Collaborative mission planning method for vehicle-mounted multi-UAV for large-scale regional reconnaissance

Abstract

The present invention discloses a vehicle-mounted multi-UAV collaborative task planning method for large-scale regional reconnaissance, including: dividing the area to be investigated into a number of grids that can be investigated by a single UAV, and dividing the area to be investigated into a number of sub-areas to obtain a sub-area division scheme; planning the driving path of the vehicle, obtaining the order in which the vehicle visits the vehicle-UAV junction points in all sub-areas, that is, the vehicle path planning scheme; planning the task set for each UAV in each sub-area, obtaining a task allocation scheme; integrating the sub-area division scheme, the vehicle path planning scheme and the task allocation scheme, obtaining the initial solution of the collaborative task planning, and adopting an optimization method based on a mixture of simulated annealing and large-scale neighborhood search to obtain a vehicle-mounted multi-UAV collaborative task planning strategy. The present invention is applied to the field of task planning and scheduling, can significantly shorten the task time, and solve the problem of regional surface coverage in regional reconnaissance.

Description

Technical Field

[0001] The present invention relates to the technical field of mission planning and scheduling, and in particular to a vehicle-mounted multi-UAV collaborative mission planning method for large-scale regional reconnaissance. Background Art

[0002] With the development and maturity of drone technology, drones have begun to play an important role in more and more application scenarios, and have gradually become a common tool for reconnaissance in target areas. Drone reconnaissance is an activity that uses drones to obtain detailed information in an area. It has become an indispensable technical support for major tasks such as strengthening social security, assisting human exploration of nature, and participating in emergency search and rescue.

[0003] In the actual regional reconnaissance process, various situations are complex and changeable. Using a drone often faces many difficulties and limitations. At this time, multiple drones are generally used to cover the target area in a coordinated mode. Multi-drone systems have many advantages, such as reducing the total mission time due to the division of coverage work; introducing fault tolerance, when a drone fails, its coverage area can be assigned to another drone. The application of multi-drone collaborative mode to regional coverage has attracted widespread attention from scholars at home and abroad. UAV regional reconnaissance research includes several aspects such as point coverage for reconnaissance missions of fixed ground targets and surface coverage for certain specific areas. In the planning of regional reconnaissance missions for larger areas, regional division, drone task allocation, and the coordination between vehicles and drones are important research contents.

[0004] At the same time, as the scale of the region continues to expand, the number of reconnaissance missions continues to increase. It is difficult for a single drone to complete the reconnaissance mission in a large area alone. The collaborative execution of multiple drones can expand the ability to execute tasks through the cooperation between members, handle various complex situations, shorten the time to complete tasks, and improve the efficiency of task completion. However, the collaborative execution of multiple drones is bound to increase the difficulty of processing, and its complexity increases exponentially with the increase in the number of drones and tasks. Therefore, it is necessary to design a corresponding task allocation algorithm to solve the regional reconnaissance problem of single vehicle and multiple drones. Task allocation algorithms are divided into two categories: exact algorithms and metaheuristic algorithms. Exact algorithms mainly include dynamic programming algorithms, branch and bound methods, column generation methods, exhaustive methods, etc. These traditional algorithms have been studied more fully, have more complete theoretical support, and have developed many commercial software to implement them. They usually have better results when the problem scale is small. The metaheuristic algorithms used to solve this type of problem mainly include genetic algorithms (GA), particle swarm optimization (PSO), ant colony optimization (ACO), tabu search algorithm (TS), etc.

