A multi-view perception oriented unmanned vehicle cluster deployment method and system

Abstract

A multi-view perception-oriented unmanned vehicle cluster deployment method and system, the method divides the target area into a limited cell grid through a cluster simulation deployment algorithm based on artificial potential field; the unmanned vehicle is divided into a limited movable direction and a field of view angle direction; the cell potential field value is updated according to the field of view range of the unmanned vehicle; the potential field moving direction is calculated according to the average value of the cell potential field value in the field of view angle direction; the repulsion moving direction is calculated according to the repulsion force between the unmanned vehicles; the moving direction of the next time step is calculated according to the repulsion moving direction, the potential field moving direction and the moving direction of the current time step, and the field of view angle direction of the next time step is taken as the field of view angle with the largest number of cells in the field of view range; so that the field of view angle direction covers the geometric center of the cluster position or rotates towards the cluster geometric center, the optimal observation position and observation field of view angle of the cluster are quickly generated, and the multi-view cooperative perception efficiency and deployment efficiency of the cluster are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned vehicle swarm technology, and in particular to a method for deploying unmanned vehicle swarms, especially a method for deploying unmanned vehicle swarms oriented towards multi-view perception. Background Technology

[0002] Unmanned vehicle swarms have significant application value in smart cities, target monitoring, and emergency rescue. By carrying visual sensors such as cameras, swarms can observe target areas from multiple perspectives from different locations, addressing the problem of target obstruction in challenging environments. However, how to efficiently deploy swarms of unmanned vehicles to both avoid the impact of obstacles on vision in complex environments and maximize field of view coverage, thereby achieving efficient and accurate multi-view collaborative perception, remains a pressing issue.

[0003] Currently, traditional deployment methods rely on manually specifying the location of unmanned vehicles or allocating their locations based on empirical rules, ignoring the impact of obstacles on field of vision. This results in unreasonable observation positions and angles for the cluster, creating blind spots in monitoring, low collaborative perception efficiency of the unmanned vehicle cluster, and low deployment efficiency of the unmanned vehicle cluster. Summary of the Invention

[0004] This disclosure provides a method for deploying unmanned vehicle clusters based on multi-view perception, so as to improve the multi-view collaborative perception efficiency and deployment efficiency of unmanned vehicle cluster systems.

[0005] The method for deploying autonomous vehicle swarms based on multi-view perception disclosed herein mainly includes the following steps:

[0006] S1: Define the target perception area;

[0007] S2: Deploy a cluster of unmanned vehicles at the boundary of the target perception area;

[0008] S3: Obtain the precise coordinates of each autonomous vehicle through the positioning module of the autonomous vehicle cluster;

[0009] S4: Based on the precise coordinates of each unmanned vehicle and the set target perception area, a cluster simulation deployment algorithm based on artificial potential field is adopted to simulate each unmanned vehicle as a point mass, and generate the optimal observation position and observation field of view of the unmanned vehicle cluster in the target perception area.

[0010] S5: Assign the generated optimal observation position and observation field of view to each unmanned vehicle in the unmanned vehicle cluster.

[0011] Furthermore, in step S4, the cluster simulation deployment algorithm based on artificial potential fields specifically includes the following steps:

[0012] S41: Divide the target perception area into a finite number of cell grids with potential field attributes, with each cell having an initial potential field value of 0; mark cells containing obstacles as impassable areas with no potential field value attribute;

[0013] S42: Divide both the autonomous vehicle's movable direction and field of view into N. d Angle θ between the movable direction and the field of view direction. d Represented as:

[0014]

[0015] Where d is the angle index;

[0016] The angular range covered by the field of view of an autonomous vehicle can be represented as [θ]. d -Δθ,θ d +Δθ], where Δθ is half of the field of view;

[0017] The field of view length is denoted as L;

[0018] S43: For driverless car i, calculate the total repulsive force between driverless car i and other driverless cars, expressed as:

[0019]

[0020] Where i and j are the indices of the autonomous vehicles, i = 1, 2, ..., N, j = 1, 2, ..., N, and N is the number of autonomous vehicles in the cluster, x i (t),y i (t) represents the coordinates of the autonomous vehicle i, x j (t),y j (t) represents the coordinates of the autonomous vehicle j, and d ij The distance between driverless vehicles i and j is expressed as:

[0021]

[0022] The repulsive movement direction of driverless car i is represented as:

[0023]

[0024] S44: For the autonomous vehicle i, for each field of view direction θ d Calculate the average potential field value of all cells within the angular range covered by this direction, and take the direction with the smallest average value as the potential field movement direction of the unmanned vehicle i.

