关游戏、磁力计姿态校准)、振动识别相关功能(比如计步器、敲击)等；至于终端500还可配置的陀螺仪、气压计、湿度计、温度计、红外线传感器等其他传感器，在此不再赘述。The terminal 500 may also include at least one sensor 550, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 451 according to the brightness of the ambient light, and the proximity sensor may turn off the display panel 451 and the display panel 451 when the terminal 500 moves to the ear. / or backlighting. As a kind of motion sensor, the gravitational acceleration sensor can detect the magnitude of acceleration in various directions (generally three axes), and can detect the magnitude and direction of gravity when it is stationary, and can be used for applications that recognize the attitude of mobile phones (such as horizontal and vertical screen switching, related games, magnetometer attitude calibration), vibration recognition-related functions (such as pedometer, tap), etc.; as for other sensors such as gyroscope, barometer, hygrometer, thermometer, infrared sensor, etc. Let me repeat.

音频电路560、扬声器561，传声器562可提供用户与终端500之间的音频接口。音频电路560可将接收到的音频数据转换后的电信号，传输到扬声器561，由扬声器561转换为声音信号输出；另一方面，传声器562将收集的声音信号转换为电信号，由音频电路560接收后转换为音频数据，再将音频数据输出处理器580处理后，经RF电路510以发送给比如另一终端，或者将音频数据输出至存储器520以便进一步处理。音频电路560还可能包括耳塞插孔，以提供外设耳机与终端500的通信。The audio circuit 560 , the speaker 561 and the microphone 562 can provide an audio interface between the user and the terminal 500 . The audio circuit 560 can transmit the electrical signal converted from the received audio data to the loudspeaker 561, and the loudspeaker 561 converts it into an audio signal output; After being received, it is converted into audio data, and then the audio data is processed by the output processor 580, and then sent to another terminal through the RF circuit 510, or the audio data is output to the memory 520 for further processing. The audio circuit 560 may also include an earplug jack to provide communication between an external earphone and the terminal 500 .

WiFi属于短距离无线传输技术，终端500通过WiFi模块570可以帮助用户收发电子邮件、浏览网页和访问流式媒体等，它为用户提供了无线的宽带互联网访问。虽然图5示出了WiFi模块570，但是可以理解的是，其并不属于终端500的必须构成，完全可以根据需要在不改变发明的本质的范围内而省略。WiFi is a short-distance wireless transmission technology. The terminal 500 can help users send and receive e-mails, browse web pages, and access streaming media through the WiFi module 570. It provides users with wireless broadband Internet access. Although FIG. 5 shows a WiFi module 570, it can be understood that it is not an essential component of the terminal 500, and can be completely omitted as required without changing the essence of the invention.

处理器580是终端500的控制中心，利用各种接口和线路连接整个手机的各个部分，通过运行或执行存储在存储器520内的软件程序和/或模块，以及调用存储在存储器520内的数据，执行终端500的各种功能和处理数据，从而对手机进行整体监控。可选的，处理器580可包括一个或多个处理核心；可选的，处理器580可集成应用处理器和调制解调处理器，其中，应用处理器主要处理操作系统、用户界面和应用程序等，调制解调处理器主要处理无线通信。可以理解的是，上述调制解调处理器也可以不集成到处理器580中。The processor 580 is the control center of the terminal 500, and uses various interfaces and lines to connect various parts of the entire mobile phone. By running or executing software programs and/or modules stored in the memory 520, and calling data stored in the memory 520, Execute various functions and process data of the terminal 500, so as to monitor the mobile phone as a whole. Optionally, the processor 580 may include one or more processing cores; optionally, the processor 580 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface and application programs etc., the modem processor mainly handles wireless communication. It can be understood that the foregoing modem processor may not be integrated into the processor 580 .

