CN110825108A - A collaborative collision avoidance method for multiple tracking UAVs in the same airspace - Google Patents

Abstract

Translated fromChinese

本发明公开一种在同一空域内多架跟踪无人机的协同防碰撞方法，该方法中每架无人机跟踪各自的目标，由于无人机之间的航路可能会有交叉，因此本发明旨在实现无人机之间无碰撞地跟踪目标。对于每架无人机，本发明首先使用OpenCV检测移动目标，然后根据目标在机载相机所获得图像上的位置信息使用PID控制器计算无人机应该执行的速度。然后使用速度障碍模型进行碰撞检测，最后给出了一种目标导向的速度优化方法。本发明考虑到复杂空域中，无人机路径交叉情况下多机协同防碰撞问题，使用速度障碍模型将多无人机的协同防碰撞问题转化为速度空间的规划问题，考虑到实时性要求，给出了一种目标导向的速度优化方法。

The invention discloses a collaborative anti-collision method for tracking multiple drones in the same airspace. In the method, each drone tracks its own target. Since the routes between the drones may intersect, the present invention It is designed to achieve collision-free tracking of targets between drones. For each drone, the present invention first uses OpenCV to detect the moving target, and then uses the PID controller to calculate the speed that the drone should perform according to the target's position information on the image obtained by the onboard camera. Then use the velocity obstacle model for collision detection, and finally give a goal-oriented velocity optimization method. The invention takes into account the multi-machine collaborative anti-collision problem in the complex airspace and the intersection of UAV paths, uses the speed obstacle model to transform the multi-UAV cooperative anti-collision problem into the planning problem of the speed space, and takes into account the real-time requirements, A goal-oriented speed optimization method is presented.

Description

Cooperative anti-collision method for multiple tracking unmanned aerial vehicles in same airspace

Technical Field

The invention relates to the field of multi-unmanned aerial vehicle collaborative motion planning under constraint conditions, in particular to a collaborative anti-collision method for multiple tracking unmanned aerial vehicles in the same airspace.

Background

With the maturity of low altitude Unmanned Aerial Vehicle (UAV) hovering, cruising and pan-tilt technologies, as well as the rise of Computer Vision and Deep Learning (Deep Learning), target tracking based on an Unmanned Aerial Vehicle platform has become a research hotspot at home and abroad, and has been used for tracking crime vehicles in urban anti-terrorism, tracking ground maneuvering targets in air war and the like. Tracking the unmanned aerial vehicle refers to an unmanned aerial vehicle with special purposes, which acquires image information by using sensors such as an airborne pan-tilt and the like, acquires position information of a target on an image (or predicts the position information of the target by state estimation and multi-sensor information fusion) through a target detection process, and calculates a control command (usually a velocity vector) through a tracking algorithm to control the motion of the unmanned aerial vehicle so as to ensure that the target is always located near a central area of a visual field obtained by an airborne camera. However, the above process is driven only by the movement of the target, and does not take into account the environment in which the drone is located (i.e. the threat of other aircraft). With the popularity of civilian drones and the increasing complexity of the tasks performed, the routes for tracking drones may cross. Therefore, a collision detection and resolution method between unmanned aerial vehicles needs to be designed to guarantee flight safety.

The problem of cooperative collision prevention of multiple unmanned aerial vehicles relates to the problem of high-dimensional combined space generated by superposition of multiple unmanned aerial vehicle degrees of freedom, the problem of optimization, the problem of static and dynamic constraint of unmanned aerial vehicles and the like. The control architecture of the multi-unmanned aerial vehicle anti-collision system comprises a centralized structure and a distributed structure. The centralized control can obtain the planning result with high efficiency and global optimization, but the centralized control is mainly suitable for static environment and is difficult to cope with the change of the environment; in a distributed system, each unmanned aerial vehicle plans its own action according to its own environmental information, which has the advantages of being adaptable to environmental changes and having the disadvantages of not being able to obtain a global optimal solution and possibly causing a Deadlock (Deadlock) problem. At present, the multi-drone anti-collision method is mostly expanded from research results of single-robot motion planning, such as Sampling-based Methods (Sampling-based Methods), Neural Network-based Methods (Neural Network-based Methods), Fuzzy Logic (Fuzzy Logic), geometric models (such as Velocity Obstacles (Velocity), Artificial Potential Field Methods (Artificial Potential fields)), mathematical optimization models (mathematical optimization or planning Methods), Swarm Intelligence or soft computing (Swarm Intelligence, software computing), and Reinforcement Learning (relationship Learning).

