Detailed Description
The application will be described in detail hereinafter with reference to the drawings in conjunction with embodiments. It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order.
Example 1
The present embodiment provides a positioning system, and fig. 3 is a functional schematic diagram of the positioning system provided according to an embodiment of the present invention, as shown in fig. 3, where the positioning system in the present embodiment includes:
An inertial measurement unit (Inertial Measurement Unit, IMU) 102 configured to obtain first measurement information and one or more sets of second measurement information, wherein the first measurement information is real-time measurement information of a target object, and the second measurement information is historical measurement information;
An ultra wideband UWB unit 104 configured to obtain one or more sets of second location information, wherein the second location information is historical location information;
The calculating unit 106 is configured to determine first position information according to the first measurement information and a preset neural network model, and position the target object according to the first position information; the neural network model is trained according to one or more groups of second measurement information and one or more groups of second position information.
Generally, the IMU includes a plurality of acceleration sensors and angular velocity sensors (or gyroscopes) to measure acceleration and angular velocity of a target object in a space. The IMU may be carried on the positioning tag, that is, the IMU is carried on the target object as a part of the positioning tag, or may be carried on the target object independently of the positioning tag, which is not limited in the present invention.
The first measurement information obtained by the IMU is real-time measurement information of the target object, that is, the real-time measurement information measured by the IMU in the positioning system where the target object is located when the first measurement information indicates that the target object is currently positioned. The second measurement information acquired by the IMU is history measurement information, that is, the second measurement information is history measurement information acquired by the IMU in the past. The second location information acquired by the UWB unit indicates that the second location information is historical location information acquired by UWB in the past.
It should be further noted that, the second location information may be location information obtained by locating the location system under the condition of not being influenced by environmental shielding or multipath effect or under the condition that the influence is within the allowable range of system error, that is, the second location information may be obtained by directly performing ranging and locating by UWB. For example, in the case that the positioning system is not affected by environmental shielding or multipath effects, UWB may directly perform ranging positioning processing on the object in the system to obtain the position information, i.e. the second position information; meanwhile, the IMU carried by the object can synchronously acquire the measurement information of the object, namely the second measurement information, but in the situation, the second measurement information does not participate in the positioning of the object and is only used for matching with the second position information to train the neural network model.
On the other hand, the second location information may be location information determined by the calculation unit according to the measurement information of the IUM and the neural network model in the present embodiment. For example, at a certain time, the calculating unit determines the first location information of the target object according to the first measurement information of the IUM and the neural network model, and then, with respect to the later time, the first location information may also be used as the historical location information, that is, the second location information, to participate in training of the neural network model.
It should be further noted that the second measurement information is not limited to the history measurement information of the target object, and is not limited to the history measurement information obtained by the IMU in the positioning system where the current target object is positioned, specifically, the second measurement information may be the reference object different from the target object, and the history measurement information measured by the IMU in another positioning system different from the IMU in the positioning system where the target object is currently positioned. Correspondingly, the second position information is not limited to the historical position information of the target object, and is not limited to the historical position information obtained by the UWB in the positioning system where the current target object is positioned, specifically, the second position information may be a reference object different from the target object, and in the positioning system where the current area different from the target object is located, the second position information may also be the historical position information measured by the UWB unit included in the positioning system in other areas. In short, the above-described second measurement information and second position information are not limited to those obtained by the positioning system currently performing positioning, but may be those obtained by other positioning systems historically.
In the positioning system of this embodiment, the neural network model obtained by training according to the second measurement information and the second position information may indicate a mapping relationship between the measurement information obtained by the IMU and the position information obtained by the corresponding UWB. Therefore, if the UWB unit cannot effectively locate the target object due to environmental occlusion or multipath effect at a certain moment, the resolving unit may input the first measurement information indicating the current real-time measurement information of the target object measured by the IMU into the neural network model, and obtain the position information corresponding to the first measurement information according to the mapping relationship between the measurement information and the position information.
It should be noted that, in this embodiment, the location information may be distance information (relative to the positioning base station) or coordinate information.
The above-mentioned resolving unit may be set in the positioning tag, or may use a computing device, such as a PC, in a scene where the positioning system is located to perform resolving processing, or may use a server set in the cloud to perform resolving processing, which is not limited in the present invention. The neural network model may be stored in the calculation unit, or may be stored in another computer-readable storage medium, which is not limited to the present invention.
With the positioning system in the above embodiment, since the training of the neural network model is performed by the one or more sets of second measurement information as the history measurement information acquired by the inertial measurement unit IMU and the one or more sets of second position information as the history position information acquired by the ultra wideband UWB unit, the positioning of the target object can be completed by further acquiring the first position information corresponding to the target by the first measurement information acquired by the inertial measurement unit IMU and the neural network model. Therefore, the positioning system in the above embodiment can solve the problem that the positioning effect of UWB positioning in the related art is easily interfered by environmental shielding or multipath effect, so as to achieve the effect of improving the accuracy of UWB positioning.
Specifically, when the positioning system is not affected by environmental shielding and the like at the historical moment, the ranging of the UWB unit is not affected, so that the positioning system can accurately determine the position information of the object, namely the second position information, and the measurement information obtained by the IMU measurement at the moment is also accurate (the IMU is not affected by environmental shielding and other factors), namely the second measurement information; therefore, the second measurement information can correspond to accurate second position information. The neural network model trained by the method can obtain the position information corresponding to the measurement information obtained by IMU measurement. And when the UWB unit is influenced by environmental shielding and the like at the current moment and cannot accurately perform ranging and positioning, the IMU can acquire the measurement information of the object at the current moment, and further calculates the corresponding position information according to the neural network model to perform positioning processing.