终端500还包括给各个部件供电的电源590(比如电池)，可选的，电源可以通过电源管理系统与处理器580逻辑相连，从而通过电源管理系统实现管理充电、放电、以及功耗管理等功能。电源590还可以包括一个或一个以上的直流或交流电源、再充电系统、电源故障检测电路、电源转换器或者逆变器、电源状态指示器等任意组件。The terminal 500 also includes a power supply 590 (such as a battery) for supplying power to various components. Optionally, the power supply can be logically connected to the processor 580 through the power management system, so that functions such as charging, discharging, and power consumption management can be implemented through the power management system. . The power supply 590 may also include one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

尽管未示出，终端500还可以包括摄像头、蓝牙模块等，在此不再赘述。具体在本实施例中，终端的显示单元是触摸屏显示器，终端还包括有存储器，以及一个或者一个以上的程序，其中一个或者一个以上程序存储于存储器中，且经配置以由一个或者一个以上处理器执行。所述一个或者一个以上程序包含用于执行以下上述轨迹预测方法。Although not shown, the terminal 500 may also include a camera, a Bluetooth module, etc., which will not be repeated here. Specifically, in this embodiment, the display unit of the terminal is a touch screen display, and the terminal also includes a memory, and one or more programs, wherein one or more programs are stored in the memory, and are configured to be processed by one or more device execution. The one or more programs are included for performing the following trajectory prediction method described above.

图6是根据一示例性实施例示出的一种轨迹预测装置600的框图。例如，装置600可以被提供为一服务器。参照图6，装置600包括处理组件622，其进一步包括一个或多个处理器，以及由存储器632所代表的存储器资源，用于存储可由处理部件622的执行的指令，例如应用程序。存储器632中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外，处理组件622被配置为执行指令，以执行上述轨迹预测方法。Fig. 6 is a block diagram of a trajectory prediction device 600 according to an exemplary embodiment. For example, the apparatus 600 may be provided as a server. Referring to FIG. 6 , apparatus 600 includes processing component 622 , which further includes one or more processors, and a memory resource represented by memory 632 for storing instructions executable by processing component 622 , such as application programs. The application program stored in memory 632 may include one or more modules each corresponding to a set of instructions. In addition, the processing component 622 is configured to execute instructions to perform the above trajectory prediction method.

装置600还可以包括一个电源组件626被配置为执行装置600的电源管理，一个有线或无线网络接口650被配置为将装置600连接到网络，和一个输入输出(I/O)接口658。装置600可以操作基于存储在存储器632的操作系统，例如Windows Server^TM，Mac OS X^TM，Unix^TM,Linux^TM，FreeBSD^TM或类似。Device 600 may also include a power component 626 configured to perform power management of device 600 , a wired or wireless network interface 650 configured to connect device 600 to a network, and an input-output (I/O) interface 658 . The apparatus 600 can operate based on an operating system stored in the memory 632, such as Windows Server^™ , Mac OS X^™ , Unix^™ , Linux^™ , FreeBSD^™ or the like.

本公开实施例提供了一种计算机可读存储介质，所述存储介质包括指令，当其在计算机上运行时，使得计算机执行上述轨迹预测方法。An embodiment of the present disclosure provides a computer-readable storage medium, and the storage medium includes instructions that, when run on a computer, cause the computer to execute the above trajectory prediction method.

本领域普通技术人员可以理解实现上述实施例的全部或部分步骤可以通过硬件来完成，也可以通过程序来指令相关的硬件完成，所述的程序可以存储于一种计算机可读存储介质中，上述提到的存储介质可以是只读存储器，磁盘或光盘等。Those of ordinary skill in the art can understand that all or part of the steps for implementing the above embodiments can be completed by hardware, and can also be completed by instructing related hardware through a program. The program can be stored in a computer-readable storage medium. The above-mentioned The storage medium mentioned may be a read-only memory, a magnetic disk or an optical disk, and the like.

以上所述仅为本公开的可选实施例，并不用以限制本公开，凡在本公开的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本公开的保护范围之内。The above descriptions are only optional embodiments of the present disclosure, and are not intended to limit the present disclosure. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the protection of the present disclosure. within range.

Claims (44)

1. A trajectory prediction method, characterized in that the method comprises:

the method comprises the steps that a plurality of first driving behavior characteristic values of a target vehicle at the current moment are obtained, the target vehicle is a vehicle around the target vehicle, the first driving behavior characteristic values correspond to driving behavior characteristics in a one-to-one mode, the driving behavior characteristics are characteristics related to driving behaviors obtained according to historical motion information of the vehicles, and the driving behaviors comprise lane changing or straight driving;

identifying whether the target vehicle has a lane changing intention at the current moment according to the plurality of first driving behavior characteristic values;

when recognizing that the target vehicle has a lane changing intention at the current moment, acquiring a relative distance between the target vehicle and the host vehicle;

and predicting the motion track of the target vehicle according to the relative distance and a preset track prediction equation.

