A multi-UAV collaborative search method based on prior information

Abstract

The present invention relates to a multi-UAV collaborative search method based on prior information, comprising mapping the latitude and longitude coordinates of a geographical point in a search area to a Cartesian coordinate space to obtain a target search area; discretizing the target search area into a pseudo-discrete grid, and discretizing the edges of each pseudo-discrete grid into search points with uniform intervals; according to the prior probability information of the target in the search area, using N search starting points as cluster centers to generate N task areas using a Kmeans clustering algorithm: auctioning pseudo-discrete grids that conflict between task areas; taking a step size of W as the number of search points; and k The mission area of ​​each UAV is rasterized to determine the final mission area of ​​each UAV, and a probability map of the final mission area of ​​the UAV is established; a search function of the UAV is established to calculate the optimal search path of the UAV in the final mission area; the optimal search path of the UAV is mapped back to the actual geographic coordinates to guide the UAV to search for the target in a real scenario.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles, and in particular to a multi-unmanned aerial vehicle collaborative search method based on prior information. Background Art

[0002] In recent years, drones have been increasingly used in military and civilian fields, including battlefield surveillance, aerial refueling, environmental monitoring, map construction, and intelligent communications. One typical application is target search, such as border patrol, post-disaster rescue, and perimeter monitoring. These tasks are usually urgent and have a wide search range, requiring multiple drones to search together. Target search path planning, as part of the search task, is crucial to improving search efficiency. In such tasks, the exact location of the target is unknown. In order to ensure that nothing is missed, a covering path planning method is generally adopted.

[0003] Covering search assumes that the probability of target occurrence is equal, and wastes time in areas with lower probability of target occurrence. Typical search theory usually divides the search area into a set of sub-areas, each of which is assigned a probability of target existence. The purpose of search planning is to properly allocate search time and other resources so as to prioritize the search of sub-areas with higher probability. More and more studies have begun to focus on target probability information, but no reasonable solutions have been proposed for the problems of unreasonable task allocation and low search efficiency that occur during multi-UAV collaborative search. Summary of the invention

[0004] In order to solve the problems existing in the background technology, the present invention provides a multi-UAV collaborative search method based on prior information, comprising:

[0005] S1: Obtain the longitude and latitude coordinates of the geographical points in the search area, and map the longitude and latitude coordinates of the geographical points in the search area to the Cartesian coordinate space through the Haversine formula to obtain the target search area;

[0006] S2: discretize the target search area into pseudo-discrete grids according to the altitude of the UAV from the ground and the downward field of view of the UAV's onboard camera, and discretize the edges of each pseudo-discrete grid into search points with uniform intervals;

[0007] S3: According to the prior probability information of the target in the search area, determine the N search points where the target is most likely to appear as the search starting points, and use the N search starting points as cluster centers to generate N task areas using the Kmeans clustering algorithm:

[0008] S4: Pseudo-discrete grid of conflicts between auction task areas; after the auction is completed, the step size is W kThe task area of ​​each UAV is rasterized to obtain the coordinates of each grid, and the final task area of ​​each UAV is determined, wherein the conflicting pseudo-discrete grids include: pseudo-discrete grids in which the search points in the pseudo-discrete grids belong to more than one task area;

[0009] S5: Establish a probability map of the UAV's final mission area based on the prior probability information of the target in the search area;

[0010] S6: establishing a search function for the UAV according to the probability map of the final mission area of ​​the search UAV and the flight speed of the UAV, and calculating the optimal search path of the UAV in the final mission area through the search function;

[0011] S7: Map the best search path of the UAV back to the actual geographic coordinates to guide the UAV to search for the target in the real scene.