It should be further noted that, the positioning system in this embodiment may also perform positioning processing according to a conventional UWB positioning manner in the related art. Whether the UWB unit is affected in the ranging process, namely whether the first measurement information obtained by IMU measurement and the neural network model are needed to be used for positioning, can be judged according to the detection of the signals of the positioning base station by the positioning tag.
In addition, when the UWB unit is affected by environmental occlusion or multipath effect, it may not cause the ranging process in the positioning of the UWB unit to be totally disabled, that is, there is a situation in which the ranging process of the UWB unit is still partially effective, in which case, the ranging result of the UWB that is partially effective may be fused with the first position information determined by the neural network model by means of filtering processing, specifically, such as kalman filtering or particle filtering, so as to implement more accurate positioning processing on the target object.
In an alternative embodiment, the IMU includes at least one of: acceleration sensor, angular velocity sensor, geomagnetic sensor; wherein,
The acceleration sensor is configured to acquire first acceleration information and second acceleration information, wherein the first acceleration information is real-time acceleration information of a target object, and the second acceleration information is historical acceleration information;
The angular velocity sensor is configured to acquire first angular velocity information, or the angular velocity sensor is configured to acquire first angular velocity information and second angular velocity information; the first angular velocity information is real-time angular velocity information of the target object, and the second angular velocity information is historical angular velocity information;
the geomagnetic sensor is configured to acquire first geomagnetic information, or the geomagnetic sensor is configured to acquire the first geomagnetic information and second geomagnetic information; the first geomagnetic information is real-time geomagnetic information of the target object, and the second geomagnetic information is historical geomagnetic information.
It should be further noted that, the first acceleration information and the second acceleration information obtained by the acceleration sensor are both used for indicating the acceleration of the measurement object; the first angular velocity information and the second angular velocity information obtained by the angular velocity sensor are used for indicating the angular velocity of the measured object; the first geomagnetic information and the second geomagnetic information obtained by the geomagnetic sensor are used for indicating geomagnetic information of the position where the measured object is located, such as geomagnetic vectors and the like. The angular velocity sensor and the geomagnetic sensor can acquire different objects in different working modes, for example, the angular velocity sensor can acquire only first angular velocity information without acquiring second angular velocity information for training a neural network model, or acquire the second angular velocity information for training the neural network model while acquiring the first angular velocity information for positioning, and the geomagnetic sensor is similar; this will be illustrated in the following alternative embodiments, and will not be described in detail here.
In an alternative embodiment, the first measurement information comprises at least first acceleration information and the second measurement information comprises at least second acceleration information;
The calculation unit is further configured to determine first rate information according to the first acceleration information and the neural network model, and determine first position information according to the first rate information and the first angular velocity information, so as to position the target object.
It should be further noted that, in the above-mentioned alternative embodiment, the first measurement information includes at least first acceleration information, and the second measurement information includes at least second acceleration information, where the mapping relationship between the second measurement information and the second location information in the neural network model is implemented based on the velocity information; specifically, the second measurement information is acceleration, that is, the rate corresponding to the second measurement information can be obtained through a calculation mode, and the corresponding second position information can also be converted into the rate corresponding to the second position information, so that training can be performed in the neural network model based on the mapping relation of the two rates. On the basis, when the first measurement information is the first acceleration information, the first acceleration information can be input into the neural network model, the first acceleration information is converted into the corresponding velocity, and the velocity corresponding to the corresponding position information is determined according to the mapping relation, namely the first velocity information in the alternative embodiment.
The first rate information is used for indicating the current rate of the target object, and based on the first rate information, the vector speed of the target object can be further determined according to the first angular speed information obtained by the IMU, and based on the vector speed, the real-time speed of the target object can be obtained, and the target object can be positioned.
In an optional embodiment, the system further includes a filtering unit configured to perform complementary filtering processing according to the first geomagnetic information and the first angular velocity information to obtain first heading information;
The resolving unit is further configured to determine first location information based on the first rate information and the first heading information to locate the target object.
It should be further noted that, the complementary filtering process is performed according to the first geomagnetic information and the first angular velocity information, so as to determine a heading information, so as to further determine a moving direction of the target object, and thus, the target object is positioned by combining the first velocity information. The complementary filtering described above can be achieved by the mahonyl algorithm.
In an alternative embodiment, the first measurement information includes at least first acceleration information and first angular velocity information, and the second measurement information includes at least second acceleration information and second angular velocity information;
The resolving unit 106 is further configured to determine first relative inertia information according to the first acceleration information and the first angular velocity information, determine first velocity information according to the first relative inertia information and the neural network model, and determine first position information according to the first velocity information, so as to position the target object;
The first relative inertia information is used for indicating speed information of the target user in a first coordinate, and the first coordinate is a station center coordinate of the target user.
It should be further noted that, in the above-mentioned alternative embodiment, the first measurement information includes at least first acceleration information and first angular velocity information, and the second measurement information includes at least second acceleration information and second angular velocity information, where the mapping relationship between the second measurement information and the second position information in the neural network model is implemented based on velocity information in the first coordinate; specifically, the second measurement information is acceleration and angular velocity, and thus, inertia information indicating velocity information corresponding to the station coordinates having the measurement object as the origin, that is, the first coordinates, can be determined. The neural network model may be trained based on the speed information indicated by the inertial information and the speed information corresponding to the corresponding second position information, so as to obtain the mapping relationship.