2. The method of claim 1, wherein identifying whether the target vehicle has an intent to switch lanes at the current time based on the plurality of first driving behavior feature values comprises:

acquiring an observation matrix of the target vehicle at the current moment according to the plurality of first driving behavior characteristic values, wherein the observation matrix is used for indicating the prior probability of different driving behaviors of the target vehicle;

obtaining a first probability transition matrix of the target vehicle at a previous time before the current time, the first probability transition matrix being used to indicate transition probabilities of the target vehicle between different driving behaviors;

determining a second probability transition matrix of the target vehicle at the current time according to the first probability transition matrix and the observation matrix;

determining the lane change probability of the target vehicle at the current moment according to the second probability transition matrix, wherein the lane change probability is used for indicating the probability of lane change of the target vehicle;

and when the lane changing probability is larger than the preset probability, recognizing that the target vehicle has the lane changing intention at the current moment.

3. The method according to claim 2, wherein the obtaining an observation matrix of the target vehicle at the current time based on the plurality of first driving behavior feature values comprises:

determining a driving behavior feature vector of the target vehicle according to the plurality of first driving behavior feature values;

acquiring an average driving behavior feature vector of the target vehicle within a preset time length before the current moment, wherein the average driving behavior feature vector comprises a plurality of average driving behavior feature values, and the average driving behavior feature values correspond to the driving behavior features one to one;

and determining an observation matrix of the target vehicle at the current moment through a preset Gaussian discrimination formula according to the driving behavior feature vector and the average driving behavior feature vector.

4. The method according to claim 1, wherein the predicting the motion trajectory of the target vehicle through a preset trajectory prediction equation according to the relative distance comprises:

acquiring a horizontal deviation value and a vertical deviation value in a preset track prediction equation;

acquiring a first shape parameter value and a second shape parameter value in the preset trajectory prediction equation;

and predicting the motion track of the target vehicle through the preset track prediction equation according to the transverse deviation value, the longitudinal deviation value, the first shape parameter value, the second shape parameter value and the relative distance.

5. The method of claim 4, wherein said obtaining the first shape parameter value and the second shape parameter value in the preset trajectory prediction equation comprises:

obtaining a first probability density distribution of a first shape parameter comprising a correspondence of a parameter value of the first shape parameter and a probability density distribution and a second probability density distribution of a second shape parameter comprising a correspondence of a parameter value of the second shape parameter and a probability density distribution;

selecting the first shape parameter value corresponding to a maximum probability density value from the first probability density distribution, and selecting the second shape parameter value corresponding to a maximum probability density value from the second probability density distribution.

6. The method of claim 5, wherein said obtaining a first probability density distribution of a first shape parameter and a second probability density distribution of a second shape parameter comprises:

acquiring the first probability density distribution and the second probability density distribution corresponding to the position information according to the position information of the target vehicle at the current moment; or,

and acquiring the first probability density distribution and the second probability density distribution corresponding to the environmental information according to the environmental information of the target vehicle at the current moment.

7. The method of claim 2, further comprising:

acquiring the relative speed of the vehicle and the target vehicle at the current moment;

determining the straight-going probability of the target vehicle at the current moment according to the lane changing probability of the target vehicle at the current moment, wherein the straight-going probability is used for indicating the probability that the target vehicle goes straight;

acquiring a weighting coefficient of each first driving behavior characteristic value in the plurality of first driving behavior characteristic values at the current moment;

and determining the movement speed of the target vehicle at the next moment of the current moment through a preset longitudinal movement equation according to the relative speed, the straight-ahead probability, the plurality of first driving behavior characteristic values and the weighting coefficient of each first driving behavior characteristic value at the current moment.

8. The method according to claim 7, wherein the obtaining a weighting coefficient for each of the plurality of first driving behavior feature values at the current time comprises:

acquiring a plurality of second driving behavior characteristic values of the target vehicle at a previous moment before the current moment, and a weighting coefficient of each second driving behavior characteristic value in the plurality of second driving behavior characteristic values at the previous moment;

and determining the weighting coefficient of each first driving behavior characteristic value at the current moment through a preset algorithm according to each second driving behavior characteristic value, the weighting coefficient of each second driving behavior characteristic value at the previous moment and the movement speed of the target vehicle at the current moment.