Among the above methods, the speed obstacle is a Local path planning (Local path planning) model (i.e. obstacle avoidance model) commonly used by multi-robot systems, but it is assumed that the obstacle is stationary or moving at a constant speed, and the model needs to solve a complex Nonlinear Optimization Problem (Nonlinear Optimization Problem) in a Relative speed Space (Relative Velocity Space), so that the real-time property is difficult to guarantee. In addition, on tracking the unmanned aerial vehicles, the speed optimization not only needs to consider potential collision among the unmanned aerial vehicles, but also needs to consider the motion trend of the tracked target, and meanwhile needs to ensure real-time response.

Disclosure of Invention

In order to solve the problems, improve the use efficiency of an airspace, ensure the safety and smoothness of the flight of the unmanned aerial vehicle and realize the continuous tracking of a target, the invention provides a cooperative anti-collision method for multiple tracking unmanned aerial vehicles in the same airspace.

The invention is realized by the following technical scheme: a cooperative anti-collision method for multiple tracking unmanned aerial vehicles in the same airspace specifically comprises the following steps:

step 1, firstly, acquiring image information from a real-time video stream issued by an unmanned aerial vehicle sensor;

step 2, using a tracker in OpenCV to locate the center coordinate (x) of the target on the image_o，y_o)；

Step 3, the central coordinate (x) in the step 2 is compared_o，y_o) Scaling to get a normalized offset Δ_x' and Delta_y' calculating the velocity vector v of the drone using a PID algorithm for tracking the target.

Step 4, using a velocity barrier model to perform collision detection on v, namely checking whether the tail end of v is in a combined velocity barrier area, and when the tail end of v is in the combined velocity barrier area, adjusting the candidate velocity by using a target-oriented velocity optimization method to eliminate collision;

and 5, repeatedly executing the processes of the steps 1-4 by the unmanned aerial vehicle until the tracking task is completed.

Further, the method for tracking the unmanned aerial vehicle specifically comprises the following steps:

step 1: collecting image information by using equipment such as an airborne holder and the like, and returning video data to a control end;

step 2: selecting a target to be tracked by the unmanned aerial vehicle from the video data, and extracting features such as HOG, HSV and the like;

and 3, step 3: positioning the central coordinate (x) of the target on the image by using the features of HOG, HSV and the like extracted in the step 2 through a tracker of OpenCV_o，y_o) I.e. the centre of the smallest enclosing circle or the centre of the smallest enclosing rectangle;

and 4, step 4: calculating the center coordinate (x)_o，y_o) With unmanned aerial vehicle field of vision central point (x)_c，y_c) The absolute offsets Δ x and Δ y in both the horizontal and vertical directions:

and 5, step 5: setting a dead zone range, wherein the width is w, the height is h, and the center of the dead zone range is superposed with the center of the visual field of the unmanned aerial vehicle; when | x_o-x_cWhen | < w/2, the speed direction of the unmanned aerial vehicle does not need to be adjusted; when y_o-y_cWhen | < h/2, the speed of the unmanned aerial vehicle is 0, namely, the unmanned aerial vehicle does not need to advance or retreat, otherwise, the step 6 is implemented;