[0025] S45: Set the inertial movement coefficient w so that the unmanned vehicle considers the movement direction of the current time step in each movement; the unmanned vehicle determines the movement direction of the next time step based on the potential field movement direction, the repulsive movement direction, and the movement direction of the current time step, then the movement direction of unmanned vehicle i in the next time step is... Represented as:

[0026]

[0027] in, I represents the direction of movement at the current time step. p Let the probability of success be p p Bernoulli distribution, p p The weight of the direction of potential field movement, I r =1-I p f d (·) represents the function for calculating the direction of movement, and normalize[·] represents the function for normalizing the direction of movement.

[0028] S46: The unmanned vehicle i moves one unit length according to the direction of movement in the next time step;

[0029] S47: Calculate the geometric center of the autonomous vehicle swarm. Represented as:

[0030]

[0031] S48: For driverless vehicle i, calculate in θ d The number of cells within the field of view under the field of view angle, where d = 1, 2, ..., N d The field of view angle with the largest number of cells is taken as the field of view direction for the next time step. Represented as:

[0032]

[0033] Where C(θ) represents the number of cells within the field of view when the field of view angle is θ;

[0034] S49: For driverless vehicle i, determine whether the geometric center of the driverless vehicle cluster is covered within the angle range of the field of view direction in the next time step;

[0035] If so, switch the field of view to the direction of the field of view in the next time step;

[0036] Otherwise, rotate the field of view towards the geometric center of the autonomous vehicle cluster. angle;

[0037] S4a: Update the potential field value of the cell grid according to the field of view of the autonomous vehicle cluster;

[0038] S4b: For each autonomous vehicle, repeat steps S43 to S49; if for S1 consecutive iterations, the moving distance of autonomous vehicle i is less than the threshold M. stop Then the stopping position of the unmanned vehicle i is iteratively optimized;

[0039] S4c: For each autonomous vehicle, repeat steps S47 to S49; if for S2 consecutive iterations, the change in the field of view of autonomous vehicle i is less than θ. stop If this happens, then the autonomous vehicle i will stop iterative optimization of the field of view direction;

[0040] S4d: Iterates the stopping position and field of view direction of all unmanned vehicles, and outputs the final position and field of view direction of the unmanned vehicle cluster as the optimal observation position and observation field of view generated by the simulation.

[0041] Furthermore, in step S4a, the method for updating the potential field value of the cell grid includes:

[0042] The potential field value of a cell is updated based on whether it is within the field of view of the autonomous vehicle and the distance from the cell to the autonomous vehicle: For cells within the field of view of the autonomous vehicle, the potential field value is obtained according to a Gaussian distribution, and the closer the cell is to the autonomous vehicle, the larger the potential field value of the cell; For cells outside the field of view of the autonomous vehicle cluster, the potential field value is 0; For cells that are outside the field of view of the autonomous vehicle cluster due to occlusion by obstacles, the potential field value is 0 and they are not included in the field of view.

[0043] Meanwhile, the potential field value of a cell is superimposed by the influence of the autonomous vehicle cluster. That is, if the field of vision of different autonomous vehicles covers the same cell, the potential field value of that cell will be superimposed.

[0044] Furthermore, the method also includes the following steps:

[0045] S6: The unmanned vehicle cluster moves to the designated optimal observation position and rotates to the optimal observation field of view to complete the deployment, and carries out multi-view collaborative perception of the target area through sensor modules.