9. The method according to claim 1, wherein the obtaining a plurality of first driving behavior feature values of a target vehicle at a current time comprises:

acquiring a plurality of driving behavior characteristics, and acquiring motion information of the target vehicle at the current moment, wherein the motion information comprises at least one of motion speed, motion acceleration and motion direction;

and determining a plurality of first driving behavior characteristic values of the target vehicle at the current moment according to the motion information and the plurality of driving behavior characteristics.

10. The method of claim 9, wherein said obtaining the plurality of driving behavior characteristics comprises:

sending an acquisition request to a server, wherein the acquisition request is used for requesting the server to send the plurality of driving behavior characteristics;

and receiving the plurality of driving behavior characteristics sent by the server according to the acquisition request, wherein the plurality of driving behavior characteristics are characteristics related to the driving behaviors, which are obtained by the server according to the historical motion information of the plurality of vehicles.

11. The method of claim 9, wherein said obtaining the plurality of driving behavior characteristics comprises:

obtaining historical motion information of the plurality of vehicles, wherein the historical motion information comprises at least one of motion speed, motion acceleration and motion direction;

and determining the driving behavior characteristics through a preset training model according to the historical motion information of the vehicles.

12. The method of claim 11, wherein determining the plurality of driving behavior characteristics from historical motion information of the plurality of vehicles via a preset training model comprises:

according to historical motion information of each vehicle in the plurality of vehicles, respectively determining the lateral speed, the linear speed and the lateral acceleration of each vehicle at a historical moment and the lateral deviation of a lane line between each vehicle and the lane line, wherein the historical moment is a moment corresponding to the historical motion information;

acquiring statistical information of each vehicle in at least one preset time before the historical moment, wherein the statistical information comprises at least one of a mean value, a standard deviation, a variation coefficient, a root mean square, shannon entropy, a logarithmic energy entropy and a normal entropy;

and determining the plurality of driving behavior characteristics through a preset training model according to the transverse speed, the linear speed and the transverse acceleration of each vehicle at the historical moment, the transverse deviation of the lane line between each vehicle and the lane line where the vehicle is located, and statistical information of each vehicle in at least one preset time before the historical moment.

13. The method according to claim 12, wherein the determining the plurality of driving behavior characteristics through a preset training model according to the lateral speed, the linear speed, the lateral acceleration and the lane line lateral deviation of each vehicle from the lane line at the historical time, and the statistical information of each vehicle in at least one preset time period before the historical time comprises:

determining a plurality of candidate driving behavior characteristics through a preset training model according to the transverse speed, the linear speed and the transverse acceleration of each vehicle at the historical moment, the transverse deviation of a lane line between each vehicle and the lane line, and the statistical information of each vehicle in at least one preset time before the historical moment;

selecting a plurality of driving behavior features associated with the driving behavior from the plurality of candidate driving behavior features.

14. The method of claim 13, wherein the selecting a plurality of driving behavior features associated with the driving behavior from the plurality of candidate driving behavior features comprises:

respectively determining the correlation between each candidate driving behavior characteristic in the plurality of candidate driving behavior characteristics and the driving behavior through a preset correlation algorithm;

and selecting a plurality of driving behavior characteristics with correlation degrees meeting preset conditions from the plurality of candidate driving behavior characteristics according to the correlation degree between each candidate driving behavior characteristic and the lane changing behavior.

15. The method according to claim 1, wherein after predicting the movement locus of the target vehicle by a preset locus prediction equation according to the relative distance, the method further comprises:

determining a driving strategy corresponding to the motion trail according to the motion trail;

and controlling the vehicle to run according to the driving strategy.

16. A trajectory prediction method, characterized in that the method comprises:

acquiring historical motion information of a plurality of vehicles, wherein the historical motion information comprises at least one of motion speed, motion acceleration and motion direction;

determining a plurality of driving behavior characteristics through a preset training model according to the historical motion information of the vehicles, wherein the driving behavior characteristics are characteristics related to driving behaviors, and the driving behaviors comprise lane changing or straight driving.

17. The method of claim 16, wherein determining a plurality of driving behavior characteristics from historical motion information of the plurality of vehicles via a preset training model comprises:

according to historical motion information of each vehicle in the plurality of vehicles, respectively determining the lateral speed, the linear speed and the lateral acceleration of each vehicle at a historical moment and the lateral deviation of a lane line between each vehicle and the lane line, wherein the historical moment is a moment corresponding to the historical motion information;

acquiring statistical information of each vehicle in at least one preset time before the historical moment, wherein the statistical information comprises at least one of a mean value, a standard deviation, a variation coefficient, a root mean square, shannon entropy, a logarithmic energy entropy and a normal entropy;

and determining the plurality of driving behavior characteristics through a preset training model according to the transverse speed, the linear speed and the transverse acceleration of each vehicle at the historical moment, the transverse deviation of the lane line between each vehicle and the lane line where the vehicle is located, and statistical information of each vehicle in at least one preset time before the historical moment.