and 6, step 6: let the resolution of the view picture of the unmanned aerial vehicle be x_r×y_rThe offsets are first normalized, scaling Δ x and Δ y to [ -1, 1]I.e. delta_x′＝Δ_x/(x_r/2)，Δ_y′＝Δ_y/(y_r/2), then the speed of the drone at the next moment is:

v＝v_maxΔ_y′

α＝α_maxΔ_x′

v_maxmaximum speed of drone, α_max(α_maxPi/2) is the maximum rotation angle of the unmanned aerial vehicle rotating around the Z axis only,

wherein when is_y' < 0, indicating that it should be reversed, otherwise it should be advanced; similarly, when Δ_xWhen the' is less than 0, the unmanned aerial vehicle rotates leftwards, otherwise, the unmanned aerial vehicle rotates rightwards;

at this time, a new velocity vector v ═ (v, α) is generated, the direction of which is relative to the current straight ahead α, and the velocity magnitude is v;

and 7, step 7: and repeating the steps 3-6 until the tracking task is completed.

Further, step 4 comprises the following substeps:

when there are multiple tracked drones in the airspace, collision check on v is also needed.

Step 1: screening the intruders by using a collision cone model, and selecting the threateners threatening the local machine, wherein the specific process comprises the following steps:

for each at u₀Unmanned plane u in sensing range_iDefining the relative velocity v_0|i＝v₀-v_iDefining the ray:

λ(p，v)＝{p+vt|t＞0}

wherein the state information of the drone is represented by S ═ (p, v), and p ═ p (p)_x，p_y) Indicating the position of the unmanned aerial vehicle, and v indicating the speed information of the unmanned aerial vehicle; at unmanned plane u₀Has u in the sensing range₁，…，u_nAlso executing the task, defining n unmanned planes u₁，…，u_nIs u₀Is the "invader", i.e. | | p₀-p_i||₂≤ρ_SR，ρ_SRA radius representing a perception range of the drone, or a radius representing an early warning range; λ (p, v) represents a ray whose origin is in p, the direction being along the direction of v;

according to u₀Radius size of (d) to u_iIs subjected to 'puffing', i.e. r_i′＝r_i+r₀Obtaining a position obstacle PO_0|iDefine the collision cone CC_0|iComprises the following steps:

any relative velocity v_t|iIs located in CC_0|iAll cause u₀And u_iBy collision of CC_0|iScreening out the pairs u₀Threatening unmanned plane denoted u₁，…，u_m(m≤n)。

Step 2, defining a speed obstacle model:

wherein

Representing a minkowski vector sum operation, i.e.:

wherein, VO_t|iIs a set of a series of speeds that result in u₀And u_iA collision at some future time;

to avoid multiple threats, multiple VOs need to be combined:

where U denotes the combination of m VO geometric regions, so when v₀When the tail end of the motor is positioned in the VO, the speed outside the VO is selected to avoid potential collision.

Step 3, speed optimization, under the condition of ensuring that the direction of a speed vector v is unchanged, under the constraint of VO, the speed v (t +1) of the unmanned aerial vehicle at the next moment is optimized, and the specific steps are as follows:

assuming that the boundary curve of VO is Ω and the extension of velocity v is defined as λ (p, v), then:

P＝λ(p，v)∩Ω

where P denotes the intersection of the candidate speed and the VO boundary curve. Of all the intersections, the intersection P closest to the distance P is selected_iAs end of the next time velocity E_vNamely:

wherein P is_iDenotes the intersection of p with the i' th VO boundary curve.

The optimized speed of the unmanned aerial vehicle is as follows:

v′＝(v′，α)＝(||E_v-p||₂，α)。

compared with the prior art, the invention has the following beneficial effects:

(1) the method has a complete geometric basis, is commonly used for the problems of obstacle avoidance and collision avoidance of a multi-robot system, and is suitable for the application scene of the method;

(2) when a plurality of intruders exist, the collision cone model is used for screening the intruders, so that the problem scale is reduced;

(3) compared with a classical velocity barrier model, the target-oriented velocity optimization method is provided in consideration of the requirements of real-time performance and directivity of the tracking task (namely the requirement that the tracking unmanned aerial vehicle needs to keep the motion trend basically consistent with the tracked target), namely a simple method for solving the optimization problem is provided.