[0046] The multi-view perception-oriented unmanned vehicle swarm system applying the above method mainly includes: unmanned vehicle-mounted equipment and a task controller, wherein:

[0047] The unmanned vehicle-mounted equipment includes: an onboard computer, an onboard ad-hoc network module, a sensor module, a positioning module, and a power supply, all configured on each unmanned vehicle in the cluster;

[0048] The onboard computer is connected to the self-organizing network module, sensor module and positioning module to control the networking communication, positioning and multi-view collaborative perception functions of the unmanned vehicle cluster;

[0049] The vehicle-mounted self-organizing network module is connected to the vehicle-mounted computer and is used to send and receive data, enabling the unmanned vehicle to join the cluster system communication network;

[0050] The sensor module is connected to the onboard computer to enable multi-view collaborative perception of the unmanned vehicle cluster;

[0051] The positioning module is connected to the onboard computer and is used to obtain the positioning coordinates of the unmanned vehicle;

[0052] The task controller includes: an ad hoc network module, a task computer, a power supply, and an interaction module;

[0053] The self-organizing network module is connected to the task computer and is used to send and receive data, enabling the task controller to join the cluster system communication network.

[0054] The task computer is connected to the self-organizing network module and the interaction module, providing computing power, video memory, memory, and storage space computing resources, and running the cluster simulation deployment algorithm based on the artificial potential field;

[0055] The interaction module is used for human interaction and to define the target perception area.

[0056] Compared with the prior art, the beneficial effects of this disclosure are:

[0057] ① By using a cluster simulation deployment algorithm based on artificial potential fields, the target perception area is divided into a finite number of cell grids, and a finite number of movable directions and field of view directions are assigned to the unmanned vehicles. The cell potential field value is updated according to the field of view of the unmanned vehicles. The potential field movement direction is calculated based on the average value of the potential field values ​​of all cells in the field of view direction. The repulsive movement direction is calculated based on the repulsive force between the unmanned vehicles. The movement direction of the next time step is calculated based on the repulsive movement direction, the potential field movement direction, and the movement direction of the current time step. The potential field value of cells blocked by obstacles is 0, and cells blocked by obstacles are not counted in the number of cells within the field of view. The field of view angle with the largest number of cells within the field of view is taken as the field of view direction of the next time step, so that the field of view direction covers the geometric center of the cluster position or rotates towards the geometric center of the cluster. This can quickly generate the optimal observation position and observation field of view of the unmanned vehicle cluster, increase the field of view coverage of the unmanned vehicle cluster, reduce the occlusion of the cluster's field of view by obstacles, increase the overlapping area of ​​the field of view of the unmanned vehicle cluster, and increase the dispersion range of the unmanned vehicle cluster, thereby improving the multi-view collaborative perception efficiency and deployment efficiency of the unmanned vehicle cluster system.

[0058] ② The target perception area is defined through the interactive interface of the task controller. The precise coordinates of each unmanned vehicle are obtained through the positioning module. The task controller inputs the precise coordinates of each unmanned vehicle and the target perception area into the cluster simulation deployment algorithm based on artificial potential field in the task computer. The algorithm generates the optimal observation position and observation field of view of the unmanned vehicle cluster in the target perception area. The generated optimal observation position and observation field of view are allocated and sent to the unmanned vehicle cluster, so that the unmanned vehicle cluster can move to the specified position and angle. This can quickly complete the deployment of unmanned vehicle clusters for multi-view perception, thereby improving the multi-view collaborative perception efficiency and deployment efficiency of the unmanned vehicle cluster system.

[0059] ③ By using a cluster simulation deployment algorithm based on artificial potential field, the cell potential field value is updated according to the field of view of the unmanned vehicle, the potential field movement direction is calculated according to the average value of the potential field values ​​of all cells in the field of view direction, the repulsion movement direction is calculated according to the repulsion force between unmanned vehicles, and the movement direction of the next time step is calculated according to the repulsion movement direction, the potential field movement direction and the movement direction of the current time step. This generates the optimal observation position and observation field of view of the unmanned vehicle cluster, which can increase the field of view coverage of the unmanned vehicle cluster and thus improve the multi-view collaborative perception efficiency of the unmanned vehicle cluster system.