[0012] The present invention has at least the following beneficial effects

[0013] The present invention fully considers the problems of unreasonable task allocation, low uniform coverage search efficiency and long time consumption when multiple drones perform search tasks, and adopts a fusion search path planning method to achieve fast search. First, according to the location information reported before the target disappears, several locations with a high probability of target appearance are inferred; secondly, the clustering method is used to initially divide the task area, auction the conflicting units on the boundary of the division, and reallocate the task area; finally, each drone comprehensively considers the probability of target appearance, and performs path planning in the sub-area through a heuristic greedy search algorithm to improve the efficiency of multi-drone collaborative search. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a flow chart of the method of the present invention;

[0015] Figure 2 It is a schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0016] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0017] See also Figure 1 and Figure 2The present invention provides a multi-UAV collaborative search method based on prior information, including: first, performing area selection, determining the boundary points of the area to be searched, and connecting the boundary points to generate an irregular polygonal area to be searched. Secondly, discretizing the search area and dividing it into pseudo-discrete grids. Thirdly, obtaining the initial task partition of each UAV through Kmeans clustering. Then, auctioning the conflicting grids and reallocating the task areas. Then, rasterizing the task areas of each UAV, and establishing a probability map based on the prior information of the search target. Designing two flight modes, at the cost of search time and probability information of the search target, creating a list of optimal search path points for each UAV to achieve fast search. Finally, restoring the path points to actual geographic coordinates, guiding the UAV to fly in a real scene.

[0018] Preferably, a multi-UAV collaborative search method based on prior information comprises:

[0019] S1: Obtain the longitude and latitude coordinates of the search area, and map the longitude and latitude coordinates of the search area to the Cartesian coordinate space to obtain the target search area;

[0020] A geographical point in the search area is randomly selected as the origin, and the longitude and latitude coordinates of the geographical point in the search area are mapped to the Cartesian coordinate space according to the longitude and latitude coordinates of the origin;

[0021]

[0022] Among them, φ1 represents the latitude of the origin, λ1 represents the longitude of the origin, R is the radius of the earth, d represents the distance from the other geographical points to the origin, φ2 represents the latitude of the other geographical points, λ2 represents the longitude of the other geographical points, x, y represent the coordinates of the other geographical points in the Cartesian coordinate system, (Δφ=φ1-φ2, Δλ=λ1-λ2).

[0023] S2: discretize the target search area into pseudo-discrete grids according to the altitude of the UAV from the ground and the downward field of view of the UAV's onboard camera, and discretize the edges of each pseudo-discrete grid into search points with uniform intervals;

[0024]

[0025] Where h is the height of the drone from the ground when flying, F is the downward field of view of the drone's onboard camera, and W is k Represents the edge length of the pseudo-discrete grid.

[0026] S3: According to the prior probability information of the target in the search area, determine the N search points where the target is most likely to appear as the target search points, and use the N target search points as cluster centers to generate N task areas using the Kmeans clustering algorithm; the prior probability information of the target is the location where the target has appeared historically;

[0027] The Kmeans clustering algorithm is used to generate N task areas. Similar to the Voronoi partitioning method, the initial task areas are obtained.

[0028] S4: Pseudo-discrete grid of conflicts between auction task areas; after the auction is completed, the step size is W k The task area of ​​each UAV is rasterized to obtain the coordinates of each grid, and the final task area of ​​each UAV is determined, wherein the conflicting pseudo-discrete grids include: pseudo-discrete grids in which the search points in the pseudo-discrete grids belong to more than one task area;

[0029] S41: Add a state deviation factor B to the drone, and scale the distance from the drone to the search starting point in its mission area to obtain the distance deviation factor of the drone:

[0030] d B (r) = d0(r) × B

[0031] Where d0(r) is the distance from the drone to the search starting point in its mission area, B represents the state deviation factor, and d B (r) represents the distance of drone r with the distance bias factor added. B is the heuristic bias factor, which determines the degree of influence of d0(r) on workload redistribution, that is, increasing B reduces the search tasks assigned to drones with larger initial flight distances. In the present invention, B is 0.5.

[0032] S42: Calculate the distance d from the drone to the search starting point in its mission area based on the drone's distance deviation factor B (r), and then compare the current number of tasks and d for each task area to which the conflict grid belongs B (r) Sum, divide this conflicting grid into the task area with the smallest sum, and complete the auction. For example, this conflicting grid belongs to the second and third task areas, and calculate the distance d from the drone to the search starting point in the second and third task areas respectively. B (r) and the current number of tasks in the second and third task areas, and the d of the two areas B (r) adding and comparing with the current task number, selecting the area with the smallest sum as the final task area of ​​the conflict unit; the task number is the number of regional grids of the task currently performed by each UAV excluding the conflict grid;

[0033] S43: After the auction is completed, the step length is W k The mission area of ​​each UAV is rasterized, and the row and column numbers determine the coordinates of each grid, representing the minimum search unit, and determining the area where each UAV will eventually perform the search mission.

