A Multi-UAV Area Coverage Flight Path Planning Method and System

Abstract

The present invention relates to a multi-UAV area coverage trajectory planning method and system, belonging to the technical field of multi-UAV cooperative area search trajectory planning. The task area is divided according to the total search capabilities of each set point to obtain the search area of each set point, and then the search area of the set point is further divided to obtain the flight area of each UAV included in the set point. Finally, the flight area is divided by using the area decomposition method to obtain a plurality of units to be merged, and the units to be merged are recombined by using the dynamic programming method to obtain several recombined areas, and trajectory planning is carried out in each recombined area to obtain the trajectory of each UAV, so as to complete the multi-UAV cooperative area coverage trajectory planning process, which has guiding significance for realizing multi-UAV cooperative work.

Description

Technical Field

[0001] The present invention relates to the technical field of multi - UAV collaborative area search trajectory planning, and particularly to a multi - UAV area coverage trajectory planning method and system based on geometric area segmentation. Background Art

[0002] With the continuous development of technology, future multi - UAVs need to adopt cluster operations to complete combat missions in order to improve the anti - risk ability. When facing the problem of multi - UAV cluster collaborative area search, it is not only necessary to reasonably allocate tasks for different UAVs according to various constraints, but also requires the UAVs to cooperate with each other and work together to maximize the overall efficiency.

[0003] In the problem of multi - UAV collaborative area search, the most common way is to reasonably divide the area, which is efficient, clear and easy to implement. However, the current UAV search trajectory planning methods mainly target single UAVs, and the research results in the field of multi - UAV collaborative search planning are still scarce. The influence of the initial positions of each UAV on collaborative search is not considered, and at the same time, there is a lack of a collaborative planning scheme for UAVs with different search capabilities, making it not highly available in actual scenarios. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi - UAV area coverage trajectory planning method and system, which can complete the multi - UAV area coverage trajectory planning process.

[0005] To achieve the above purpose, the present invention provides the following solutions:

[0006] A multi - UAV area coverage trajectory planning method, the trajectory planning method includes:

[0007] Dividing the task area according to the total search capabilities of each set of points to obtain the search area of each set of points; the set of points includes several UAVs;

[0008] For the search area of each set of points, dividing the search area of the set of points to obtain the flight area of each UAV included in the set of points;

[0009] For the flight area of each UAV, using the area decomposition method to divide the flight area to obtain multiple units to be merged; using the dynamic programming method to reorganize the units to be merged to obtain several reorganized areas; and performing trajectory planning within each reorganized area to obtain the trajectory of the UAV.

[0010] A multi - UAV area coverage trajectory planning system, the trajectory planning system includes:

[0011] A search area division module, configured to divide a task area according to the total search capabilities of each set point to obtain the search area of each said set point; each said set point includes a number of unmanned aerial vehicles.

[0012] A flight area division module, configured to divide the search area of each said set point to obtain the flight area of each unmanned aerial vehicle included in the set point for the search area of each said set point.

[0013] A flight path planning module, configured to divide the flight area of each unmanned aerial vehicle by using a region decomposition method to obtain a plurality of units to be merged; use a dynamic programming method to reorganize the units to be merged to obtain a number of reorganized regions; perform flight path planning within each said reorganized region to obtain the flight path of the unmanned aerial vehicle.

[0014] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention:

[0015] The present invention is used to provide a multi - unmanned - aerial - vehicle area - coverage flight path planning method and system. The task area is divided according to the total search capabilities of each set point to obtain the search area of each set point, then the search area of the set point is divided to obtain the flight area of each unmanned aerial vehicle included in the set point. Finally, the flight area is divided by using a region decomposition method to obtain a plurality of units to be merged, and a dynamic programming method is used to reorganize the units to be merged to obtain a number of reorganized regions, and flight path planning is performed within each reorganized region to obtain the flight path of each unmanned aerial vehicle, so that the area - coverage flight path planning process of multi - unmanned - aerial - vehicle collaboration can be completed, which has guiding significance for realizing multi - unmanned - aerial - vehicle collaborative work. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a flowchart of the flight path planning method provided in Embodiment 1 of the present invention.

[0018] Figure 2 It is a schematic diagram of the geometric division of the task area when two set points are used in Embodiment 1 of the present invention.

[0019] Figure 3Schematic diagram of geometric partitioning of the task area when using multiple collection points provided in Embodiment 1 of the present invention;

[0020] Figure 4 Schematic diagram of geometric partitioning of the task area when using three collection points with different ratios provided in Embodiment 1 of the present invention;

[0021] Figure 5 Schematic diagram of the division of the trapezoidal decomposition example provided in Embodiment 1 of the present invention;

[0022] Figure 6 Unit connection diagram of the trapezoidal decomposition example provided in Embodiment 1 of the present invention;

[0023] Figure 7 Schematic diagram of the division of the rotated trapezoidal decomposition example provided in Embodiment 1 of the present invention;

[0024] Figure 8 Schematic diagram of the minimum convex polygon of the trapezoidal decomposition example provided in Embodiment 1 of the present invention;

[0025] Figure 9 Schematic diagram of the division of the minimum convex polygon of the trapezoidal decomposition example provided in Embodiment 1 of the present invention;

[0026] Figure 10 Schematic diagram of the recombination of the minimum convex polygon of the trapezoidal decomposition example provided in Embodiment 1 of the present invention;

[0027] Figure 11 Process schematic diagram of the dynamic programming method provided in Embodiment 1 of the present invention;