[0060] ④ By using a cluster simulation deployment algorithm based on artificial potential field, the potential field value of cells occluded by obstacles is made to be 0, and cells occluded by obstacles are not counted in the number of cells within the field of view. The potential field movement direction is calculated based on the average value of the potential field values ​​of all cells in the field of view direction. The field of view angle with the largest number of cells within the field of view is taken as the field of view angle direction of the next time step, generating the optimal observation position and observation field of view angle of the unmanned vehicle cluster. This can reduce the occlusion of the cluster's field of view by obstacles, increase the field of view coverage of the unmanned vehicle cluster, and thus improve the multi-view collaborative perception efficiency of the unmanned vehicle cluster system.

[0061] ⑤ By using a cluster simulation deployment algorithm based on artificial potential field, the target perception area is divided into a finite number of cell grids, and a finite number of movable directions and field of view directions are assigned to the unmanned vehicle. The optimal observation position and field of view of the unmanned vehicle cluster are obtained through a finite number of iterations and optimizations. The unmanned vehicle is simulated as a point mass, which can quickly generate the optimal observation position and field of view of the unmanned vehicle cluster, thereby improving the deployment efficiency of multi-view collaborative perception of the unmanned vehicle cluster system.

[0062] ⑥ By using a cluster simulation deployment algorithm based on artificial potential field, the field of view can cover the geometric center of the cluster location or rotate towards the geometric center of the cluster, generating the optimal observation position and field of view of the unmanned vehicle cluster. This can increase the overlapping area of ​​the field of view of the unmanned vehicle cluster, thereby improving the multi-view collaborative perception efficiency of the unmanned vehicle cluster system.

[0063] ⑦ By using a cluster simulation deployment algorithm based on artificial potential field, the repulsion movement direction is calculated according to the repulsion force between unmanned vehicles, and the optimal observation position and observation field of view of the unmanned vehicle cluster are generated. This can increase the dispersion range of the unmanned vehicle cluster, thereby improving the multi-view collaborative perception efficiency of the unmanned vehicle cluster system. Attached Figure Description

[0064] The above and other objects, features and advantages of this disclosure will become more apparent from the more detailed description of exemplary embodiments of this disclosure taken in conjunction with the accompanying drawings, in which the same reference numerals generally represent the same components.

[0065] Figure 1 This is a schematic diagram of an exemplary unmanned vehicle cluster system structure according to this disclosure;

[0066] Figure 2 This is a schematic diagram illustrating the cluster simulation deployment algorithm based on artificial potential fields according to this disclosure;

[0067] Figure 3 A flowchart of the method for deploying unmanned vehicle swarms based on multi-view perception according to this disclosure;

[0068] Figure 4 This is a flowchart of the cluster simulation deployment algorithm based on artificial potential fields. Detailed Implementation

[0069] Preferred embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0070] This disclosure provides a method for deploying unmanned vehicle swarms for multi-view perception and an unmanned vehicle swarm system applying this method.

[0071] In one exemplary implementation:

[0072] An autonomous vehicle swarm system for multi-view perception is shown in the attached figure. Figure 1 As shown, it mainly includes: unmanned vehicle clusters and task controllers;

[0073] (1) The unmanned vehicle cluster consists of multiple unmanned vehicles, each of which includes an onboard computer, a self-organizing network module, a sensor module, a positioning module and a power supply.

[0074] The onboard computer is connected to the self-organizing network module, sensor module and positioning module. As the core module of unmanned vehicle control, it controls the networking communication, positioning and multi-view collaborative perception functions of the unmanned vehicle cluster.

[0075] The self-organizing network module connects to the onboard computer, enabling the unmanned vehicle to join the cluster system communication network for sending and receiving data.

[0076] The sensor module is connected to the onboard computer for multi-view collaborative perception of the unmanned vehicle swarm.

[0077] The positioning module is connected to the onboard computer and is used to obtain the positioning coordinates of the unmanned vehicle.

[0078] The power supply provides power to the autonomous vehicle system.

[0079] (2) The task controller consists of a self-organizing network module, a task computer, a power supply and an interactive interface.