On the basis, when the IMU acquires the inertial information corresponding to the first measurement information, the IMU can acquire the corresponding speed information according to the neural network model to serve as the first speed information, so that the possible movement position and movement direction of the target object at the current moment, namely the movement state of the target object, are determined. Based on the above, the target object can be positioned at the current moment by combining the motion path of the previous target object or the ranging information of other positioning base stations, so as to determine the first position information. In the actual calculation process, the speed information may be directly subjected to integral processing to obtain corresponding position information.
It should be further noted that, since the angular velocity information obtained by the angular velocity sensor in the IMU is based on the relative coordinate system of the sensor itself, that is, the station coordinate with the target object as the origin, the first coordinate is a relative coordinate system, and the corresponding inertial information obtained according to the first measurement information is the first relative inertial information in the alternative embodiment.
In an alternative embodiment, the first measurement information further includes first geomagnetic information, and the second measurement information further includes second geomagnetic information;
The resolving unit 106 is further configured to determine first relative inertial information according to the first acceleration information and the first angular velocity information, and convert the first relative inertial information according to the first geomagnetic information to obtain first absolute inertial information;
Determining first speed information according to the first absolute inertia information and the neural network model, and determining first position information according to the first speed information so as to position a target object;
the first absolute inertial information is used for indicating speed information of the target user in a second coordinate, and the second coordinate is a geocentric coordinate.
It should be further noted that, in the above-mentioned alternative embodiment, geomagnetic information is introduced into the first measurement information and the second measurement information, so that, for the neural network model, the second measurement information referred to in the training stage is actually a new inertial information, that is, absolute inertial information, determined by converting the inertial information by geomagnetic information after acquiring corresponding relative inertial information according to acceleration information and angular velocity information; specifically, after geomagnetic information is introduced, the geomagnetic information may be used as a conversion coefficient (the coefficient may be a geomagnetic vector, that is, an included angle between an object and a magnetic north direction), so as to convert relative inertial information, thereby obtaining the absolute inertial information, where the absolute inertial information is used to indicate a coordinate of the earth center, specifically, a local northeast coordinate where the positioning system is located, that is, corresponding speed information in the second coordinate. The neural network model may be trained based on the speed information indicated by the absolute inertia information and the speed information corresponding to the corresponding second position information, so as to obtain a mapping relationship thereof.
On the basis, when the IMU acquires absolute inertia information corresponding to the first measurement information, corresponding speed information can be acquired according to the neural network model to serve as first speed information, and further the possible movement position and movement direction of the target object at the current moment, namely the movement state of the target object, are determined. Based on the above, the target object can be positioned at the current moment by combining the motion path of the previous target object or the ranging information of other positioning base stations, so as to determine the first position information. In the actual calculation process, the speed information may be directly subjected to integral processing to obtain corresponding position information.
In addition, the correspondence relationship may be directly associated with the speed information without using the coordinate information, and specifically, the calculation means may be configured to:
determining measured velocity information based on the first acceleration information and the first angular velocity information,
And determining first speed information according to the measured speed information and the neural network model, and determining first position information according to the first speed information so as to locate the target object.
The measured speed information is speed information obtained by calculating according to the first acceleration information and the first angular speed information of the target object measured by the IMU.
In an alternative embodiment, the neural network model includes a neural network model weight, wherein the neural network model weight is used to indicate a mapping relationship between the second measurement information and the corresponding one or more sets of second rate information;
Determining the weight of the neural network model according to the regression relation between an input sample and an output sample of the neural network model, wherein the input sample is one or more groups of second measurement information, and the output sample is one or more groups of second rate information;
the second rate information is used for indicating historical rate information, and one or more groups of rate information are acquired according to one or more groups of second position information.
It should be further noted that the training of the neural network model is applicable to a case where the first measurement information is first acceleration information and the second measurement information is second acceleration information. As described in the foregoing optional embodiment, the mapping relationship between the second measurement information and the second location information in the neural network model is implemented based on rate information; specifically, after the UWB unit acquires the second position information, the second position information may be converted to obtain corresponding velocity information, that is, the second velocity information in the above alternative embodiment, and the second acceleration information acquired by the IMU may also be converted to velocity information by a calculation manner, so as to correspond to the second velocity information. The process of determining the correspondence relationship, that is, the process of determining the neural network weight in the neural network model, where the neural network weight may indicate the mapping relationship between the second acceleration information and the second rate information, and further indicate the mapping relationship between the second measurement information and the second rate information.
In this alternative embodiment, the neural network model may be a BP neural network model, or an equivalent neural network model; the structure of the BP neural network model and the training process are known to those skilled in the art, and are not described in detail herein.
In an alternative embodiment, the resolving unit 106 is further configured to:
the first acceleration information is used as an input parameter of a neural network model, a corresponding output parameter is obtained according to the neural network model, and the output parameter is used as first rate information;
and determining first position information according to the first speed information and the first angular velocity information so as to position the target object.
In an alternative embodiment, the neural network model includes a neural network model weight, wherein the neural network model weight is used to indicate a mapping relationship between the second measurement information and the corresponding one or more sets of second speed information;
Determining the weight of the neural network model according to the regression relation between an input sample and an output sample of the neural network model, wherein the input sample is one or more groups of second measurement information, and the output sample is one or more groups of second speed information;
the second speed information is used for indicating historical speed information, and one or more groups of speed information are acquired according to one or more groups of second position information.