18. The method of claim 17, wherein the determining the plurality of driving behavior characteristics through a preset training model according to the lateral speed, the linear speed, the lateral acceleration and the lane line lateral deviation of each vehicle from the lane line at the historical time, and the statistical information of each vehicle in at least one preset time period before the historical time comprises:

determining a plurality of candidate driving behavior characteristics through a preset training model according to the transverse speed, the linear speed and the transverse acceleration of each vehicle at the historical moment, the transverse deviation of a lane line between each vehicle and the lane line, and the statistical information of each vehicle in at least one preset time before the historical moment;

selecting a plurality of driving behavior features associated with the driving behavior from the plurality of candidate driving behavior features.

19. The method of claim 18, wherein the selecting a plurality of driving behavior features associated with the driving behavior from the plurality of candidate driving behavior features comprises:

respectively determining the correlation between each candidate driving behavior characteristic in the plurality of candidate driving behavior characteristics and the driving behavior through a preset correlation algorithm;

and selecting a plurality of driving behavior characteristics with correlation degrees meeting preset conditions from the plurality of candidate driving behavior characteristics according to the correlation degree between each candidate driving behavior characteristic and the lane changing behavior.

20. The method of claim 16, wherein after determining a plurality of driving behavior characteristics from the historical motion information of the plurality of vehicles via a preset training model, the method further comprises:

receiving an acquisition request sent by a vehicle, wherein the acquisition request is used for requesting the server to send the plurality of driving behavior characteristics;

and sending the plurality of driving behavior characteristics to the vehicle according to the acquisition request.

21. A trajectory prediction device, characterized in that the device comprises:

the device comprises a first acquisition unit, a second acquisition unit and a control unit, wherein the first acquisition unit is used for acquiring a plurality of first driving behavior characteristic values of a target vehicle at the current moment, the target vehicle is a vehicle around the vehicle, the first driving behavior characteristic values correspond to a plurality of driving behavior characteristics one to one, the driving behavior characteristics are characteristics related to driving behaviors obtained according to historical motion information of the vehicles, and the driving behaviors comprise lane changing or straight driving;

the identification unit is used for identifying whether the target vehicle has a lane changing intention at the current moment according to the plurality of first driving behavior characteristic values;

a second acquisition unit, configured to acquire a relative distance between the target vehicle and the host vehicle when it is recognized that the target vehicle has a lane change intention at the current time;

and the prediction unit is used for predicting the motion trail of the target vehicle according to the relative distance and a preset trail prediction equation.

22. The apparatus of claim 21,

the identification unit is further configured to obtain an observation matrix of the target vehicle at the current moment according to the plurality of first driving behavior feature values, where the observation matrix is used to indicate prior probabilities of different driving behaviors of the target vehicle; obtaining a first probability transition matrix of the target vehicle at a previous time before the current time, the first probability transition matrix being used to indicate transition probabilities of the target vehicle between different driving behaviors; determining a second probability transition matrix of the target vehicle at the current time according to the first probability transition matrix and the observation matrix; determining the lane change probability of the target vehicle at the current moment according to the second probability transition matrix, wherein the lane change probability is used for indicating the probability of lane change of the target vehicle; and when the lane changing probability is larger than the preset probability, recognizing that the target vehicle has the lane changing intention at the current moment.

23. The apparatus of claim 22,

the identification unit is further used for determining a driving behavior feature vector of the target vehicle according to the plurality of first driving behavior feature values; acquiring an average driving behavior feature vector of the target vehicle within a preset time length before the current moment, wherein the average driving behavior feature vector comprises a plurality of average driving behavior feature values, and the average driving behavior feature values correspond to the driving behavior features one to one; and determining an observation matrix of the target vehicle at the current moment through a preset Gaussian discrimination formula according to the driving behavior feature vector and the average driving behavior feature vector.