[0080] The self-organizing network module connects to the task computer, enabling the task controller to join the cluster system communication network for sending and receiving data.

[0081] The task computer is connected to the self-organizing network module and the interactive interface, providing computing resources such as computing power, video memory, memory, and storage space, and running a cluster simulation deployment algorithm based on an artificial potential field.

[0082] The power supply provides power to the task controller system.

[0083] The interactive interface is used for manual interaction to delineate the target perception area.

[0084] Figure 2-4 The diagram below illustrates the deployment process and method of an unmanned vehicle cluster for multi-view perception in this embodiment, mainly including the following steps:

[0085] Step 1: Define the target perception area through the interactive interface of the task controller.

[0086] Step 2: Deploy the cluster of unmanned vehicles at the boundary of the target perception area.

[0087] Step 3: The unmanned vehicle cluster obtains the precise coordinates of each unmanned vehicle through the positioning module and sends them to the task controller through the self-organizing network module.

[0088] Step 4: The task controller reads the precise coordinates of each unmanned vehicle through the self-organizing network module. In the task computer, the precise coordinates of each unmanned vehicle and the target perception area are input into the cluster simulation deployment algorithm based on the artificial potential field.

[0089] Step 5: Run the cluster simulation deployment algorithm based on artificial potential field, treat each unmanned vehicle as a point mass for simulation, and generate the optimal observation position and observation field of view of the unmanned vehicle cluster in the target perception area.

[0090] Step 5-1: Divide the target perception area into a finite number of cell grids with potential field properties to reduce algorithm complexity and improve the speed at which the algorithm generates the optimal observation position and observation field of view for the autonomous vehicle cluster. The initial potential field value of each cell is 0.

[0091] Cells containing obstacles are marked as impassable areas and have no potential field value attribute. The magnitude of the cell's potential field value varies depending on whether it is within the autonomous vehicle's field of view and its distance from the vehicle. For cells within the autonomous vehicle's field of view, the potential field value is obtained according to a Gaussian distribution; the closer the cell is to the vehicle, the larger the potential field value. For cells outside the autonomous vehicle cluster's field of view, the potential field value is 0. Cells that are outside the autonomous vehicle cluster's field of view due to obstacle occlusion have a potential field value of 0 and are not included in the field of view.

[0092] The potential field value of a cell is superimposed by the influence of the autonomous vehicle cluster. That is, if the field of vision of different autonomous vehicles covers the same cell, the potential field value of that cell will be superimposed.

[0093] Step 5-2: Divide the autonomous vehicle's movable direction and field of view into N. d With a limited number of movable directions and field-of-view directions, the algorithm's complexity is reduced, and the speed at which it generates optimal observation positions and field-of-view angles for the autonomous vehicle swarm is improved. The angle θ between the movable directions and the field-of-view directions... d It can be represented as:

[0094]

[0095] Where d is the angle index.

[0096] The angular range covered by the field of view of an autonomous vehicle can be expressed as [θ]. d -Δθ,θ d +Δθ], where Δθ is half of the field of view. The field of view length can be expressed as L.

[0097] Step 5-3: For driverless car i, calculate the total repulsive force between driverless car i and other driverless cars, which can be expressed as:

[0098]

[0099] Where i and j are the indices of the autonomous vehicles, i = 1, 2, ..., N, j = 1, 2, ..., N, and N is the number of autonomous vehicles in the cluster, x i (t),y i (t) represents the coordinates of the autonomous vehicle i, x j (t),y j (t) represents the coordinates of the autonomous vehicle j, and d ij The distance between driverless vehicles i and j can be expressed as:

[0100]

[0101] The repulsive movement direction of the unmanned vehicle i can be expressed as:

[0102]

[0103] Calculating the repulsive movement direction based on the repulsive force between autonomous vehicles can increase the dispersion range of the autonomous vehicle swarm, thereby improving the multi-view collaborative perception efficiency of the autonomous vehicle swarm system.