It should be further noted that, the training of the neural network model is applicable to the case that the first measurement information is the first acceleration information, the first angular velocity information, and the first geomagnetic information, and the second measurement information is the second acceleration information, the second angular velocity information, and the second geomagnetic information. As in the previous alternative embodiment, the mapping relationship between the second measurement information and the second location information in the neural network model is implemented based on velocity information; specifically, after the IMU acquires the second acceleration information, the second angular velocity information, and the second geomagnetic information, the second acceleration information, the second angular velocity information, and the second geomagnetic information may be converted into second absolute inertial information, so as to indicate velocity information of the object in the geocentric coordinates; and further, the second absolute inertia information is corresponding to the speed information corresponding to the second position information acquired by the UWB unit, so that the corresponding relation between the second measurement information and the second position information can be determined. The process of determining the correspondence relationship, that is, the process of determining the neural network weight in the neural network model, where the neural network weight may indicate a mapping relationship between the second absolute inertial information and the velocity information corresponding to the second position information, and further indicate a mapping relationship between the second measurement information and the second position information.
It should be further noted that the second speed information may be obtained by performing calculation processing on the second location information, for example, deriving from the second location information.
In this alternative embodiment, the neural network model may be a BP neural network model, or an equivalent neural network model; the structure of the BP neural network model and the training process are known to those skilled in the art, and are not described in detail herein.
In an alternative embodiment, the resolving unit 106 is configured to:
Taking the first acceleration information, the first angular velocity information and the first geomagnetic information as input parameters of a neural network model, acquiring corresponding output parameters according to the neural network model, and taking the output parameters as first velocity information;
and determining first position information according to the first speed information, and positioning the target object according to the first position information.
In an alternative embodiment, the one or more sets of second measurement information include: one or more sets of second measurement information corresponding to the target object and/or one or more sets of second measurement information corresponding to the reference object;
the one or more sets of second location information include: one or more sets of second position information corresponding to the target object, and/or one or more sets of second position information corresponding to the reference object.
It should be further noted that the target object is the target object in the positioning system in this embodiment, and the reference object may be a target object or another object other than the target object. For example, in the same positioning system, the object a, the object B and the object C are respectively positioned at different times in the past, so that the object a, the object B and the object C can be respectively used as reference objects to obtain one or more sets of second measurement signals and one or more sets of second position information corresponding to the object a, the object B and the object C, thereby realizing training of the neural network model.
It is assumed that the object D is located in the positioning system at the current moment, that is, the object D is taken as a target object, and at this time, the trained neural network model can be used for positioning processing as the target object.
Furthermore, the reference object may not be limited to the same positioning system, and different positioning systems distributed in different areas may participate in training of the neural network model. For example, the positioning system M in beijing can obtain one or more sets of second measurement information of the object M positioned at different times through the IMU, and obtain one or more sets of second location information of the object M positioned at different times through the UWB unit; meanwhile, the positioning system N in Shanghai city can acquire one or more groups of second measurement information of the object N positioned at different moments through the IMU, and acquire one or more groups of second position information of the object N positioned at different moments through the UWB unit. The second measurement information and the second position information acquired by the positioning system M, and the second measurement information and the second position information acquired by the positioning system N may be summarized in a unified manner, for example, uploaded to a cloud server to perform training of the neural network model in a unified manner.
The object O in guangzhou city is assumed to be positioned in the positioning system O, that is, the object O is taken as a target object, and the trained neural network model can be utilized to perform positioning processing on the target object.
Since the processing accuracy of the neural network model increases with the number of training samples, in the above-described alternative embodiment, the accuracy of the neural network model can be significantly improved by restricting the training samples of the neural network model to no longer be the target object.
Meanwhile, because the training of the neural network model is not limited to the target object, the neural network model can start training before a specific positioning system works. Under the condition that the UWB unit cannot effectively position due to environmental shielding or multipath effect in the positioning system in the initial stage of positioning processing, the target object can be positioned by matching the IMU with the neural network model.
In an alternative embodiment, the resolving unit 106 is further configured to:
And acquiring third position information of the target object through the UWB unit, and positioning the target object according to the third position information.
It should be further noted that, the third location information indicates that the UWB unit directly performs ranging and positioning processing on the target object, that is, the positioning system in this embodiment may perform positioning processing on the target object without using the first measurement information measured by the IMU and the neural network model, and performing positioning processing on the target object by using a conventional UWB unit.
The following describes the operation of the positioning system in this embodiment by way of specific embodiments.
Example 1
Fig. 4 is a schematic operation diagram of a positioning system according to an embodiment of the present invention, and as shown in fig. 4, the positioning system includes an IMU composed of an acceleration sensor, an angular velocity sensor, and a geomagnetic sensor, and a UWB unit.
In the working process of the positioning system, at a first moment, the IMU acquires the acceleration of the target object, and the UWB unit acquires the ranging value (namely the position information in the embodiment) of the target object, and trains the acceleration and the ranging value as input samples and output samples of the neural network model. Specifically, the neural network model is a mapping relationship between the acceleration and the velocity corresponding to the ranging value, which is determined according to a regression relationship between the velocity corresponding to the acceleration and the velocity corresponding to the ranging value.