[0028] Figure 12 Schematic diagram of the division of the polygon area example provided in Embodiment 1 of the present invention;

[0029] Figure 13 Unit connection diagram of the polygon area example provided in Embodiment 1 of the present invention;

[0030] Figure 14 Search network schematic diagram of the dynamic programming method provided in Embodiment 1 of the present invention;

[0031] Figure 15 Recombination schematic diagram of the polygon area example provided in Embodiment 1 of the present invention;

[0032] Figure 16 Schematic diagram of the straight-line scanning track provided in Embodiment 1 of the present invention;

[0033] Figure 17 Schematic diagram of the turning track provided in Embodiment 1 of the present invention;

[0034] Figure 18 Schematic diagram of the area to be searched during the simulation provided in Embodiment 1 of the present invention;

[0035] Figure 19 Schematic diagram of the trapezoidal decomposition of the outer area of the area to be searched during the simulation provided in Embodiment 1 of the present invention;

[0036] Figure 20 Schematic diagram of the recombination of the area to be searched during the simulation provided in Embodiment 1 of the present invention;

[0037] Figure 21 Schematic diagram of the track planning of the area to be searched during the simulation provided in Embodiment 1 of the present invention;

[0038] Figure 22 Schematic diagram of the task area division of two aircraft with a single rendezvous point during the simulation provided in Embodiment 1 of the present invention;

[0039] Figure 23 Schematic diagram of the search track of the task area of two aircraft with a single rendezvous point during the simulation provided in Embodiment 1 of the present invention;

[0040] Figure 24 Schematic diagram of the task area division of two rendezvous points during the simulation provided in Embodiment 1 of the present invention;

[0041] Figure 25 Schematic diagram of the simplification of the task area division of two rendezvous points during the simulation provided in Embodiment 1 of the present invention;

[0042] Figure 26 Schematic diagram of the search track of the task area of two rendezvous points during the simulation provided in Embodiment 1 of the present invention;

[0043] Figure 27 System block diagram of the track planning system provided in Embodiment 2 of the present invention. Detailed implementation manners

[0044] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0045] The purpose of the present invention is to provide a multi-UAV area coverage track planning method and system, which can complete the multi-UAV area coverage track planning process.

[0046] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] Embodiment 1:

[0048] Currently, for the task planning problem of area coverage search, the following planning methods mainly include: (1) By proportionally dividing the polygon area, the area allocation of multiple unmanned aerial vehicles (UAVs) is completed, and the path planning problem of a single area is solved by the parallel search method within the allocated area. (2) Using a distributed online heuristic strategy to solve the multi-UAV cooperative coverage problem, each UAV maintains a probability grid map in the form of a locally stored matrix, and designs two evaluation functions and related technical strategies, enabling the UAVs to make state transfer or area transfer decisions in an online self-organizing manner. The simulation results show that this strategy has high search efficiency, good robustness, and fault tolerance. (3) An optimal full-coverage aerial self-organizing network construction scheme is proposed. While obtaining the maximum network coverage, the minimum number of UAVs is deployed to essential locations. The simulation results show that this scheme is superior to several peer algorithms in terms of traversal time and redundant access rate. (4) A multi-UAV area coverage model based on the central Voronoi configuration is proposed, which realizes the optimal deployment of the search coverage area and search time. The cooperative search strategy based on map information update and fusion realizes the autonomous cooperative control of UAVs. (5) A multi-UAV cooperative disaster inspection search method is proposed for the rescue problem. Fully considering factors such as terrain, inspection time, search area coverage, and the number of UAVs, a multi-aircraft cooperative flight model under primary and secondary disasters is established. Combining the fastest descent method and the A* algorithm to obtain the route plan, the maximum coverage search is completed using UAVs within the specified time, and the global search path under secondary disasters is obtained using the tabu search algorithm. (6) A coverage search algorithm based on basic behavior combination and environment mapping is proposed. Using a discrete map, a simulation model of UAV random shape coverage search is established to update the environment and task changes in a timely manner. The comparison between simulation analysis and dynamic programming shows that this method has scalability and can effectively change the search strategy. However, looking at the current multi-UAV cooperative search planning, the research results are still scarce and remain at the basic theoretical level. In this embodiment, the advantages of each method are integrated and extended to cooperative planning, various basic algorithms are fused, and the best path planning is achieved through hybrid application.

[0049] This embodiment is used to provide a multi - UAV area - coverage trajectory planning method, making reasonable basic assumptions about the multi - UAV collaborative search problem. The UAVs adopt a particle - based model. The positions of the UAVs and the positions of the mission areas can be described by coordinates. The sensor coverage areas of each UAV are clear and can be different from each other, and the detection size is much smaller than the area of the mission area. The boundary of the mission area can be described by straight lines, that is, a polygon. As Figure 1 shown, the trajectory planning method includes:

[0050] S1: Divide the mission area according to the total search capabilities of each set of points to obtain the search area of each set of points; the set of points includes several UAVs;

[0051] Area - coverage search can be divided into two categories: the multi - UAV single - set - point problem and the multi - UAV multi - set - point problem. For the multi - UAV single - set - point problem, directly use the mission area as the search area of this single set of points, and use S2 to determine the flight area of each UAV. For the multi - UAV multi - set - point problem, it can be decomposed and simplified into two sub - problems: each set of points includes only one equivalent UAV and one set of points includes multiple UAVs. Use S1 to solve the sub - problem that each set of points includes only one equivalent UAV to determine the search areas of each set of points; use S2 to solve the sub - problem that one set of points includes multiple UAVs to determine the flight area of each UAV.