[0034] S431: Determine the minimum coordinate point (x min ,ymin ) and the maximum coordinate point (x max ,y max );

[0035] S432: Convert the irregular task area into a w×h matrix grid, where w=x max —x min , h=y max —y min , the side length of each grid is W k .

[0036] S5: Establish a probability map of the UAV's final mission area based on the prior probability information of the target in the search area;

[0037] S51: The grid where the search start point is located is regarded as the peak grid and each peak grid is given a score PS G(i，j) ;

[0038] S52: In the final mission area of ​​the drone, the scores are gradually reduced from the peak grid to the outer grids until each grid in the final mission area of ​​the drone obtains a score, wherein the scores of grids that do not belong to the final mission area of ​​the drone are set to 0. For example, the peak grid is given a score of 5, and the scores of the outer grids of the peak grid are gradually reduced to 70% to 90% until all grids have score values;

[0039] S53: Calculate the probability P(i, j) of finding the target in each grid of the final mission area of ​​the drone according to the score obtained in each grid of the final mission area of ​​the drone, and create a probability map of the final mission area;

[0040] P(i, j) = PS G(i，j) / S

[0041] Among them, P(i, j) represents the probability of finding the target in the jth grid of the i-th column, PS G(i，j) represents the score of the peak grid, and S represents the sum of the scores of all grids.

[0042] S6: establishing a search function for the UAV according to the probability map of the UAV's final mission area and the flight speed of the UAV, and calculating the optimal search path of the UAV in the final mission area through the search function;

[0043] S61: Calculate the time it takes the drone to move between two grids based on the drone's flight speed:

[0044]

[0045]

[0046] Among them, (x i ,y i) is the coordinate of the grid where the drone is currently located, (x i-1 ,y i-1 ) is the coordinate of the next grid that the drone will reach, is the distance between two coordinate points, V m is the moving speed of the drone, The time it takes for the drone to move from one grid to the next.

[0047] S62: Establish a search function to calculate the score of each grid according to the time when the UAV moves between two grids and the probability map of the UAV's final mission area;

[0048]

[0049] Among them, P(i, j) represents the probability of finding the target in each grid, Represents the distance between two points, V m is the moving speed of the drone, T s The time it takes for the drone to search a grid.

[0050] S63: Taking the search starting point of the UAV as the initial point, calculating the scores of all other unsearched grids, selecting the grid with the largest score as the next grid to be searched, and stopping the search until the target is found, thereby generating a search path for the UAV;

[0051]

[0052] S7: Mapping the best search path of the UAV back to the actual geographic coordinates to guide the UAV to search for the target in the real scene;

[0053] Restore search point coordinates to geographic coordinates

[0054] P g =P c C -1 G T

[0055] Among them, P g is a point on the geographic plane, P c is a Cartesian point, C is a Cartesian transformation matrix, and G is a boundary point longitude and latitude matrix.

[0056] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-UAV collaborative search method based on prior information, characterized in that: include: S1: Obtain the longitude and latitude coordinates of the geographical points in the search area, and map the longitude and latitude coordinates of the geographical points in the search area to the Cartesian coordinate space through the Haversine formula to obtain the target search area; S2: discretize the target search area into pseudo-discrete grids according to the altitude of the UAV from the ground and the downward field of view of the UAV's onboard camera, and discretize the edges of each pseudo-discrete grid into search points with uniform intervals; S3: According to the prior probability information of the target in the search area, determine the N search points where the target is most likely to appear as the search starting points, and use the N search starting points as cluster centers to generate N task areas using the Kmeans clustering algorithm; S4: Pseudo-discrete grid of conflicts between auction task regions; After the auction is completed, the step length is W k The task area of ​​each UAV is rasterized to obtain the coordinates of each grid, and the final task area of ​​each UAV is determined, wherein the conflicting pseudo-discrete grids include: pseudo-discrete grids in which the search points in the pseudo-discrete grids belong to more than one task area; S5: Establish a probability map of the UAV's final mission area based on the prior probability information of the target in the search area; S6: establishing a search function for the UAV according to the probability map of the UAV's final mission area and the flight speed of the UAV, and calculating the optimal search path of the UAV in the final mission area through the search function; S7: Map the best search path of the UAV back to the actual geographic coordinates to guide the UAV to search for the target in the real scene.