At the second time (the first time is a historical time relative to the second time), the UWB unit has a dead zone due to environmental occlusion or multipath factors, i.e., the data obtained by the UWB is partially invalid, so that an accurate ranging value cannot be obtained. At the moment, the current acceleration of the target object can be obtained through the IMU, the speed information corresponding to the acceleration is determined through the neural network model, meanwhile, the angular velocity sensor and the geomagnetic sensor in the IMU are subjected to complementary filtering to obtain the heading information of the target object, and the vector velocity information of the target object can be determined by integrating the speed information and the heading information of the target object.
The possible motion state of the target object can be determined based on the vector velocity information of the target object, at this time, the motion state of the target object and normal information which can be obtained by the UWB, such as a motion path or motion state before the target object, or a coordinate position or distance of the target object, are fused, specifically, particle filtering or kalman filtering, so that the current position of the target object can be correspondingly obtained, and positioning is further realized.
It should be further noted that, in the above embodiments, the main body of the neural network model training may be different, and the process of the neural network model training and the positioning system to position the target object is further described in the following embodiments.
Example 2
Fig. 5 is an interaction schematic diagram (one) of a positioning system according to an embodiment of the present invention, as shown in fig. 5, in the working process of the positioning system, a corresponding resolving engine (i.e. the resolving unit in the above embodiment) obtains a ranging value or a position coordinate of a positioning base station, so as to calculate a second motion state of a target object, and sends the second motion state of the target object to a positioning tag carried by the target object. The positioning label can acquire corresponding measurement information according to the second motion state sent by the resolving engine and the IMU carried by the positioning label, so that training of the neural network model is completed.
At a later time, the positioning tag can rely on corresponding measurement information acquired by the carried IMU to determine the first motion state of the target object according to the trained neural network model. The positioning label further sends the first motion state to a resolving engine, and the resolving engine determines the second motion state or the first motion state of the target object at the current moment of selecting the target object as the motion state of the target object, namely the speed information of the target object according to the motion mode of the target object at the previous moment, so that the positioning processing of the target object is completed.
Example 3
Fig. 6 is an interaction schematic diagram (two) of a positioning system according to an embodiment of the present invention, as shown in fig. 6, in the working process of the positioning system, a corresponding resolving engine (i.e. the resolving unit in the above embodiment) obtains a ranging value or a position coordinate of a positioning base station so as to calculate a second motion state of a target object, and at the same time, the positioning tag sends corresponding measurement information obtained according to an IMU carried by itself to the resolving engine, so that the resolving engine performs training of a neural network model according to the second motion state and the measurement information.
After the calculation engine finishes training the neural network model, the weight of the neural network model can be sent to the positioning label, namely the calculation engine directly informs the relation between the measurement information of the positioning label and the corresponding motion state. And at the later moment, the positioning tag can acquire the first motion state of the target object according to the measurement information at the current moment and the weight of the neural network model. The positioning tag sends the first motion state to a resolving engine, and the resolving engine determines the second motion state or the first motion state of the target object at the current moment of selecting the target object as the motion state of the target object, namely the speed information of the target object according to the motion mode of the target object at the previous moment, so that the positioning processing of the target object is completed.
Example 3
Fig. 7 is an interaction schematic diagram (iii) of a positioning system according to an embodiment of the present invention, as shown in fig. 7, in the working process of the positioning system, a corresponding resolving engine (i.e. the resolving unit in the above embodiment) obtains a ranging value or a position coordinate of a positioning base station so as to calculate a second motion state of a target object, and at the same time, the positioning tag sends corresponding measurement information obtained according to an IMU carried by itself to the resolving engine, so that the resolving engine performs training of a neural network model according to the second motion state and the measurement information.
And the positioning label can send the measurement information of the current moment to the resolving engine so that the resolving engine can obtain the corresponding first motion state according to the measurement information of the current moment and the trained neural network model on the one hand, and can also calculate the second motion state of the target object through the ranging value of the positioning base station obtained at the current moment. The resolving engine further determines the second motion state or the first motion state of the target object at the current moment of the target object as the motion state of the target object, namely the speed information of the target object, according to the motion mode of the target object at the previous moment, so that the positioning processing of the target object is completed.
Example 2
The present embodiment provides a database, and fig. 8 is a schematic diagram of the database provided according to an embodiment of the present invention, as shown in fig. 8, where the database in the present embodiment includes:
A database, comprising:
one or more sets of second measurement information, one or more sets of second location information, and correspondence between the one or more sets of second measurement information and the one or more sets of second location information;
wherein, one or more groups of second measurement information are obtained through one or more Inertial Measurement Units (IMU), and one or more groups of second position information are obtained through one or more ultra wideband UWB units.
It should be further noted that, the database in this embodiment indicates the database storing the second measurement information, the second location information, and the correspondence between the second measurement information and the second location information. The corresponding relation between the second measurement information and the second position information can be obtained through a neural network training mode or a fitting mode. The database in this embodiment may be stored in a server or a terminal in a scene where a certain positioning system is located, or may be stored in a cloud server, which is not limited in the present invention.