[0052] This embodiment uses an irregular area - segmentation method based on geometric segmentation for the segmentation of the mission area, then S1 can include:

[0053] (1) Take the sum of the search capabilities of all UAVs included in the set of points as the total search capability of this set of points to obtain the total search capabilities of each set of points;

[0054] Each set of points has multiple UAVs. For each set of points, all UAVs included in this set of points can be equivalent to one equivalent UAV, and the sum of the search capabilities of all UAVs included in this set of points represents the total search capability of this set of points. Among them, the search capability of the UAV is determined by the coverage range of the sensor carried on the UAV and is known in advance.

[0055] (2) Divide the mission area according to the ratio of the total search capabilities of each set of points to obtain the search area of each set of points; the ratio of the total search capabilities of each set of points is the same as the ratio of the areas of the search areas of each set of points.

[0056] Divide the mission area among each set of points according to the ratio of the total search capabilities of each set of points. Inside the search area responsible for each set of points, re - divide according to the method of the multi - UAV single - set - point to determine the flight area of each UAV.

[0057] Specifically, when completing tasks such as reconnaissance and rescue, the tasks need to be divided as much as possible according to the search capabilities of each UAV, so that the working hours of each UAV are the same and minimized. When the speed of the UAV is v and the time is t, the flight range s is: s = v * t. In a multi-UAV cluster, the sensors of each UAV have different detection ranges. When flying the same distance, the performance of the sensor determines the size of its detection range. Assuming that the detection radius of a certain UAV is r, then the detection width is 2r, and the area C of the task area that can be detected by it i is: C i = s * 2r. The search capability described in this embodiment is the area of the task area that the UAV can detect per unit time.

[0058] The total coverage search area C of all UAVs sum is:

[0059]

[0060] where m is the total number of UAVs, and r i is the detection radius of the i-th UAV.

[0061] It can be seen from the above analysis that when multiple UAVs cooperate for coverage search, the detection area of each UAV is proportional to its sensor search capability. Suppose there are m UAVs U1, U2, ···, U m , when allocating the task area, it is necessary to first perform normalization processing on the search capabilities as the performance indexes of the UAVs, defined as K1, K2, ···, K m , where:

[0062]

[0063] Therefore, for the task area R with an area of S R , it needs to be decomposed into m sub-task areas R1, R2, ···, R m , and their areas are expressed as S R1 , S R2 , ···, S Rm , satisfying:

[0064]

[0065] After dividing the area according to the ratio of the search capabilities of each UAV, these sub-areas are assigned to the designated UAVs one by one to perform the search tasks in their own areas, so as to achieve the task goal of multiple UAVs cooperating to complete area coverage search. The ratio of the detection areas of UAVs is the ratio of the search capabilities.

[0066] After calculating the total search capabilities of each set point in this embodiment, the task area can be divided according to the above principle to obtain the search areas corresponding to each set point.

[0067] As Figure 2 shown, the equivalent UAVs of the two set points are respectively located at the five-pointed stars in Figure 2 . The dashed line represents the line connecting the two aircraft and its perpendicular bisector. Intuitively, sub-regions 1-5 are closer to UAV1, and sub-region 6 is closer to UAV2. However, when dividing according to the search ability ratio, if the search performances are the same, then sub-regions 1-3 will be assigned to UAV1, and 4-6 will be assigned to UAV2. By changing the ability ratio of the two aircraft in turn, the dividing line is the intersection line of each sub-region in the figure. After mathematical analysis, each intersection line is a hyperbola with foci at the two aircraft, and the points on it have the same difference in distance to the two aircraft.

[0068] As Figure 3 shown, when the number of set points is greater than three, the task segmentation will show the figures in the Voronoi diagram method, and the dividing line is a hyperbola. The Voronoi diagram is also called the Thiessen polygon or Dirichlet diagram, which is composed of a set of continuous polygons formed by the perpendicular bisectors of the lines connecting two adjacent points. For the multi-set point problem, it can be solved by the incremental method. Its principle is to use the perpendicular bisectors of the three set points as the initial dividing lines, and then iterate multiple times of assignment and moving the dividing lines for modification, as Figure 4 shown. When dividing the region, determine the intersection line of the region with the most excess performance ratio and the region with the least performance ratio as the dividing line to be adjusted, and adjust it with a fixed increment until the division is completed. Figure 4 For three equivalent UAVs jointly searching a task area, the dotted line is the initial perpendicular bisector, and the final dividing method obtained through the incremental adjustment loop. The task division ratio of the three set point aircraft is (0.4, 0.2, 0.4).

[0069] After obtaining the search areas responsible for each set point, the next step is to solve the problem of multiple UAVs with a single set point. The problem of multiple UAVs with a single set point is regarded as a special case of multiple UAVs with multiple set points, that is, the set points overlap and the other conditions remain unchanged. At this time, the region division method is exactly the same as the previous one, except that the hyperbola for dividing the region degenerates into a straight line passing through the set point itself, so that the area ratio of the regions divided by the straight line is equal to the search ability ratio of each aircraft.