The method is characterized in that: the collision problem that probably takes place between them when having solved many tracking unmanned aerial vehicle flights fast high-efficiently.

Drawings

FIG. 1 is a flow chart embodying the present invention;

FIG. 2 is a diagram of an implementation of the present invention;

FIG. 3 is a diagram showing a relationship between an actual position and an ideal position of a target on a visual field picture of an airborne tripod head of an unmanned aerial vehicle;

FIG. 4 is a schematic diagram of the screening of "intruding" drones based on velocity barriers in the present invention;

FIG. 5 is a schematic diagram of a target-oriented speed optimization method introduced in the present invention.

Detailed Description

The present invention will be described in further detail with reference to the following detailed description and accompanying drawings. Fig. 1 shows a specific flowchart of a cooperative anti-collision method for multiple tracked drones in the same airspace, which specifically includes:

step 1, firstly, acquiring image information from a real-time video stream issued by an unmanned aerial vehicle sensor;

step 2, using a tracker in OpenCV to locate the center coordinate (x) of the target on the image_o，y_o)；

Step 3, the central coordinate (x) in the step 2 is compared_o，y_o) Normalized offset Δ_x' and Delta_y' calculating the velocity vector v of the drone using a PID algorithm for tracking the target.

Step 4, using a velocity barrier model to perform collision detection on v, namely checking whether the tail end of v is in a combined velocity barrier area, and when the tail end of v is in the combined velocity barrier area, adjusting the candidate velocity by using a target-oriented velocity optimization method to eliminate collision;

and 5, repeatedly executing the processes of the steps 1-4 by the unmanned aerial vehicle until the tracking task is completed.

Fig. 2 shows an implementation architecture diagram for implementing the method of the invention on a real drone platform. The bottom layer of the architecture is an unmanned aerial vehicle hardware platform which comprises a battery, a point machine, a sensor, an antenna, an onboard processor and the like, and a Linux operating system is installed on the onboard processor of the unmanned aerial vehicle. In order to realize global information sharing between drones, a Robot Operation System (ROS) is considered to be deployed on a Linux Operation System as a Meta-Operation System. In the ROS, a Master Node (Master Node) is responsible for message communication between nodes. Therefore, each unmanned aerial vehicle can be used as a node in one ROS, all unmanned aerial vehicles register own information with a main node, and each unmanned aerial vehicle broadcasts own information such as position, speed and the like through a Topic (Topic). Similarly, other drones may subscribe to the topic to obtain information for the drone. This enables asynchronous global communication between drones. On the basis of an ROS layer, the cooperative anti-collision system for the tracking unmanned aerial vehicle comprises two main functional modules, namely target tracking and cooperative anti-collision. The target tracking module mainly comprises two parts, namely target detection based on OpenCV and speed control based on PID; the cooperative collision avoidance mainly comprises a collision detection based on speed obstacle and a speed optimization part of target guidance. A detailed process description of these two modules is given below:

the specific implementation steps of the target tracking process are as follows:

step 1: collecting image information by using equipment such as an airborne holder and the like, and returning video data to a control end;

step 2: selecting a target to be tracked by the unmanned aerial vehicle from the video data, and extracting features such as HOG, HSV and the like;

and 3, step 3: locating the central coordinates (x) of the target on the image by using the features of HOG, HSV, etc. extracted in step 2 through the 'tracker' of OpenCV, such as MIL, KCF, TLD, etc_o，y_o) I.e. the center of the smallest enclosing circle or the center of the smallest enclosing rectangle, such as the center of the red circle in fig. 3;

and 4, step 4: calculating the center coordinate (x)_o，y_o) With unmanned aerial vehicle field of vision central point (x)_c，y_c) The absolute offsets Δ x and Δ y in both the horizontal and vertical directions:

and the delta x and the delta y are used for judging whether the unmanned aerial vehicle needs to carry out speed adjustment at the current moment or not and calculating a control instruction by using a PID algorithm.