24. The apparatus of claim 21,

the prediction unit is further used for acquiring a horizontal deviation value and a vertical deviation value in a preset track prediction equation; acquiring a first shape parameter value and a second shape parameter value in the preset trajectory prediction equation; and predicting the motion track of the target vehicle through the preset track prediction equation according to the transverse deviation value, the longitudinal deviation value, the first shape parameter value, the second shape parameter value and the relative distance.

25. The apparatus of claim 24,

the prediction unit is further configured to obtain a first probability density distribution of a first shape parameter and a second probability density distribution of a second shape parameter, the first probability density distribution including a correspondence of a parameter value of the first shape parameter and a probability density distribution, the second probability density distribution including a correspondence of a parameter value of the second shape parameter and a probability density distribution; selecting the first shape parameter value corresponding to a maximum probability density value from the first probability density distribution, and selecting the second shape parameter value corresponding to a maximum probability density value from the second probability density distribution.

26. The apparatus of claim 25,

the prediction unit is further configured to acquire the first probability density distribution and the second probability density distribution corresponding to the position information according to the position information of the target vehicle at the current time; or,

the prediction unit is further configured to acquire the first probability density distribution and the second probability density distribution corresponding to the environment information according to the environment information of the target vehicle at the current time.

27. The apparatus of claim 22, further comprising:

a third obtaining unit, configured to obtain a relative speed between the host vehicle and the target vehicle at the current time;

a first determining unit, configured to determine, according to a lane change probability of the target vehicle at the current time, a straight-ahead probability of the target vehicle at the current time, where the straight-ahead probability is used to indicate a probability that the target vehicle makes a straight-ahead movement;

a fourth obtaining unit, configured to obtain a weighting coefficient of each of the plurality of first driving behavior characteristic values at the current time;

and the second determining unit is used for determining the movement speed of the target vehicle at the next moment of the current moment through a preset longitudinal movement equation according to the relative speed, the straight-ahead probability, the plurality of first driving behavior characteristic values and the weighting coefficient of each first driving behavior characteristic value at the current moment.

28. The apparatus of claim 27,

the fourth obtaining unit is further configured to obtain a plurality of second driving behavior feature values of the target vehicle at a previous time before the current time, and a weighting coefficient of each of the plurality of second driving behavior feature values at the previous time; and determining the weighting coefficient of each first driving behavior characteristic value at the current moment through a preset algorithm according to each second driving behavior characteristic value, the weighting coefficient of each second driving behavior characteristic value at the previous moment and the movement speed of the target vehicle at the current moment.

29. The apparatus of claim 21,

the first obtaining unit is further configured to obtain the plurality of driving behavior characteristics, and obtain motion information of the target vehicle at the current time, where the motion information includes at least one of a motion speed, a motion acceleration, and a motion direction; and determining a plurality of first driving behavior characteristic values of the target vehicle at the current moment according to the motion information and the plurality of driving behavior characteristics.

30. The apparatus of claim 29,

the first obtaining unit is further configured to send an obtaining request to a server, where the obtaining request is used to request the server to send the plurality of driving behavior characteristics; and receiving the plurality of driving behavior characteristics sent by the server according to the acquisition request, wherein the plurality of driving behavior characteristics are characteristics related to the driving behaviors, which are obtained by the server according to the historical motion information of the plurality of vehicles.

31. The apparatus of claim 29,

the first acquisition unit is further used for acquiring historical motion information of the plurality of vehicles, wherein the historical motion information comprises at least one of motion speed, motion acceleration and motion direction; and determining the driving behavior characteristics through a preset training model according to the historical motion information of the vehicles.

32. The apparatus of claim 31,

the first obtaining unit is further configured to respectively determine, according to historical motion information of each vehicle in the plurality of vehicles, a lateral speed, a linear speed, a lateral acceleration, and a lane line lateral deviation between each vehicle and a lane line where the vehicle is located at a historical time, where the historical time is a time corresponding to the historical motion information; acquiring statistical information of each vehicle in at least one preset time before the historical moment, wherein the statistical information comprises at least one of a mean value, a standard deviation, a variation coefficient, a root mean square, shannon entropy, a logarithmic energy entropy and a normal entropy; and determining the plurality of driving behavior characteristics through a preset training model according to the transverse speed, the linear speed and the transverse acceleration of each vehicle at the historical moment, the transverse deviation of the lane line between each vehicle and the lane line where the vehicle is located, and statistical information of each vehicle in at least one preset time before the historical moment.