[0070] S2: For the search area of each said set point, divide the search area of the set point to obtain the flight areas of each UAV included in the set point;

[0071] Regarding the problem of dividing the search area of the collection point, the search area is rasterized and divided according to the principle that the grid points assigned to a certain UAV are as close as possible to the position of the UAV. Specifically, S2 may include:

[0072] (1) Rasterize the search area to obtain a rasterized area; the rasterized area includes a plurality of grid points;

[0073] (2) Randomly assign all grid points to the subsets corresponding to each UAV according to the ratio of the search capabilities of each UAV, and the ratio of the search capabilities of each UAV is the same as the ratio of the number of grid points included in the subset corresponding to each UAV;

[0074] (3) Cyclically exchange the grid points in each subset according to the cost matrix until the objective function value after the exchange is greater than or equal to the objective function value before the exchange; the objective function value is the sum of the distance sums corresponding to each subset, and the distance sum corresponding to each subset is the sum of the distances from each grid point included in the subset to the UAV corresponding to the subset.

[0075] Among them, the construction method of the cost matrix is: calculate the distance from each grid point to each UAV, construct the cost matrix, and the element in the i-th row and j-th column of the cost matrix is the distance from the i-th UAV to the j-th grid point.

[0076] Among them, cyclically exchanging the grid points in each subset according to the cost matrix may include: cyclically exchanging through pairwise exchange. For the selected two subsets, calculate the cost difference of each grid point according to the cost matrix, use the cost difference as the standard for whether the grid points in the subset are exchanged, and cyclically exchange the grid points in each subset according to the cost difference.

[0077] Next, taking the collection point including two aircraft as an example, the above principle that the grid points assigned to a certain UAV are as close as possible to the position of the UAV is described by a formula. Assume that the rasterized scatter point set of the search area is K, which contains points {k1, k2, k3,..., k n}. K1 and K2 represent the subsets assigned to two UAVs, and n1 and n2 represent the number of points.

[0078] When the coverage search performance of the two UAVs is the same, then:

[0079]

[0080] The final ideal area division effect is as follows:

[0081]

[0082] Among them, is the sum of the distances between each grid point of the subset K1 and the corresponding unmanned aerial vehicle of the subset K1, is the sum of the distances between each grid point of the subset K2 and the corresponding unmanned aerial vehicle of the subset K2.

[0083] In this embodiment, the above allocation objective is achieved by establishing a cost matrix. When two unmanned aerial vehicles are allocated, the cost matrix is a 2×n matrix, and the element c ij represents the cost from unmanned aerial vehicle i to grid point j. Therefore, the cost difference Δc j of a single grid point can be calculated by the following formula: Δc j = c 1j - c 2j . After calculating the cost differences of each grid point, each grid point is randomly assigned to the subset K1 or the subset K2 according to the search ability ratio of the two unmanned aerial vehicles, and then the grid points within the subset are cyclically exchanged until the objective function value after the exchange is greater than or equal to the objective function value before the exchange.

[0084] S3: For the flight area of each of the unmanned aerial vehicles, use the region decomposition method to divide the flight area to obtain a plurality of units to be merged; use the dynamic programming method to reorganize the units to be merged to obtain several reorganized regions; perform trajectory planning within each of the reorganized regions to obtain the trajectory of the unmanned aerial vehicle.

[0085] In this embodiment, the polygon region segmentation method can be used to divide the flight area. If the flight area is a concave polygon, it is necessary to first decompose the polygon into several convex polygons. The unmanned aerial vehicle motion planning is achieved by decomposing a flight area into a set of geometric figures. The centers of these geometric figures can be used as potential trajectory points for finding the fastest trajectory, so as to make the overall trajectory planning process run as fast and efficiently as possible and reduce the calculation time. The division of the flight area generally selects trapezoidal decomposition because the trapezoid is a more effective geometric shape, with a smaller cost when using the S-shaped scanning trajectory to cover. The long and thin trapezoids generated require significantly fewer turns of the unmanned aerial vehicle than triangles. At the same time, the trapezoidal decomposition generates fewer units than triangulation or approximate decomposition techniques, and the number of decomposed units can be kept at a relatively low level. This segmentation method can accelerate the regional reorganization and optimization to a certain extent.

[0086] Specifically, in this embodiment, trapezoidal decomposition of the polygon can be performed, showing the vertical slicing lines generated at each vertex. Each unit formed by each vertical slicing line and the boundary of the polygon is the unit obtained after decomposition; a unit connection diagram of the trapezoidal decomposition can also be obtained, where each node represents a unit and each edge represents unit adjacency. As Figure 5As shown, it is a partitioning schematic diagram of a trapezoidal decomposition example. The vertical slicing lines intersecting the polygon define the vertices of all cells. Using the trapezoidal decomposition method, a solution can be found within O(nlogn) time, where n is the number of vertices, which means that polygons with a high number of vertices will have an acceptable computation time. Another important output of the trapezoidal decomposition is the cell connection graph, which will be used in the region recombination phase to determine which cells can be combined. The cell connection graph of the trapezoidal decomposition example is as shown in Figure 6 as shown.

[0087] Since the slicing is always vertical, the final decomposition of a specific polygon depends on its initial rotation. If the polygon is rotated, decomposed, and then rotated back to its original orientation, there are infinitely many decomposition possibilities, which means that there may be different and possibly more desirable decompositions at other rotation angles. In this embodiment, by simply using the longest axis of the cell and minimizing the number of turns, the optimal flight path can be found, that is, the vertical slicing lines defined in this embodiment are all perpendicular to the long axis of the flight region. As shown in Figure 7 is a partitioning schematic diagram after rotating the trapezoidal decomposition example by 90 degrees.