And 5, step 5: in order to reduce the "jitter" problem of the drone when the target center coordinates are near the ideal value, a dead zone range is set, with width w and height h, as shown by the small rectangle in fig. 3. The center of the dead zone range is superposed with the center of the visual field of the unmanned aerial vehicle; when | x_o-x_cWhen the | is less than or equal to w/2, the speed direction of the unmanned aerial vehicle does not need to be adjusted, namely, the unmanned aerial vehicle does not need to be rotated left or right; when y_o-y_cWhen | < h/2, the speed of the unmanned aerial vehicle is 0, namely, the unmanned aerial vehicle does not need to advance or retreat, otherwise, the step 6 is implemented;

and 6, step 6: arriving at this step to illustrate that the position of the target on the image has already gone out of the dead zone range, the present invention calculates the drone speed using a PID controller, only the P control is given here as an example. Let the resolution of the view picture of the unmanned aerial vehicle be x_r×y_rThe offsets are first normalized, scaling Δ x and Δ y to [ -1, 1]I.e. delta_x′＝Δ_x/(x_r/2)，Δ_y′＝Δ_y/(y_r/2), then the speed of the drone at the next moment is:

v＝v_maxΔ_y′

α＝α_maxΔ_x′

v_maxmaximum speed of drone, α_max(α_maxPi/2) is the maximum rotation angle of the unmanned aerial vehicle rotating around the Z axis only,

wherein when is_y' < 0, indicating that it should be reversed, otherwise it should be advanced; similarly, when Δ_xWhen the' is less than 0, the unmanned aerial vehicle rotates leftwards, otherwise, the unmanned aerial vehicle rotates rightwards;

at this time, a new velocity vector v ═ (v, α) is generated, the direction of which is relative to the current straight ahead α, and the velocity magnitude is v;

and 7, step 7: and repeating the steps 3-6 until the tracking task is completed.

However, the speed generated by the above process cannot guarantee that no collision occurs between the drones, and for this reason, a synergistic collision avoidance process is given. The anti-collision process provided by the invention is based on a speed obstacle model, is commonly used for the problems of obstacle avoidance and collision avoidance of a multi-robot system, and is suitable for the application scene of the invention. The specific implementation steps of the cooperative anti-collision process are as follows:

step 1: screening the intruders by using a collision cone model, and selecting the threateners threatening the local machine, wherein the specific process comprises the following steps:

for each at u₀Unmanned plane u in sensing range_iDefining the relative velocity v_0|i＝v₀-v_iDefining the ray:

λ(p，v)＝{p+vt|t＞0}

wherein the state information of the drone is represented by S ═ (p, v), and p ═ p (p)_x，p_y) Indicating the position of the unmanned aerial vehicle, and v indicating the speed information of the unmanned aerial vehicle; at unmanned plane u₀Has u in the sensing range₁，…，u_nAlso executing the task, defining n unmanned planes u₁，…，u_nIs u₀Is the "invader", i.e. | | p₀-p_i||₂≤ρ_SR，ρ_SRA radius representing a perception range of the drone, or a radius representing an early warning range; λ (p, v) represents a ray whose origin is in p, the direction being along the direction of v;

according to u₀Radius size of (d) to u_iIs subjected to 'puffing', i.e. r_i′＝r_i+r₀Obtaining a position obstacle PO_0|iDefine the collision cone CC_0|iComprises the following steps:

any relative velocity v_t|iIs located in CC_0|iAll cause u₀And u_iSuch as the shaded portion on the right in fig. 4. By CC_0|iScreening out the pairs u₀Threatening unmanned plane denoted u₁，…，u_m(m≤n)。