33. The apparatus of claim 32,

the first obtaining unit is further configured to determine a plurality of candidate driving behavior features through a preset training model according to the lateral speed, the linear speed and the lateral acceleration of each vehicle at a historical time, the lateral deviation of each vehicle from a lane line where the vehicle is located, and statistical information of each vehicle in at least one preset time before the historical time; selecting a plurality of driving behavior features associated with the driving behavior from the plurality of candidate driving behavior features.

34. The apparatus of claim 33,

the first obtaining unit is further configured to respectively determine a correlation between each candidate driving behavior feature of the plurality of candidate driving behavior features and the driving behavior through a preset correlation algorithm; and selecting a plurality of driving behavior characteristics with correlation degrees meeting preset conditions from the plurality of candidate driving behavior characteristics according to the correlation degree between each candidate driving behavior characteristic and the lane changing behavior.

35. The apparatus of claim 21, further comprising:

the control unit is used for determining a driving strategy corresponding to the motion trail according to the motion trail; and controlling the vehicle to run according to the driving strategy.

36. A trajectory prediction device, characterized in that the device comprises:

a fifth acquisition unit configured to acquire historical motion information of a plurality of vehicles, the historical motion information including at least one of a motion speed, a motion acceleration, and a motion direction;

and the third determining unit is used for determining a plurality of driving behavior characteristics according to the historical motion information of the plurality of vehicles through a preset training model, wherein the plurality of driving behavior characteristics are characteristics related to driving behaviors, and the driving behaviors comprise lane changing or straight driving.

37. The apparatus of claim 36,

the third determining unit is further configured to determine, according to historical motion information of each of the plurality of vehicles, a lateral velocity, a linear velocity, a lateral acceleration, and a lane line lateral deviation between each of the vehicles and a lane line located therebetween at a historical time, where the historical time is a time corresponding to the historical motion information; acquiring statistical information of each vehicle in at least one preset time before the historical moment, wherein the statistical information comprises at least one of a mean value, a standard deviation, a variation coefficient, a root mean square, shannon entropy, a logarithmic energy entropy and a normal entropy; and determining the plurality of driving behavior characteristics through a preset training model according to the transverse speed, the linear speed and the transverse acceleration of each vehicle at the historical moment, the transverse deviation of the lane line between each vehicle and the lane line where the vehicle is located, and statistical information of each vehicle in at least one preset time before the historical moment.

38. The apparatus of claim 37,

the third determining unit is further configured to determine a plurality of candidate driving behavior features through a preset training model according to the lateral speed, the linear speed and the lateral acceleration of each vehicle at a historical time, the lateral deviation of each vehicle from a lane line where the vehicle is located, and statistical information of each vehicle in at least one preset time period before the historical time; selecting a plurality of driving behavior features associated with the driving behavior from the plurality of candidate driving behavior features.

39. The apparatus of claim 38,

the third determining unit is further configured to determine, through a preset correlation algorithm, a correlation between each candidate driving behavior feature of the plurality of candidate driving behavior features and the driving behavior; and selecting a plurality of driving behavior characteristics with correlation degrees meeting preset conditions from the plurality of candidate driving behavior characteristics according to the correlation degree between each candidate driving behavior characteristic and the lane changing behavior.

40. The apparatus of claim 36, further comprising:

the receiving unit is used for receiving an acquisition request sent by a vehicle, and the acquisition request is used for requesting the server to send the plurality of driving behavior characteristics;

a sending unit, configured to send the plurality of driving behavior characteristics to the vehicle according to the acquisition request.

41. An apparatus, characterized in that the apparatus comprises: a processor, a memory, a communication interface and a bus;

wherein the memory, the processor and the communication interface are connected by the bus, the memory having stored thereon programmable instructions, the processor invoking the programmable instructions stored on the memory for performing the method of any of claims 1-15.

42. An apparatus, characterized in that the apparatus comprises: a processor, a memory, a communication interface and a bus;

wherein the memory, the processor and the communication interface are connected by the bus, the memory having stored thereon programmable instructions, the processor invoking the programmable instructions stored on the memory for performing the method of any of claims 16-20.

43. A computer-readable storage medium, comprising instructions which, when executed on a computer, cause the computer to perform the method of any one of claims 1 to 15.

44. A computer-readable storage medium, comprising instructions which, when executed on a computer, cause the computer to perform the method of any one of claims 16 to 20.