It should be further noted that, in the above database, one or more sets of second measurement information may be acquired by one or more IMUs, and specifically, the second measurement information in the database may be measurement information of the same object or different objects acquired by the same IMU, or measurement information of different objects acquired by multiple IMUs respectively; correspondingly, one or more groups of second position information is acquired through one or more UWB units, and the second position information in the specific indication database can be the position information of the same object or different objects acquired by the same UWB, or the position information of different objects acquired by a plurality of UWB respectively. The IMU and the UWB unit in this embodiment refer to the IMU and the UWB unit in the positioning system shown in embodiment 1, that is, in this embodiment, when a certain IMU obtains a set of second measurement information, the UWB unit located in the same positioning system as the IMU also obtains a set of corresponding second location information, and the second location information is synchronously uploaded to the database in this embodiment.
In an alternative embodiment, the database further comprises: a preset neural network model;
The database is further configured to determine a mapping relationship between the second measurement information and second location information corresponding to the second measurement information according to the neural network model.
It should be further noted that, in the above alternative embodiment, the mapping relationship between the second measurement information and the second location information corresponding to the second measurement information is determined by training the neural network model. According to the difference of the second measurement information, the neural network model may adopt different training manners, specifically, the training manner of the neural network model in this alternative embodiment corresponds to the training manner of the neural network model in embodiment 1, so that the description thereof will not be repeated here.
Example 3
The present embodiment provides a positioning method, and fig. 9 is a schematic diagram of the positioning method provided according to the embodiment of the present invention, as shown in fig. 9, where the positioning method in the present embodiment includes:
S302, the IMU acquires first measurement information and one or more groups of second measurement information; wherein the first measurement information is real-time measurement information of the target object, and the one or more sets of second measurement information are historical measurement information;
S304, determining first position information according to the first measurement information and a preset neural network model by a resolving unit, and positioning a target object according to the first position information;
The neural network model is trained according to one or more groups of second measurement information and one or more groups of second position information, and the one or more groups of second position information is historical position information.
The technical features and effects of the positioning method in this embodiment correspond to those of the positioning system in embodiment 1, and thus are not described herein.
In an alternative embodiment, the first measurement information comprises at least first acceleration information and the second measurement information comprises at least second acceleration information;
determining first location information according to the first measurement information and a preset neural network model, including:
Determining first rate information according to the first acceleration information and the neural network model;
first angular velocity information is acquired, and first position information is determined according to the first velocity information and the first angular velocity information.
In an alternative embodiment, the method further comprises: acquiring first geomagnetic information, and performing complementary filtering processing according to the first geomagnetic information and first angular velocity information to acquire first heading information;
determining first position information according to the first rate information and the first angular velocity information, including:
First location information is determined based on the first speed information and the first heading information.
In an alternative embodiment, the first measurement information includes at least first acceleration information and first angular velocity information, and the second measurement information includes at least second acceleration information and second angular velocity information;
determining first location information according to the first measurement information and a preset neural network model, including:
Determining first relative inertia information according to the first acceleration information and the first angular velocity information, determining first velocity information according to the first relative inertia information and the neural network model, and determining first position information according to the first velocity information so as to position a target object;
The first relative inertia information is used for indicating speed information of the target user in a first coordinate, and the first coordinate is a station center coordinate of the target user.
In an alternative embodiment, the first measurement information further includes first geomagnetic information, and the second measurement information further includes second geomagnetic information;
determining first position information according to the first measurement information and a preset neural network model, and further comprising:
Determining first relative inertial information according to the first acceleration information and the first angular velocity information, and converting the first relative inertial information according to the first geomagnetic information to obtain first absolute inertial information; determining first speed information according to the first absolute inertia information and the neural network model, and determining first position information according to the first speed information so as to position a target object;
the first absolute inertial information is used for indicating speed information of the target user in a second coordinate, and the second coordinate is a geocentric coordinate.
In an alternative embodiment, the neural network model includes a neural network model weight, wherein the neural network model weight is used to indicate a mapping relationship between the second measurement information and the corresponding one or more sets of second rate information;
Determining the weight of the neural network model according to the regression relation between an input sample and an output sample of the neural network model, wherein the input sample is one or more groups of second measurement information, and the output sample is one or more groups of second rate information;
the second rate information is used for indicating historical rate information, and one or more groups of rate information are acquired according to one or more groups of second position information.
In an alternative embodiment, determining the first rate information from the first acceleration information and the neural network model includes:
And taking the first acceleration information as an input parameter of the neural network model, acquiring a corresponding output parameter according to the neural network model, and taking the output parameter as first speed information.
In an alternative embodiment, the neural network model includes a neural network model weight, wherein the neural network model weight is used to indicate a mapping relationship between the second measurement information and the corresponding one or more sets of second rate information;
Determining the weight of the neural network model according to the regression relation between an input sample and an output sample of the neural network model, wherein the input sample is one or more groups of second measurement information, and the output sample is one or more groups of second rate information;
the second rate information is used for indicating historical rate information, and one or more groups of rate information are acquired according to one or more groups of second position information.
In an alternative embodiment, determining the first location information according to the first measurement information and the preset neural network model includes:
Taking the first acceleration information, the first angular velocity information and the first geomagnetic information as input parameters of a neural network model, acquiring corresponding output parameters according to the neural network model, and taking the output parameters as first velocity information;
and determining first position information according to the first speed information, and positioning the target object according to the first position information.
In an alternative embodiment, the one or more sets of second measurement information include: one or more sets of second measurement information corresponding to the target object and/or one or more sets of second measurement information corresponding to the reference object;
the one or more sets of second location information include: one or more sets of second position information corresponding to the target object and/or one or more sets of second position information corresponding to the reference object.
In an alternative embodiment, the method further comprises:
And acquiring third position information of the target object through the UWB unit, and positioning the target object according to the third position information.