[0104] Step 5-4: For driverless vehicle i, for each field of view direction θ d Calculate the average potential field value of all cells within the angular range covered by this direction, and take the direction with the smallest average value as the potential field movement direction of the unmanned vehicle i. The cell potential field value is updated based on the cluster's field of view and obstacles. Introducing obstacle factors into the movement of unmanned vehicles can reduce the occlusion of the cluster's field of view by obstacles, increase the field of view coverage of the unmanned vehicle cluster, and thus improve the multi-view collaborative perception efficiency of the unmanned vehicle cluster system.

[0105] Step 5-5: Set the inertial movement coefficient w so that the autonomous vehicle considers the direction of movement in the current time step each time it moves. The autonomous vehicle determines the direction of movement in the next time step based on the potential field direction of movement, the repulsive direction of movement, and the direction of movement in the current time step. Therefore, the direction of movement of autonomous vehicle i in the next time step is... It can be represented as:

[0106]

[0107] in, I represents the direction of movement at the current time step. p Let the probability of success be p p Bernoulli distribution, p p The weight of the direction of potential field movement, I r =1-I p f d (·) represents the function for calculating the direction of movement, and normalize[·] represents the function for normalizing the direction of movement.

[0108] Calculating the movement direction of the next time step based on the repulsion movement direction, the potential field movement direction, and the movement direction of the current time step can increase the field of view coverage of the unmanned vehicle swarm, thereby improving the multi-view collaborative perception efficiency of the unmanned vehicle swarm system.

[0109] Steps 5-6: The unmanned vehicle i moves one unit length according to the direction of movement in the next time step.

[0110] Steps 5-7: Calculate the geometric center of the autonomous vehicle swarm It can be represented as:

[0111]

[0112] Steps 5-8: For driverless vehicle i, calculate θ d The number of cells within the field of view under the field of view angle, where d = 1, 2, ..., N d The field of view angle with the largest number of cells is taken as the field of view direction for the next time step. It can be represented as:

[0113]

[0114] Where C(θ) represents the number of cells within the field of view when the field of view angle is θ.

[0115] The number of cells is calculated based on the cluster's field of view and obstacles. The field of view optimization of the unmanned vehicle by incorporating obstacle factors can reduce the occlusion of the cluster's field of view by obstacles, increase the field of view coverage of the unmanned vehicle cluster, and thus improve the multi-view collaborative perception efficiency of the unmanned vehicle cluster system.

[0116] Steps 5-9: For driverless vehicle i, determine whether the angle range of the field of view direction in the next time step covers the geometric center of the driverless vehicle cluster. If yes, switch the field of view to the field of view direction of the next time step; otherwise, rotate the field of view direction towards the geometric center of the driverless vehicle cluster. Angle. By aligning the field of view with the geometric center of the cluster or rotating it toward the cluster's geometric center, the overlap area of ​​the autonomous vehicle cluster's field of view can be increased, thereby improving the multi-view collaborative perception efficiency of the autonomous vehicle cluster system.

[0117] Steps 5-10: Update the potential field value of the cell grid according to the field of view of the autonomous vehicle cluster.

[0118] Step 5-11: For each autonomous vehicle, repeat steps 5-3 to 5-9. If for S1 consecutive iterations, the distance traveled by autonomous vehicle i is less than the threshold M... stop Then the stopping position of the unmanned vehicle i is iteratively optimized.

[0119] Step 5-12: For each autonomous vehicle, repeat steps 5-7 to 5-9. If, after S2 consecutive iterations, the change in the field of view of autonomous vehicle i is less than θ... stop If the unmanned vehicle i stops its iterative optimization of the field of view direction, then the unmanned vehicle i will stop.

[0120] Steps 5-13: Iterate the stopping position and field of view direction of all unmanned vehicles, and output the final position and field of view direction of the unmanned vehicle cluster as the optimal observation position and observation field of view generated by the simulation.

[0121] Step 6: Assign the generated optimal observation position and field of view to the unmanned vehicle cluster, and send the optimal observation position and field of view assigned to each unmanned vehicle to each unmanned vehicle through the self-organizing network module.

[0122] Step 7: The unmanned vehicle cluster moves to the designated optimal observation position and rotates to the optimal observation field of view to complete the deployment. It then conducts multi-view collaborative perception of the target area through the sensor module.