[0088] Each vertex of the polygon must be rotated around the polygon by an angle ψ p , assuming that P is a polygon with n vertices, then the rotation of P is shown as follows, where P r is the rotated polygon and ψ p is the rotation angle.

[0089]

[0090] Different from ground search, aerial flight paths are usually not only restricted within the mission area (unless there are obstacles such as no-fly zones). They can fly over non-mission areas and directly transfer between multiple convex polygon cell areas. Therefore, the overall flight cost can be reduced by filling the external areas. This embodiment proposes a method of extended trapezoidal decomposition including potential external cells that may reduce the overall flight time. The method of finding and generating these external cells is to find the convex hull around the concave polygon. Finding the convex hull around the concave polygon is the method to achieve finding these external cells. The convex hull is the simplest convex polygon that contains all the vertices of the concave polygon.

[0091] As shown in Figure 8 is the convex hull of the trapezoidal decomposition example. Subtracting the original concave polygon from the convex hull generates the external optional polygon. The external optional polygon itself can obtain all optional cells through trapezoidal decomposition, as shown in Figure 9For the areas 10-13 shown, since the flight area and the external optional polygon share vertices, the decomposed cells will always be aligned with each other in the vertical slice direction of the trapezoidal decomposition. This means that the cell merging algorithm will generate longer and thinner convex polygons afterwards, which is more conducive to fast search. Figure 10 Shown is an example of recombination between cells 1-4 and the optional cell 10, which provides a larger single convex area for coverage search. Without the optional cell 10, cells 1-4 would impose a greater cost on the search.

[0092] Based on the above theory, in S3, the flight area is divided using the area decomposition method, and the multiple cells to be merged obtained may include:

[0093] (1) Determine the minimum convex polygon surrounding the flight area; the minimum convex polygon is the convex hull as described above.

[0094] (2) For each vertex of the minimum convex polygon, determine a dividing line passing through the vertex and along the direction perpendicular to the long side of the minimum convex polygon; all the dividing lines divide the minimum convex polygon into multiple cells to be merged. The long side direction is the major axis direction of the minimum convex polygon.

[0095] After obtaining multiple cells to be merged, a cell connection graph can also be obtained by taking each cell to be merged as a node and adding connection edges between two adjacent cells to be merged.

[0096] Dynamic programming is a technique for decomposing complex problems into simple sub-problems. By optimally combining these solutions with smaller problems, a complete solution can be constructed. Like the traversal method, dynamic programming examines each possible solution to ensure the optimality of the solution. The advantage of this idea is that the solutions to simple problems can be stored or memoized, and the solutions to the same sub-problems can be reused in iterative calculations. If there are obvious overlaps in sub-problems, a large amount of computing time can be saved, thereby achieving the purpose of reducing the computing cost. In this embodiment, the area cell merging can be decomposed into smaller sub-problems, and then the dynamic programming recursive algorithm is used to compare the optimal solution and the solutions of each sub-problem with the solution of the entire area to obtain the optimal solution. The optimal objective it seeks is expressed as follows: J(G) = min{C(G), min[J(G1) + J(G2)]};

[0097] Among them, J(G) represents the optimal cost of the search area G, which is the minimum of the simple cost C(G) of directly searching the entire area and the sum of the optimal costs of searching any two sub-areas G1 and G2, where G1 and G2 cover the entire search area. Then, the algorithm can be recursively applied by solving the two sub-problems J(G1) and J(G2) using the same equation. Since there will be redundant flights in concave areas, a penalty cost is added for selection. If the polygon is convex, the cost C(G) of the search area G is defined as the flight time here, and a very high penalty cost is given when the area is concave:

[0098]

[0099] Among them, the flight time t is the ratio of the search cost L to the flight speed of the UAV. The calculation method of the search cost L is as follows: Assume that the UAV speed is constant at v, the width of the area that the UAV can cover during a straight-line search is ′, and the minimum turning radius of the UAV is r, and the constraint d≥2r needs to be satisfied. Since the UAV speed is constant, the search cost can be represented by the total length of the coverage flight path:

[0100]

[0101] Among them, L is the length of the overall flight path, i is the number of the scan line, is the length of the scan flight path of the i-th scan line, is the length of the turning flight path of the i-th scan. According to the definition of the Dubins path, the length of the turning flight path can be obtained: l t =d+(π - 2)r.

[0102] In this embodiment, a bottom-up dynamic programming method is selected. The ways of region merging can be roughly divided into two types. One is the "bottom-up" planning method, that is, starting from the smallest unit and continuously splicing, and the other is the splitting method from the whole to local units. The latter starts from the cost C(G) of the global G, first searches the entire area, and then gradually decomposes the area into smaller sub-areas G1 and G2. The former method starts from a single unit area and gradually constructs a search graph by recombining adjacent unit areas or keeping them unmerged, so as to finally cover the complete search area. This method is the same as the above formula in mathematics, but the calculation order is slightly different in the algorithm. According to the idea of recursive merging, to ensure that the results of sub-problems can be reused, the former method is adopted here, ensuring a lower memory usage.