[0005] In practical applications, small drones are generally used to cover and scan the entire reconnaissance target area. Small drones have the advantages of low cost, good concealment, and low-altitude high-precision scanning. They are widely used in regional coverage scanning in the military and civilian fields. However, small drones have poor endurance, and the area covered by a single drone in one flight is very limited. Introducing the vehicle-machine collaborative mode for regional reconnaissance operations can greatly improve the efficiency of regional reconnaissance work and expand the reconnaissance range. The vehicle-machine collaborative mode refers to a mode in which ground vehicles and drones work together as a community to complete tasks such as delivery / collection of items, reconnaissance, inspection, and monitoring. Compared with the use of vehicles or drones alone, the vehicle-machine collaborative mode allows vehicles and drones to complement each other's advantages to improve work efficiency. The vehicle-machine collaborative mode has been widely used in real life, such as the collection, detection and monitoring of specific target intelligence, the delivery of packages, and the inspection of circuits. In terms of collaborative methods, there are four main ways of vehicle-machine collaboration: drones and vehicles as synchronous working units, drones and vehicles operating independently, drone-based working units and vehicle-assisted working units, and vehicle-based working units and drone-assisted working units. In terms of problem models, according to the number of drones and vehicles, the vehicle-machine collaboration problem model can be divided into the following three categories: single-vehicle single-machine model, single-vehicle multi-machine model, and multi-vehicle multi-machine model. In terms of algorithms, there are two main heuristic algorithm design ideas: "drone first, then vehicle" and "vehicle first, then drone".

[0006] In the current existing technologies, the single-vehicle multi-machine collaborative mode can effectively improve the task efficiency. At present, there are few studies on the application of this mode in regional reconnaissance, and this research has great application prospects, so it is of practical significance. When considering the regional coverage of multiple drones, there are three ideas for regional division: gridding the region separately; planning the drone path for the entire region, cutting the entire regional path according to the drone range to form sub-regions; and implementing polygon decomposition on irregular regions to obtain sub-regions. At present, most of the existing technologies use one of the above three ideas alone, and rarely combine the above ideas for sub-region division. In addition, in drone task allocation, the exact algorithm usually has a better effect when the problem scale is small. When the problem scale increases, its solution time will increase sharply, and the solution efficiency is low; and the exact algorithm can usually only solve a small number of constraints and model-specific problems. For the actual complex problem application of multi-drone multi-constraint task planning, the ability is still slightly insufficient. Summary of the invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a vehicle-mounted multi-UAV collaborative mission planning method for large-scale regional reconnaissance, which can significantly shorten the mission time.

[0008] To achieve the above object, the present invention provides a vehicle-mounted multi-UAV collaborative mission planning method for large-scale regional reconnaissance, which is characterized by comprising the following steps:

[0009] Step 1: divide the area to be surveyed into a number of grids that can be surveyed by a single drone according to the range of the drone, and divide the area to be surveyed into a number of sub-areas, each of which has a vehicle-drone junction and a number of grids, to obtain a sub-area division scheme;

[0010] Step 2: Based on the sub-area division results, the vehicle's driving path is planned to obtain the order in which the vehicle visits the vehicle-UAV junction points in all sub-areas, i.e., the vehicle path planning scheme;

[0011] Step 3: Based on the sub-area division results, plan the task set for each UAV in each sub-area to obtain a task allocation plan;

[0012] Step 4: The sub-area division scheme, vehicle path planning scheme and task allocation scheme are integrated to obtain the initial solution of collaborative task planning. An optimization method based on a mixture of simulated annealing and large-scale neighborhood search is used to obtain a vehicle-mounted multi-UAV collaborative task planning strategy with the goal of completing all reconnaissance tasks in the shortest time.

[0013] Compared with the prior art, the present invention has the following beneficial technical effects:

[0014] 1. Adopt a vehicle-machine collaborative mode with vehicles as the auxiliary and drones as the main force to expand the scope of small drones to perform reconnaissance missions, thereby completing reconnaissance missions covering larger areas.

[0015] 2. Use a sub-region division method that combines polygon decomposition and gridding to discretize tasks to facilitate task allocation and improve task completion efficiency;

[0016] 3. Perform multi-UAV task allocation based on the averaged task set to minimize the mutual waiting time between UAVs and thus shorten the total task time;

[0017] 3. The adaptive large-area search method with fusion annealing mechanism is used to optimize the initial solution, which can effectively solve the problem of collaborative task planning of multiple vehicle-mounted UAVs and improve the collaborative efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0019] Figure 1 This is an example diagram of a single-vehicle multi-machine collaborative area reconnaissance in an embodiment of the present invention;