Step 2, combining potential collisions: in order to directly use the absolute velocity v₀Performing collision detection, and proposing a speed obstacle model;

wherein

Representing a minkowski vector sum operation, i.e.:

wherein, VO_t|iIs a set of a series of speeds that result in u₀And u_iA collision at some future time;

to avoid multiple threats, multiple VOs need to be combined, as in fig. 5 a VO consists of two speed barriers:

where U denotes the combination of m VO geometric regions, so when v₀When the tail end of the motor is positioned in the VO, the speed outside the VO is selected to avoid potential collision.

Step 3, collision handling (speed optimization). In the classical velocity barrier model, the following optimization problem is solved

The speed of the robot is adjusted. However, the process is time-consuming, and therefore, the method for optimizing the target-oriented speed is provided, and under the condition that the direction of the speed vector v is guaranteed to be unchanged to the maximum extent and under the constraint of the VO, the speed v (t +1) of the unmanned aerial vehicle at the next moment is optimized, and the method specifically comprises the following steps:

assuming that the boundary curve of VO is Ω and the extension of velocity v is defined as λ (p, v), then:

P＝λ(p，v)∩Ω

where P denotes the intersection of the candidate speed and the VO boundary curve. Of all the intersections, the intersection P closest to the distance P is selected_iAs end of the next time velocity E_vNamely:

wherein P is_iDenotes the intersection of p with the i' th VO boundary curve.

The optimized speed of the unmanned aerial vehicle is as follows:

v′＝(v′，α)＝(||E_v-p||₂，α)。

the present invention is not limited to the above-described embodiments, and those skilled in the art can implement the present invention in other various embodiments based on the disclosure of the present invention. Therefore, the design of the invention is within the scope of protection, with simple changes or modifications, based on the design structure and thought of the invention.

Claims (3)

1. A cooperative anti-collision method for multiple tracking unmanned aerial vehicles in the same airspace is characterized by comprising the following steps:

step 1, firstly, acquiring image information from a real-time video stream issued by an unmanned aerial vehicle sensor;

step 2, using a tracker in OpenCV to locate the center coordinate (x) of the target on the image_o，y_o)；

Step 3, the central coordinate (x) in the step 2 is compared_o，y_o) Scaling to get a normalized offset Δ_x' and Delta_y' calculating the velocity vector v of the drone using a PID algorithm for tracking the target.

Step 4, using a velocity barrier model to perform collision detection on v, namely checking whether the tail end of v is in a combined velocity barrier area, and when the tail end of v is in the combined velocity barrier area, adjusting the candidate velocity by using a target-oriented velocity optimization method to eliminate collision;

and 5, repeatedly executing the processes of the steps 1-4 by the unmanned aerial vehicle until the tracking task is completed.

2. The cooperative anti-collision method according to claim 1, wherein the method for tracking the drone is specifically:

step 1: collecting image information by using equipment such as an airborne holder and the like, and returning video data to a control end;

step 2: selecting a target to be tracked by the unmanned aerial vehicle from the video data, and extracting features such as HOG, HSV and the like;

and 3, step 3: positioning the central coordinate (x) of the target on the image by using the features of HOG, HSV and the like extracted in the step 2 through a tracker of OpenCV_o，y_o) I.e. the centre of the smallest enclosing circle or the centre of the smallest enclosing rectangle;

and 4, step 4: calculating the center coordinate (x)_o，y_o) With unmanned aerial vehicle field of vision central point (x)_c，y_c) The absolute offsets Δ x and Δ y in both the horizontal and vertical directions:

and 5, step 5: setting a dead zone range, wherein the width is w, the height is h, and the center of the dead zone range is superposed with the center of the visual field of the unmanned aerial vehicle; when | x_o-x_cWhen | < w/2, the speed direction of the unmanned aerial vehicle does not need to be adjusted; when y_o-y_cWhen | < h/2, the speed of the unmanned aerial vehicle is 0, namely, the unmanned aerial vehicle does not need to advance or retreat, otherwise, the step 6 is implemented;