From the description of the above embodiments, it will be clear to a person skilled in the art that the method according to the above embodiments may be implemented by means of software plus the necessary general hardware platform, but of course also by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) comprising several instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method of the various embodiments of the present invention.
Example 4
Fig. 10 is a flowchart of a neural network model training method according to an embodiment of the present invention, as shown in fig. 10, where the neural network model training method in the embodiment includes:
S402, acquiring one or more groups of second measurement information, wherein the one or more groups of second measurement information are acquired through an Inertial Measurement Unit (IMU);
s404, acquiring one or more groups of second position information, wherein the one or more groups of second position information are acquired through an ultra wideband UWB unit;
and S406, training the preset neural network model according to one or more groups of second measurement information and one or more groups of second position information to obtain a trained neural network model.
It should be further noted that, in the above embodiment, the IMU and the UWB unit are the IMU and the UWB unit in the positioning system shown in embodiment 1, that is, in this embodiment, when a certain IMU obtains a set of second measurement information, a UWB unit located in the same positioning system as the IMU also obtains a set of corresponding second location information, and the second location information is synchronously uploaded to the database in this embodiment. In step S406, the training process of the preset neural network model according to the one or more sets of second measurement information and the one or more sets of second position information is to use the second measurement information and the second position information acquired by the IMU and the UWB unit in the same time as the second measurement information and the second position information in the same positioning system as training samples of the neural network model, that is, in the training process of the neural network model, when a set of second measurement information is input, a set of second position information corresponding to the second measurement information is correspondingly input to complete training.
It should be further noted that the execution subject of steps S402 to S406 is a processing unit, such as a resolving engine, in which the subject of the neural network model is stored.
In an alternative embodiment, training the preset neural network model according to one or more sets of second measurement information and one or more sets of second location information includes:
Taking one or more groups of second measurement information as an input sample of the neural network model, and taking one or more groups of second position information as an output sample of the neural network model;
Determining a neural network model weight according to a regression relationship between an input sample and an output sample of the neural network model, wherein the neural network model weight is used for indicating a mapping relationship between second measurement information and second position information corresponding to the second measurement information;
and training the neural network model according to the weight of the neural network model.
It should be further noted that, the regression relationship between the input samples and the output samples of the neural network model may be determined by fitting or the like. The neural network model may be a BR neural network model.
In an alternative embodiment, the second measurement information includes at least: second acceleration information;
determining a neural network model weight from a regression relationship between an input sample and an output sample of the neural network model, comprising:
Acquiring second rate information corresponding to the second position information according to the second position information;
And determining the weight of the neural network model according to the regression relation between the second acceleration information and the second rate information.
The rate information obtained by converting the second rate information, that is, the second position information; and determining the weight of the neural network model according to the regression relation between the second acceleration information and the second rate information, wherein the second acceleration information can be converted into corresponding rate information (for example, through integral processing) so as to realize the correspondence with the second rate information, and thus, training of the neural network model based on the rate information can be completed.
In an alternative embodiment, the second measurement information includes: second acceleration information, second angular velocity information;
determining a neural network model weight from a regression relationship between an input sample and an output sample of the neural network model, comprising:
Determining corresponding second relative inertia information according to the second acceleration information and the second angular velocity information, and acquiring second velocity information corresponding to the second position information according to the second position information; the second relative inertia information is used for indicating speed information of the target object in a first coordinate, and the first coordinate is a station center coordinate of the target object;
And determining the weight of the neural network model according to the regression relation between the second relative inertia information and the second speed information.
It should be further noted that the second relative inertial information indicates relative inertial information of the object at the historical moment determined according to the second acceleration information and the second angular velocity information, specifically, station coordinates corresponding to the object at the historical moment, that is, velocity information in the first coordinates. And determining the weight of the neural network model according to the regression relation between the second relative inertia information and the second speed information, so that training of the neural network model based on the relative coordinate system can be completed.
In an alternative embodiment, the second measurement information further comprises: second geomagnetic information;
determining a neural network model weight from a regression relationship between an input sample and an output sample of the neural network model, comprising:
Determining corresponding second absolute inertial information according to second acceleration information, second angular velocity information and second geomagnetic information, wherein the second absolute inertial information is used for indicating velocity information of the target user in a second coordinate, and the second coordinate is a geocentric coordinate;
and determining the weight of the neural network model according to the regression relation between the second absolute inertia information and the second speed information.
It should be further noted that the second absolute inertial information indicates absolute inertial information of the object at the historical moment determined according to the second acceleration information, the second angular velocity information and the second geomagnetic information, specifically, the geocentric coordinates corresponding to the object at the historical moment, that is, velocity information in the second coordinates. And determining the weight of the neural network model according to the regression relation between the second absolute inertia information and the second speed information, so that training of the neural network model based on the absolute coordinate system can be completed.
From the description of the above embodiments, it will be clear to a person skilled in the art that the method according to the above embodiments may be implemented by means of software plus the necessary general hardware platform, but of course also by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) comprising several instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method of the various embodiments of the present invention.
Example 5
The present embodiment provides a positioning device, which is used to implement the foregoing embodiments and preferred embodiments, and will not be described in detail. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. While the means described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.