[0020] Figure 2 Flow chart of a method for planning a collaborative mission of multiple UAVs mounted on a vehicle in an embodiment of the present invention;

[0021] Figure 3 This is an example diagram of area division in an embodiment of the present invention;

[0022] Figure 4 This is an example diagram of task allocation in an embodiment of the present invention;

[0023] Figure 5 It is a task allocation flow chart in an embodiment of the present invention;

[0024] Figure 6 This is a flow chart of strategy update in an embodiment of the present invention;

[0025] Figure 7 This is an example diagram of the area and the vehicle-drone convergence point in Example 1 of the present invention;

[0026] Figure 8 This is a schematic diagram of the initial sub-region division result of Example 1 of the present invention;

[0027] Fig. 9This is a schematic diagram of the maximum waiting time of a drone in a sub-area based on TSA in Example 1 of the present invention;

[0028] Fig.10 This is a schematic diagram of vehicle and drone path planning based on TSA in Example 1 of the present invention;

[0029] Fig.11 This is a schematic diagram of the maximum waiting time of drones in a sub-area based on ALNSAM, which is an example of the present invention;

[0030] Fig.12 This is a schematic diagram of the total mission time of the ALNSAM method and the TSA method under different numbers of drones in Example 1 of the present invention;

[0031] Fig.13 Example 1 of the present invention is a schematic diagram of the vehicle and drone paths using three drones based on the ALNSAM method;

[0032] Fig.14 This is a schematic diagram of a part of Changsha City and the gridding results according to the example of the present invention.

[0033] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0036] In addition, the technical solutions between the various embodiments of the present invention can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0037] This embodiment discloses a vehicle-mounted multi-UAV collaborative task planning method for large-scale regional reconnaissance, which is used to complete large-scale regional information collection tasks. The method first establishes a collaborative planning model that optimizes the vehicle ground driving path and the multi-UAV collaborative task allocation. On the basis of the established collaborative planning model, the region is divided based on gridding and polygon decomposition to obtain a sub-region division scheme, and then the vehicle path planning scheme and the task allocation scheme considering the multi-UAV task set are obtained according to the sub-region division scheme, so that the three constitute the initial solution of the collaborative task planning. On this basis, an optimization method based on a mixture of simulated annealing and large-scale neighborhood search is adopted, with the goal of completing all reconnaissance tasks in the shortest time, to obtain a vehicle-mounted multi-UAV collaborative task planning strategy.

[0038] like Figure 1 The following is a simple example of a single-vehicle multi-machine coordinated area reconnaissance. The task completion condition is Figure 1 The area shown in the figure is covered and scanned to detect the corresponding information of this area. Assuming that there is a known road network in this area, there are several known vehicle-drone convergence points on the road network where vehicles can park and drones can take off and land. It is necessary to use drones and vehicles to complete full coverage reconnaissance of the area to collect the corresponding information of this area. In order to facilitate the task allocation of drones, the entire area is gridded, and the size of each grid is the maximum scanning area of ​​a single drone at a time. A single drone scans one grid to complete a task. Since the area is large and the range of a single drone is short, in order to reduce the time of regional reconnaissance, multiple drones are used to cover the entire area, and vehicles are introduced as mobile platforms and charging stations for drones. When drones perform tasks, vehicles wait at the vehicle-drone convergence point and provide battery replacement services until all drones complete their respective tasks and return to the vehicle, and the vehicle carries all drones to the next vehicle-drone convergence point.

[0039] The entire regional reconnaissance mission planning process can be described as follows: In order to facilitate the division of the entire reconnaissance mission, the area is divided into several sub-areas. Each sub-area corresponds to a vehicle-drone rendezvous point. The drones launched at the rendezvous point must complete the coverage of their corresponding sub-areas. A vehicle carrying multiple drones departs from the base and drives along the planned vehicle path. When the vehicle drives to the vehicle-drone rendezvous, all drones are launched and wait for the drones in place. The drones then conduct reconnaissance of the sub-area corresponding to the rendezvous point, and the drones scan in a "Zigzag" manner within the grid. After completing a mission, the drone returns to the vehicle to replace the battery, and then continues to take off and scan the remaining part of the sub-area, and repeats this cycle until all drones complete their respective reconnaissance missions and return to the vehicle, and the vehicle carries all drones to the next vehicle-drone rendezvous point. In this way, all sub-areas have been reconnaissanced and the vehicle carries the drones back to the base station.