[0123] The multi-view perception-oriented unmanned vehicle swarm deployment method adopted in this embodiment divides the target perception area into a finite number of cell grids using a swarm simulation deployment algorithm based on artificial potential fields. This divides the unmanned vehicle into a finite number of movable directions and field-of-view directions. The cell potential field value is updated based on the unmanned vehicle's field of view. The potential field movement direction is calculated based on the average of all cell potential field values ​​in the field-of-view direction. The repulsive movement direction is calculated based on the repulsive force between unmanned vehicles. Finally, the movement direction for the next time step is calculated based on the repulsive movement direction, the potential field movement direction, and the movement direction at the current time step. This ensures that vehicles obstructed by obstacles... The cell potential field value is 0. Cells obstructed by obstacles are not counted in the number of cells within the field of view. The field of view angle with the largest number of cells within the field of view is taken as the field of view angle direction for the next time step. This ensures that the field of view angle direction covers the geometric center of the cluster location or rotates towards the geometric center of the cluster. This can quickly generate the optimal observation position and observation field of view of the unmanned vehicle cluster, increase the field of view coverage of the unmanned vehicle cluster, reduce the obstruction of the cluster's field of view by obstacles, increase the overlapping area of ​​the field of view of the unmanned vehicle cluster, and increase the dispersion range of the unmanned vehicle cluster, thereby improving the multi-view collaborative perception efficiency and deployment efficiency of the unmanned vehicle cluster system.

[0124] The above technical solutions are merely exemplary embodiments of the present invention. For those skilled in the art, based on the application methods and principles disclosed in the present invention, it is easy to make various types of improvements or modifications, and not limited to the methods described in the specific embodiments of the present invention. Therefore, the methods described above are merely preferred and not restrictive.

Claims

1. A method for deploying an unmanned vehicle swarm oriented towards multi-view perception, comprising the following steps: S1: Define the target perception area; S2: Deploy a cluster of unmanned vehicles at the boundary of the target perception area; S3: Obtain the precise coordinates of each autonomous vehicle through the positioning module of the autonomous vehicle cluster; S4: Based on the precise coordinates of each unmanned vehicle and the set target perception area, a cluster simulation deployment algorithm based on artificial potential field is adopted to simulate each unmanned vehicle as a point mass, and generate the optimal observation position and observation field of view of the unmanned vehicle cluster in the target perception area. S5: Assign the generated optimal observation position and observation field of view to each unmanned vehicle in the unmanned vehicle cluster; In step S4, the cluster simulation deployment algorithm based on artificial potential field specifically includes the following steps: S41: Divide the target perception area into a finite number of cell grids with potential field attributes, with each cell having an initial potential field value of 0; mark cells containing obstacles as impassable areas with no potential field value attribute; S42: Divide the autonomous vehicle's movable direction and field of view into... N d Angles in the movable direction and field of view direction. θ d Represented as: in, d Angle index; The angular range covered by the field of view of the autonomous vehicle can be shown as follows: ,in, It is half the range of the field of view; Field of view length is expressed as L ; S43: For driverless cars i Calculate driverless cars i The sum of the repulsive forces between the vehicle and other autonomous vehicles is expressed as: in, i and j For indexing autonomous vehicles, i = 1, 2, …, N , j = 1, 2, …, N , N The number of driverless cars in the cluster. x i ( t ), y i ( t For driverless cars i coordinates x j ( t ), y j ( t For driverless cars j coordinates d ij For driverless cars i and j The distance between them is expressed as: Then driverless car i The direction of repulsive movement is represented as: ； S44: For driverless cars i For each field of view direction θ d Calculate the average potential field value of all cells within the angular range covered by this direction, and take the direction with the smallest average value as the autonomous vehicle's direction. i direction of potential field movement ; S45: Set the inertial movement coefficient w This allows the autonomous vehicle to consider the direction of movement at the current time step during each movement; the autonomous vehicle determines the direction of movement at the next time step based on the potential field direction, the repulsive direction, and the direction of movement at the current time step. i Direction of movement in the next time step Represented as: in, The direction of movement at the current time step. The probability of success is Bernoulli distribution, That is, the weight of the direction of potential field movement. , The function represents the direction of movement. This represents the normalization function for the direction of movement; S46: Driverless car i Move one unit length according to the direction of movement in the next time step; S47: Calculate the geometric center of the autonomous vehicle swarm. , represented as: S48: For driverless cars i Calculated in The number of cells within the field of view under the field of view angle, of which The field of view angle with the largest number of cells is taken as the field of view direction for the next time step. , represented as: in, C ( θ ) indicates the field of view is θ The number of cells within the field of view at that time; S49: For driverless cars i Determine whether the geometric center of the unmanned vehicle cluster is covered within the angular range of the field of view direction in the next time step; If so, switch the field of view to the direction of the field of view in the next time step; Otherwise, rotate the field of view towards the geometric center of the autonomous vehicle cluster. angle; S4a: Update the potential field value of the cell grid according to the field of view of the autonomous vehicle cluster; S4b: For each driverless vehicle, repeat steps S43 to S49; if consecutive S One iteration, driverless car i The moving distances are all less than the threshold. M stop Then driverless cars i Iterative optimization of stopping position; S4c: For each autonomous vehicle, repeat steps S47 to S49; if consecutive S Two iterations, driverless cars i The change in the direction of the field of view is less than θ stop Then driverless cars i Stop iterative optimization of the field of view direction; S4d: Iterates the stopping position and field of view direction of all unmanned vehicles, and outputs the final position and field of view direction of the unmanned vehicle cluster as the optimal observation position and observation field of view generated by the simulation.