Fig. 11 is a block diagram of a positioning device according to an embodiment of the present invention, and as shown in fig. 11, the positioning device in this embodiment includes:
A measurement module 502, configured to obtain first measurement information and one or more sets of second measurement information; wherein the first measurement information is real-time measurement information of the target object, and the one or more sets of second measurement information are historical measurement information;
the resolving module 504 is configured to determine first location information according to the first measurement information and a preset neural network model, and locate the target object according to the first location information;
The neural network model is trained according to one or more groups of second measurement information and one or more groups of second position information, and the one or more groups of second position information is historical position information.
Other technical features and technical effects of the positioning device in this embodiment correspond to those of the positioning method in embodiment 3, and thus are not described herein.
It should be noted that each of the above modules may be implemented by software or hardware, and for the latter, it may be implemented by, but not limited to: the modules are all located in the same processor; or the above modules may be located in different processors in any combination.
Example 6
The embodiment provides a neural network model training device, which is used for implementing the foregoing embodiments and preferred embodiments, and is not described in detail. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. While the means described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.
Fig. 12 is a block diagram of a neural network model training apparatus according to an embodiment of the present invention, and as shown in fig. 12, the neural network model training apparatus in this embodiment includes:
A first obtaining module 602, configured to obtain one or more sets of second measurement information, where the one or more sets of second measurement information are obtained by an inertial measurement unit IMU;
A second obtaining module 604, configured to obtain one or more sets of second location information, where the one or more sets of second location information are obtained by an ultra wideband UWB unit;
the training module 606 is configured to train the preset neural network model according to one or more sets of second measurement information and one or more sets of second location information, so as to obtain a trained neural network model.
Other technical features and technical effects of the neural network model training device in this embodiment correspond to the neural network model training method in embodiment 4, and thus are not described herein.
It should be noted that each of the above modules may be implemented by software or hardware, and for the latter, it may be implemented by, but not limited to: the modules are all located in the same processor; or the above modules may be located in different processors in any combination.
Example 7
Embodiments of the present invention also provide a computer readable storage medium having a computer program stored therein, wherein the computer program is arranged to perform the steps of any of the method embodiments described above when run.
Alternatively, in the present embodiment, the above-described computer-readable storage medium may be configured to store a computer program for executing the steps of:
S1, acquiring measurement information of an Inertial Measurement Unit (IMU) for measuring a target object, and acquiring a motion state of the target object according to the measurement information and a preset neural network model;
the neural network model is obtained by training according to historical measurement information of the IMU for measuring the target object and the historical motion state of the target object.
Alternatively, in the present embodiment, the above-described computer-readable storage medium may include, but is not limited to: a usb disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory RAM), a removable hard disk, a magnetic disk, or an optical disk, or other various media capable of storing a computer program.
Example 8
Embodiments of the present invention also provide a computer readable storage medium having a computer program stored therein, wherein the computer program is arranged to perform the steps of any of the method embodiments described above when run.
Alternatively, in the present embodiment, the above-described computer-readable storage medium may be configured to store a computer program for executing the steps of:
S1, acquiring historical measurement information of an Inertial Measurement Unit (IMU) for measuring a target object;
s2, acquiring a historical motion state of a target object;
And S3, training a preset neural network model according to the historical measurement information and the historical motion state to obtain a trained neural network model.
Alternatively, in the present embodiment, the above-described computer-readable storage medium may include, but is not limited to: a usb disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory RAM), a removable hard disk, a magnetic disk, or an optical disk, or other various media capable of storing a computer program.
Example 9
An embodiment of the invention also provides an electronic device comprising a memory having stored therein a computer program and a processor arranged to run the computer program to perform the steps of any of the method embodiments described above.
Optionally, the electronic apparatus may further include a transmission device and an input/output device, where the transmission device is connected to the processor, and the input/output device is connected to the processor.
Alternatively, in the present embodiment, the above-described processor may be configured to execute the following steps by a computer program:
S1, acquiring measurement information of an Inertial Measurement Unit (IMU) for measuring a target object, and acquiring a motion state of the target object according to the measurement information and a preset neural network model;
the neural network model is obtained by training according to historical measurement information of the IMU for measuring the target object and the historical motion state of the target object.
Alternatively, specific examples in this embodiment may refer to examples described in the foregoing embodiments and optional implementations, and this embodiment is not described herein.
Example 10
An embodiment of the invention also provides an electronic device comprising a memory having stored therein a computer program and a processor arranged to run the computer program to perform the steps of any of the method embodiments described above.
Optionally, the electronic apparatus may further include a transmission device and an input/output device, where the transmission device is connected to the processor, and the input/output device is connected to the processor.
Alternatively, in the present embodiment, the above-described processor may be configured to execute the following steps by a computer program:
S1, acquiring historical measurement information of an Inertial Measurement Unit (IMU) for measuring a target object;
s2, acquiring a historical motion state of a target object;
And S3, training a preset neural network model according to the historical measurement information and the historical motion state to obtain a trained neural network model.
Alternatively, specific examples in this embodiment may refer to examples described in the foregoing embodiments and optional implementations, and this embodiment is not described herein.
It will be appreciated by those skilled in the art that the modules or steps of the invention described above may be implemented in a general purpose computing device, they may be concentrated on a single computing device, or distributed across a network of computing devices, they may alternatively be implemented in program code executable by computing devices, so that they may be stored in a memory device for execution by computing devices, and in some cases, the steps shown or described may be performed in a different order than that shown or described, or they may be separately fabricated into individual integrated circuit modules, or multiple modules or steps within them may be fabricated into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the principle of the present invention should be included in the protection scope of the present invention.