[0040] like Figure 1 As shown, the areas enclosed by different solid thick lines represent different sub-areas, the ground driving path of the vehicle is represented by the dashed thick line, the flight trajectory of the drone in each sub-area is represented by the dashed line, and the dashed lines of different forms represent the paths of different drones. During the entire reconnaissance process, the triangle represents the vehicle-drone meeting point in each sub-area, and all drones take off and land at the meeting point, and the square represents the base.

[0041] In this embodiment, the following constraints are made to facilitate calculation:

[0042] (1) Vehicles can only travel on the road network and cannot enter the area to perform reconnaissance missions;

[0043] (2) UAVs can only take off and land at vehicle-UAV rendezvous points visited by vehicles;

[0044] (3) The UAV scans only one grid in one flight, which is called a reconnaissance mission;

[0045] (4) Both the vehicle and the drone are operating at a constant speed;

[0046] (5) Ignore the time for drone takeoff, landing, and battery replacement;

[0047] (6) When the drone is performing a reconnaissance mission, the vehicle stays in place and waits for the drone to land;

[0048] (7) At each vehicle-drone rendezvous point, the vehicle must wait until all drones have completed their reconnaissance missions and returned to the vehicle before it can leave for the next vehicle-drone rendezvous point.

[0049] (8) When the UAV scans the grid, the grid vertex closest to the vehicle-UAV junction is selected as the scanning start point, and the grid vertex second closest to the vehicle-UAV junction is selected as the scanning end point.

[0050] The specific symbols used in this embodiment are shown in Table 1 below:

[0051] Table 1 Symbol Description

[0052]

[0053]

[0054] The collaborative planning model in this embodiment is specifically:

[0055] First, the objective function is to minimize the time to complete all reconnaissance missions. The time Z to complete all reconnaissance missions includes two parts: the first part is the time T that the vehicle carrying multiple drones travels between the base and the vehicle-drone rendezvous point to be visited. G; The second part is the time T that multiple UAVs need to complete their reconnaissance missions in all areas D ,Right now:

[0056] MinZ=T G +T D (1)

[0057] The time T of the vehicle carrying multiple drones in the first part traveling between the base and the meeting point of the vehicles and drones to be visited G Determined by the vehicle's travel path, it is:

[0058] T G =∑ i∈N ∑ j∈N x ij L ij / v G (2)

[0059] The time T required for multiple UAVs in the second part to complete reconnaissance missions in all areas D It is calculated based on the time that the vehicle waits for m drones in each area. Every time a vehicle reaches a vehicle-drone convergence point, it starts timing after releasing the drone until all drones return to the vehicle. The total time is the time to complete the reconnaissance mission in the sub-area. This time is determined by the total mission time of the drone that takes the longest time to execute the mission set in each area, which is:

[0060]

[0061] The sum of the time required for the drones to complete the reconnaissance mission in each sub-area is calculated as the time T required for multiple drones to complete the reconnaissance mission in all areas. D ,Right now:

[0062]

[0063] The constraints of the collaborative planning model are:

[0064] t0=(T max -l D / v D ) (5)

[0065] The above constraint (5) represents the time required to calculate a UAV to scout a grid;

[0066] ∑ j∈N x 0j =∑ i∈N x i0 =1 (6)

[0067] The above constraint (6) indicates that the vehicle carrying the UAV must start from the base and return to the same base after completing all coverage tasks;

[0068] ∑ i,j∈N,i≠j x ij =∑ i,j∈N,i≠j x ji (7)

[0069] The above constraint (7) indicates that the out-degree of each vehicle-UAV junction is equal to the in-degree, thus ensuring the connectivity of the vehicle driving route;

[0070] ∑ i∈N ∑ j∈N x ij =1 (8)