[0103] As Figure 11 shown is the display of the merging idea of this method. Assume that there are currently five unit areas, units 1 - 3 are internal units, and 4 - 5 are external supplementary units. The corresponding area examples are as Figure 12As shown. In the algorithm, operations can start from any initial point. Taking units 1&2 as an example, this unit can be merged with any adjacent unit (such as Figure 11 the upper branch), or it can be considered that the merging of the current gray area has been completed, and then a new starting unit (white) is taken from any other adjacent unit in the search area for merging (such as Figure 11 the lower branch). The algorithm will work recursively starting from each new starting point until every internal unit in the search area is covered. The cost of the current merging is calculated and stored at each step, and these stored costs are reused in subsequent calculation steps. For example Figure 11 subsequent steps of the lower branch will combine G{4,5} multiple times, and the cost C(G{4,5}) needs to be calculated for each merging. If the calculation result when this situation is first encountered is stored, it can be simply called from memory at any subsequent time. In this recursive merging method, all possible unit areas are traversed and merged until all possible solutions that fully contain each required (internal) unit are found. The flow of the recursive function algorithm described above is summarized as follows:

[0104] a) Arbitrarily select the starting unit area, list all its neighbor areas, and merge them one by one;

[0105] b) Calculate the cost after merging the current unit area and store it;

[0106] c) Arbitrarily select a new starting area from the internal area (excluding the merged units) and list its neighbor units;

[0107] d) Calculate the cost after merging the current unit area and store it;

[0108] e) When the internal unit area is fully covered, return all the stored cost values, select the minimum cost, and its merging method is the optimal one.

[0109] For the problem of detecting the convexity condition of polygons during cost calculation, in this embodiment, the gift wrapping algorithm is selected to evaluate their convexity clockwise according to the order of points to complete the judgment on whether the merged area is convex or concave.

[0110] Based on the above theory, in S3, the dynamic programming method is used to reorganize the units to be merged, and several reorganized areas that can be obtained include:

[0111] (1) According to the unit connection graph, use the bottom-up dynamic programming method to reorganize the units to be merged to obtain multiple reorganization schemes; each reorganization scheme includes several alternative areas;

[0112] (2) Calculate the cost of each recombination scheme, and select the recombination scheme with the minimum cost as the optimal recombination scheme; several alternative regions included in the optimal recombination scheme are several recombined regions.

[0113] Among them, calculating the cost of each recombination scheme may include:

[0114] For each alternative region included in each recombination scheme, use the gift wrapping algorithm to determine whether the alternative region is a convex polygon; if so, the cost of the alternative region is the flight time; otherwise, the cost of the alternative region is ∞; calculate the sum of the costs of all alternative regions to obtain the cost of the recombination scheme. The calculation method of the flight time is as follows: conduct a flight path planning for the alternative region, calculate the search cost L, and then divide the search cost L by the flight speed of the UAV to obtain the flight time.

[0115] The following uses a recombination example to elaborate in detail the process of unit recombination using the unit region merging method based on dynamic programming:

[0116] Figure 12 shows a decomposed flight region, which consists of three internal units 1-3 and two external units 4 and 5. Figure 13 is a unit connection graph used to represent the regional adjacency relationship, and this graph shows the relationship list of merging these unit regions.

[0117] The first step of dynamic programming is to establish Figure 14 the search network shown in the figure. By using a recursive algorithm to branch from the left node 1 and move towards the exit node on the right, this graph is constructed. Once reaching a node where no forward branching is possible (such as the exit node), the program backtracks one step and tests alternative solutions, and the recursion continues until all possible branches are exhausted. Any node that contains all internal units can directly branch to the exit node, representing a possible final solution. It should be noted that 1|2 means that cells 1 and 2 are separate polygons, and 1&2 means that they have been recombined. Figure 14Details show certain optimization effects applied to the recombination process. First, it can be seen that in the first step, the program checks the recombination with two adjacent cells 2 and 4 and the new polygon expanding from cell area 2 (since cell area 4 is an optional external object cell). Second, it can be seen at the top of the figure that the node "1|2|3" has a branch with a direct exit because this scheme already contains all the necessary internal cells and is thus a solution in itself. In addition, although the node "1&2&3" contains all the necessary cell areas, this node does not branch to the exit node because the recombination of G{1,2,3} will result in concave cells, the cost of which is defined as infinite, so the planner will choose to continue with the branch of further merging with other cells. Finally, "1&2&3&5" has no branch because in this case, the only viable additional cell is cell 4, but the sub-region "1&2&3&4&5" has already obtained the result from the branch of "1&2&3&4", so there is no need to calculate again. Similarly, there are no branches from "1&2&4&5" and "1&2&5".

[0118] Figure 15 Shows the two final sub-regions G{1} and G{2,3,5} after decomposition and recombination. The lowest-cost segmentation method contains an optional region #5. The combined region contains the original external cell 5, while region 4 is not in it. The introduction of region 5 merges with cell areas 2 and 3 to generate a convex sub-region, and in this case, the search cost of the drone is minimized. By recording the number of calculation groups, the efficiency of the recursive merging calculation can be compared with the direct traversal method, and it is found that the number of calculations is greatly reduced. The recursive merging calculation method can effectively reduce the amount of calculation and keep the planning speed at a relatively high level.

[0119] The coverage search trajectory consists of two different flight states: the straight-line scanning trajectory during search and the turning trajectory for transitioning to the next scan. The start and end of each scan can be represented by trajectory points, which define the positions where the scan line intersects the polygon of the mission area, as Figure 16 shown. The line connecting the two trajectory points represents a search scan line, with coordinates where i is the scan index:

[0120]

[0121]

[0122] Among the intersection points of the drone with the scanning trajectory on the boundary of the mission area, the starting trajectory point is defined as [x f , y f , and its heading is the scanning angle ψ sIn the direction, the track point where the scan ends is defined as [x0, y0], and the course is ψ s Take the intersection points of the outermost two scan lines and the regional boundary as corner points. The UAV can choose to start the search from one of the four corner points and follow the correct track point order to achieve the coverage patrol of the mission area.