2. The method according to claim 1, characterized in that, In step S4a, the method for updating the potential field value of the cell grid includes: The potential field value of a cell is updated based on whether it is within the field of view of the autonomous vehicle and the distance from the cell to the autonomous vehicle: For cells within the field of view of the autonomous vehicle, the potential field value is obtained according to a Gaussian distribution, and the closer the cell is to the autonomous vehicle, the larger the potential field value of the cell; For cells outside the field of view of the autonomous vehicle cluster, the potential field value is 0; For cells that are outside the field of view of the autonomous vehicle cluster due to occlusion by obstacles, the potential field value is 0 and they are not included in the field of view. Meanwhile, the potential field value of a cell is superimposed by the influence of the autonomous vehicle cluster. That is, if the field of vision of different autonomous vehicles covers the same cell, the potential field value of that cell will be superimposed.

3. The method according to claim 1 or 2, further comprising the following step: S6: The unmanned vehicle cluster moves to the designated optimal observation position and rotates to the optimal observation field of view to complete the deployment, and carries out multi-view collaborative perception of the target area through sensor modules.

4. A multi-view perception-oriented unmanned vehicle swarm system applying the method described in any one of claims 1-3, characterized in that, include: Unmanned vehicle-mounted equipment and mission controllers, including: The unmanned vehicle-mounted equipment includes: an onboard computer, an onboard ad-hoc network module, a sensor module, a positioning module, and a power supply, all configured on each unmanned vehicle in the cluster; The onboard computer is connected to the self-organizing network module, sensor module and positioning module to control the networking communication, positioning and multi-view collaborative perception functions of the unmanned vehicle cluster; The vehicle-mounted self-organizing network module is connected to the vehicle-mounted computer and is used to send and receive data, enabling the unmanned vehicle to join the cluster system communication network; The sensor module is connected to the onboard computer to enable multi-view collaborative perception of the unmanned vehicle cluster; The positioning module is connected to the onboard computer and is used to obtain the positioning coordinates of the unmanned vehicle; The task controller includes: an ad hoc network module, a task computer, a power supply, and an interaction module; The self-organizing network module is connected to the task computer and is used to send and receive data, enabling the task controller to join the cluster system communication network. The task computer is connected to the self-organizing network module and the interaction module, providing computing power, video memory, memory, and storage space computing resources, and running the cluster simulation deployment algorithm based on the artificial potential field; The interaction module is used for human interaction and to define the target perception area.

Images