and 6, step 6: let the resolution of the view picture of the unmanned aerial vehicle be x_r×y_rThe offsets are first normalized, scaling Δ x and Δ y to [ -1, 1]I.e. delta_x′＝Δ_x/(x_r/2)，Δ_y′＝Δ_y/(y_r/2), then the speed of the drone at the next moment is:

v＝v_maxΔ_y′

α＝α_maxΔ_x′

v_maxmaximum speed of drone, α_max(α_maxPi/2) is the maximum rotation angle of the unmanned aerial vehicle rotating around the Z axis only,

wherein when is_y′<0, indicating that it should be backed up, otherwise it should be advanced;similarly, when Δ_x′<When 0, the unmanned aerial vehicle rotates leftwards, otherwise, the unmanned aerial vehicle rotates rightwards;

at this time, a new velocity vector v ═ (v, α) is generated, the direction of which is relative to the current straight ahead α, and the velocity magnitude is v;

and 7, step 7: and repeating the steps 3-6 until the tracking task is completed.

3. The cooperative anti-collision method according to claim 1, wherein step 4 comprises the following sub-steps:

when there are multiple tracked drones in the airspace, collision check on v is also needed.

Step 1: screening the intruders by using a collision cone model, and selecting the threateners threatening the local machine, wherein the specific process comprises the following steps:

for each at u₀Unmanned plane u in sensing range_iDefining the relative velocity v_0|i＝v₀-v_iDefining the ray:

λ(p，v)＝{p+vt|t>0}

wherein the state information of the drone is represented by S ═ (p, v), and p ═ p (p)_x，p_y) Indicating the position of the unmanned aerial vehicle, and v indicating the speed information of the unmanned aerial vehicle; at unmanned plane u₀Has u in the sensing range₁，…，u_nAlso executing the task, defining n unmanned planes u₁，…，u_nIs u₀Is the "invader", i.e. | | p₀-p_i||₂≤ρ_SR，ρ_SRA radius representing a perception range of the drone, or a radius representing an early warning range; λ (p, v) represents a ray whose origin is in p, the direction being along the direction of v;

according to u₀Radius size of (d) to u_iIs subjected to 'puffing', i.e. r_i′＝r_i+r₀Obtaining a position obstacle PO_0|iDefine the collision cone CC_0|iComprises the following steps:

any relative velocity v_t|iIs located in CC_0|iAll cause u₀And u_iBy collision of CC_0|iScreening out the pairs u₀Threatening unmanned plane denoted u₁，…，u_m(m≤n)。

Step 2, defining a speed obstacle model:

wherein

Representing a minkowski vector sum operation, i.e.:

wherein, VO_t|iIs a set of a series of speeds that result in u₀And u_iA collision at some future time; to avoid multiple threats, multiple VOs need to be combined:

wherein ∪ denotes the combination of m VO geometric regions, so when v is₀When the tail end of the motor is positioned in the VO, the speed outside the VO is selected to avoid potential collision.

Step 3, speed optimization, under the condition of ensuring that the direction of a speed vector v is unchanged, under the constraint of VO, the speed v (t +1) of the unmanned aerial vehicle at the next moment is optimized, and the specific steps are as follows:

assuming that the boundary curve of VO is Ω and the extension of velocity v is defined as λ (p, v), then:

P＝λ(p，v)∩Ω

where P denotes the intersection of the candidate speed and the VO boundary curve. In all intersections, distances are selectedThe intersection point P nearest to P_iAs end of the next time velocity E_vNamely:

wherein P is_iDenotes the intersection of p with the i' th VO boundary curve.

The optimized speed of the unmanned aerial vehicle is as follows:

v′＝(v′，α)＝(||E_v-p||₂，α)。

Images