[0123] In Figure 16 a S-shaped cruise track is generated with a scanning angle of 45 degrees, which is obviously not the best because some straight sections are very short and there are many turns. A simple but effective method to select the scanning angle is to find the minimum bounding box of the enclosed polygon, align the scan with the long axis of the polygon, then take the angle of the long axis of the known bounding box as the scanning angle, and define the intersection points of the scan line and the mission area polygon as straight track points, representing the traversal order of all scans in the convex polygon. Generally, the starting point of the coverage search is from one of the polygon corner points, so that the generated cruise path has no repetition and is conducive to reducing the program operation complexity, thus ensuring the planning rate.

[0124] After obtaining the scan line, to make the UAV transition from one scan track to the next, a turning operation is still required, which is specifically manifested as alternating between left turns and right turns and involves a 180-degree course adjustment. To complete the entire flight time model, this operation needs to be specifically defined, and this part can be achieved by using the Dubins path. The Dubins path is the shortest curve connecting two points in the two-dimensional plane, and its curvature is restricted (in this case, it is the maximum turning rate of the UAV), as Figure 17 shown. The Dubins path provides a continuous flight track for the transition between adjacent orbits. It can be seen from the figure that these tracks are composed of two turning circles (whose radii satisfy the minimum turning radius constraint) relative to the track start and end points and part of the common tangent. All parts of this track are simple geometric shapes, which makes it simple and easy to calculate the length and flight time.

[0125] Based on the above theory, in S3, the track planning is carried out within each reorganized area, and the specific track of each UAV obtained includes:

[0126] (1) For each reorganized area, take the angle between the reorganized area and the horizontal line as the scanning angle, and plan the straight scan track according to the width of the area covered during the straight search of the UAV and the scanning angle to obtain the straight scan track;

[0127] Specifically, the width of the area covered during the straight-line search of the UAV is determined by the detection width of the UAV sensor, and the maximum value of the width of the area covered during the straight-line search of the UAV is equal to the detection width. The value of the scanning angle is determined as follows: determine the minimum area bounding box of the recombined area, and use the angle between the minimum area bounding box and the horizontal line as the value of the scanning angle. Starting from any vertex of the recombined area, with the scanning angle as the extension direction, plan the first straight-line scanning track. Then, use the width of the area covered by the UAV's straight-line search as the distance between the next straight-line scanning track and the current straight-line scanning track, and use the scanning angle as the extension direction to plan the second straight-line scanning track. And so on, until the next straight-line scanning track has no intersection with the recombined area.

[0128] (2) Use the Dubins path to plan the turning track between two adjacent straight-line scanning tracks to obtain the turning track; all the straight-line scanning tracks and all the turning tracks together form the track of each UAV.

[0129] Next, the trajectory planning method introduced in this embodiment is simulated. Given the search area of a single UAV as Figure 18 shown, rotate the search area by a predetermined angle, and this predetermined angle is the angle between the long axis of the search area and the horizontal direction. Since this area is non-convex, first find its enclosing convex polygon, perform trapezoidal segmentation on the rotated search area, and the segmentation result is as Figure 19 shown. Use the dynamic programming algorithm to calculate the optimal polygon decomposition and merging structure from bottom to top and then rotate back to the original angle. The recombination result is shown in Figure 20. Plan and connect the coverage search tracks according to the merged area respectively to obtain the final completed search track as follows, as Figure 21 shown.

[0130] For the multi-UAV problem, in real life, generally setting two rally points is sufficient. Therefore, this embodiment only simulates the cases of two UAVs with a single rally point and two UAVs with two rally points:

[0131] (1) The case of two UAVs with a single rally point:

[0132] Two aircraft with the same search ability are located at the point (0, 0) ( Figure 22 (a)) and the point (-134, 351) (as Figure 22 (b)) respectively. First, perform area segmentation according to the search ability of the UAVs. Since the two UAVs start from the same point to complete the search, the segmentation line is a straight line. After completing the area segmentation, then use the algorithm for search track planning described above within their respective areas to complete the final planned track for area search. The track is as Figure 23 shown.

[0133] (2) The case of two UAVs with two rally points:

[0134] Two aircraft with the same search ability are located at points (-100, 0) and (100, 0) respectively. First, the search area is divided by hyperbolas according to the search ability ratio, as Figure 24 shown. To simplify the curve description, the straight line between the intersection points of the hyperbola and the region boundary is used to replace the hyperbola, and the final region division result is as Figure 25 . The search track planning algorithm in the single aircraft mission planning technology is used respectively to obtain the final track planning result as Figure 26 (If there are multiple aircraft at each assembly point, then re-segment and plan according to the division method of multiple aircraft at a single assembly point). In the collaborative coverage area search stage, multiple UAVs scan in a zigzag pattern. By preferentially walking long straight line segments, the number of turns is reduced, the search cost is effectively reduced, the utilization efficiency of the UAVs is increased, and at the same time, the requirement that the distance from the task points assigned to a certain aircraft to that aircraft is as short as possible is met in the area allocation, and the effect meets the expectation.