2. The multi-UAV collaborative search method based on prior information according to claim 1 is characterized in that: Mapping the longitude and latitude coordinates of the geographical points in the search area to the Cartesian coordinate space by the Haversine formula includes: A geographical point in the search area is randomly selected as the origin, and the longitude and latitude coordinates of the geographical point in the search area are mapped to the Cartesian coordinate space according to the longitude and latitude coordinates of the origin; Among them, φ1 represents the latitude of the origin, λ1 represents the longitude of the origin, R is the radius of the earth, d represents the distance from the other geographical points to the origin, φ2 represents the latitude of the other geographical points, λ2 represents the longitude of the other geographical points, x, y represent the coordinates of the other geographical points in the Cartesian coordinate system, Δφ=φ1-φ2, Δλ=λ1-λ2.

3. The multi-UAV collaborative search method based on prior information according to claim 1 is characterized in that: The pseudo-discrete grid comprises: Where h is the height of the drone from the ground when flying, F is the downward field of view of the drone's onboard camera, and W is k Indicates the width of the pseudo-discrete grid.

4. The multi-UAV collaborative search method based on prior information according to claim 1 is characterized in that: The step S4 specifically includes: S41: adding a state deviation factor B to the UAV, scaling the distance from the UAV to the search starting point in its mission area to obtain the distance deviation factor of the UAV; S42: Calculate the sum of the current number of tasks in each task area to which the conflicting grid belongs and the distance from the UAV to the search starting point in its task area according to the distance deviation factor of the UAV, divide the conflicting grid into the task area with the smallest sum, and complete the auction; S43: After the auction is completed, the step length is W k The mission area of ​​each UAV is rasterized, and the row and column numbers determine the coordinates of each grid, representing the minimum search unit, and determining the area where each UAV will eventually perform the search mission.

5. The multi-UAV collaborative search method based on prior information according to claim 1 is characterized in that: The method of establishing a probability map of the final mission area of ​​the drone based on the prior probability information of the target in the search area includes: S51: The grid where the search start point is located is regarded as the peak grid and each peak grid is given a score PS G(i,j) ; S52: reducing the score from the peak grid to the outer grid layer by layer in the final mission area of ​​the drone until each grid in the final mission area of ​​the drone obtains a score, wherein the score of the grid that does not belong to the final mission area of ​​the drone is set to 0; S53: Calculate the probability P(i, j) of finding the target in each grid of the final mission area of ​​the drone according to the score obtained in each grid of the final mission area of ​​the drone, and create a probability map of the final mission area.

6. The multi-UAV collaborative search method based on prior information according to claim 5 is characterized in that: The probability of finding the target in each grid in the final mission area of ​​the UAV includes: P(i,j)=PS G(i,j) / S Among them, PS G(i,j) represents the score of the peak grid, S is the sum of the scores of the grids in each area, and P(i,j) represents the probability of finding the target in each grid.

7. The multi-UAV collaborative search method based on prior information according to claim 1 is characterized in that: The calculation steps of the optimal search path of the UAV in the final mission area include: S61: Calculate the time it takes the drone to move between two grids based on the drone's flight speed: S62: Establish a search function to calculate the score of each grid according to the time when the UAV moves between two grids and the probability map of the UAV's final mission area; S63: Taking the search starting point of the UAV as the initial point, calculating the scores of all other unsearched grids, selecting the grid with the largest score as the next grid to be searched, and stopping the search until the target is found, thereby generating the search path of the UAV.

8. The multi-UAV collaborative search method based on prior information according to claim 7 is characterized in that: The time for the drone to move between two grids includes: Among them, (x i ,y i ) is the coordinate of the grid where the drone is currently located, (x i-1 ,y i-1 ) is the coordinate of the next grid that the drone reaches, is the distance between two coordinate points, V m is the moving speed of the drone, The time it takes for the drone to move from one grid to the next.

9. The multi-UAV collaborative search method based on prior information according to claim 7 is characterized in that: The score of each grid includes: Among them, P(i,j) represents the probability of finding the target in each grid. Represents the distance between two points, V m is the moving speed of the drone, T s The time it takes for the drone to search a grid.

Images