[0135] Embodiment 2:

[0136] This embodiment is used to provide a multi-UAV area coverage track planning system, as Figure 27 shown. The track planning system includes:

[0137] A search area division module M1, which is used to divide the task area according to the total search ability of each assembly point to obtain the search area of each said assembly point; the assembly point includes several UAVs;

[0138] A flight area division module M2, which is used to divide the search area of each said assembly point to obtain the flight area of each UAV included in the assembly point;

[0139] A track planning module M3, which is used to divide the flight area of each UAV by using the area decomposition method to obtain multiple units to be merged; and use the dynamic programming method to reorganize the units to be merged to obtain several reorganized areas; perform track planning in each said reorganized area to obtain the track of the UAV.

[0140] What each embodiment in this specification focuses on explaining is the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method part.

[0141] In this article, specific examples are used to illustrate the principles and implementation modes of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation modes and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A multi-UAV area coverage trajectory planning method, characterized in that The described trajectory planning method includes: Dividing the task area according to the total search capabilities of each set point to obtain the search area of each set point; the set point includes several unmanned aerial vehicles (UAVs). For the search area of each set point, dividing the search area of the set point to obtain the flight area of each UAV included in the set point. For the flight area of each UAV, using the region decomposition method to divide the flight area to obtain multiple units to be merged; using the dynamic programming method to reorganize the units to be merged to obtain several reorganized regions; performing trajectory planning within each reorganized region to obtain the trajectory of the UAV.

2. The track planning method according to claim 1, wherein The specific steps of dividing the task area according to the total search capabilities of each set point to obtain the search area of each set point include: Taking the sum of the search capabilities of all UAVs included in the set point as the total search capability of the set point to obtain the total search capabilities of each set point. Dividing the task area according to the ratio of the total search capabilities of each set point to obtain the search area of each set point; the ratio of the total search capabilities of each set point is the same as the ratio of the areas of the search areas of each set point.

3. The trajectory planning method according to claim 1, wherein The specific steps of dividing the search area of the set point to obtain the flight area of each UAV included in the set point include: Performing rasterization processing on the search area to obtain a rasterized area; the rasterized area includes multiple raster points. Allocating all the raster points to the corresponding subsets of each UAV according to the ratio of the search capabilities of each UAV; the ratio of the search capabilities of each UAV is the same as the ratio of the number of raster points included in the corresponding subset of each UAV. Performing cyclic exchange on the raster points within each subset according to the cost matrix until the objective function value after the exchange is greater than or equal to the objective function value before the exchange; the objective function value is the sum of the distance sums corresponding to each subset; the distance sum corresponding to a subset is the sum of the distances from each raster point included in the subset to the UAV corresponding to the subset.

4. The track planning method according to claim 3, wherein, Before performing cyclic exchange on the raster points within each subset according to the cost matrix, the trajectory planning method further includes: calculating the distance from each raster point to each UAV and constructing a cost matrix; the element in the i-th row and j-th column of the cost matrix is the distance from the i-th UAV to the j-th raster point.

5. The track planning method according to claim 1, characterized in that, The specific steps of using the region decomposition method to divide the flight area to obtain multiple units to be merged include: Determining the minimum convex polygon surrounding the flight area. For each vertex of the minimum convex polygon, determining a dividing line passing through the vertex and along the direction perpendicular to the long side of the minimum convex polygon; all the dividing lines divide the minimum convex polygon into multiple units to be merged.

6. The track planning method according to claim 5, wherein Before reorganizing the to-be-merged cells using the dynamic programming method, the trajectory planning method further includes: taking each of the to-be-merged cells as a node, and adding connection edges between two adjacent to-be-merged cells to obtain a cell connection graph.

7. The track planning method according to claim 6, wherein The reorganizing the to-be-merged cells using the dynamic programming method to obtain several reorganized regions specifically includes: According to the cell connection graph, using the dynamic programming method to reorganize the to-be-merged cells to obtain multiple reorganization schemes; each reorganization scheme includes several alternative regions; Calculating the cost of each reorganization scheme, and selecting the reorganization scheme with the minimum cost as the optimal reorganization scheme; the several alternative regions included in the optimal reorganization scheme are the several reorganized regions.

8. The track planning method according to claim 7, characterized in that, The calculating the cost of each reorganization scheme specifically includes: For each alternative region included in each reorganization scheme, using the gift wrapping algorithm to determine whether the alternative region is a convex polygon; if so, the cost of the alternative region is the flight time; otherwise, the cost of the alternative region is ∞; Calculating the sum of the costs of all the alternative regions to obtain the cost of the reorganization scheme.

9. The track planning method according to claim 1, wherein, The performing trajectory planning within each reorganized region to obtain the trajectory of the UAV specifically includes: For each reorganized region, planning a straight-line scanning trajectory according to the width and scanning angle of the area covered during the straight-line search of the UAV; the scanning angle is the angle between the reorganized region and the horizontal line; Using the Dubins path to plan the turning trajectory between two adjacent straight-line scanning trajectories to obtain the turning trajectory; all the straight-line scanning trajectories and all the turning trajectories together form the trajectory of each UAV.

10. A multi-UAV area coverage trajectory planning system, characterized in that, The trajectory planning system includes: A search area division module, configured to divide the task area according to the total search capabilities of each set point to obtain the search area of each set point; the set point includes several UAVs; A flight area division module, configured to divide the search area of each set point to obtain the flight area of each UAV included in the set point for the search area of each set point; A trajectory planning module, configured to divide the flight area using the region decomposition method to obtain multiple to-be-merged cells for the flight area of each UAV; reorganizing the to-be-merged cells using the dynamic programming method to obtain several reorganized regions; and performing trajectory planning within each reorganized region to obtain the trajectory of the UAV.

Images