[0071] The above constraint (8) indicates that each vehicle-UAV rendezvous point can only be visited once;

[0072]

[0073] The above constraint (9) means that each grid in each sub-area can only be scouted by one UAV;

[0074]

[0075] The above constraint (10) indicates that the time required for the UAV to complete the reconnaissance of a grid each time shall not exceed its maximum endurance time;

[0076]

[0077] The above constraint condition (11) indicates that the maximum total scanning area of ​​all drones in each area must be greater than or equal to the area of ​​the area, ensuring that the area can be fully covered, that is, ensuring that each grid contained in the area is detected by the drone;

[0078]

[0079] The above constraint condition (12) means that after the vehicle reaches the vehicle-drone junction, all drones on the vehicle are released;

[0080]

[0081]

[0082] The above constraint equations (13) and (14) define the value range of the 0-1 variable.

[0083] refer to Figure 2 Based on the established collaborative planning model, the vehicle-mounted multi-UAV collaborative mission planning method for large-scale regional reconnaissance in this embodiment specifically includes the following steps:

[0084] Step 1: divide the area to be surveyed into a number of grids that can be surveyed by a single drone according to the range of the drone, and divide the area to be surveyed into a number of sub-areas, each of which has a vehicle-drone junction and a number of grids, to obtain a sub-area division scheme;

[0085] Step 2: Based on the sub-area division results, the vehicle's driving path is planned to obtain the order in which the vehicle visits the vehicle-UAV junction points in all sub-areas, i.e., the vehicle path planning scheme;

[0086] Step 3: Based on the sub-area division results, plan the task set for each UAV in each sub-area to obtain a task allocation plan;

[0087] Step 4: The sub-area division scheme, vehicle path planning scheme and task allocation scheme are integrated to obtain the initial solution of collaborative task planning. An optimization method based on a mixture of simulated annealing and large-scale neighborhood search is used to obtain a vehicle-mounted multi-UAV collaborative task planning strategy with the goal of completing all reconnaissance tasks in the shortest time.

[0088] In step 1, the process of obtaining the sub-region division scheme mainly includes two parts: gridding of the area to be surveyed and sub-region division:

[0089] When a drone conducts reconnaissance in an area, the entire area is gridded to facilitate the task assignment of the drone. The specific method is as follows: first analyze the size of the area, then calculate the maximum scanning area of ​​the drone based on the range of the drone (reserving a maneuvering range), and finally divide the entire area to be reconnaissance into unitized grids, with the size of each grid equal to the maximum scanning area of ​​the drone. The scanning range of the drone can be regarded as a square area with a side length of DW.

[0090] Assuming that all drones are of the same type, the maximum scanning area SD of the drone is max ,for:

[0091] SD max =(T max ×v D -l D )×DW (15)

[0092] The area to be detected is then divided into several grids, and the area of ​​each grid is proportional to the maximum scanning area SD of the drone. max same.

[0093] When the drone performs the coverage mission, the vehicle only serves as a charging station for the drone, waiting at the vehicle-drone convergence point in the sub-area and providing battery replacement services until all drones complete their respective tasks and return to the vehicle. The vehicle acts as a mobile platform to carry all drones to the next sub-area to converge with the vehicle-drone. Since the area to be reconnaissance is large, in order to facilitate the division of the entire reconnaissance mission, area A is divided into several sub-areas, each of which corresponds to a vehicle-drone convergence point, and the drone released at the convergence point needs to complete the coverage of its corresponding sub-area. According to the results of regional gridding, the main purpose of sub-area division is to allocate the grids to be scanned for each vehicle-drone convergence point to form a set of reconnaissance sub-areas, i.e., sub-areas, contained in each vehicle-drone convergence point. The strategy adopted in this embodiment is the principle of proximity, that is, the grids are assigned to the vehicle-drone convergence point closest to it. First, the distance matrix between the center point of each grid and each vehicle-drone convergence point is calculated, and then each grid is assigned to the vehicle-drone convergence point closest to it to form a preliminary sub-area division result, i.e., a sub-area division scheme is obtained.

[0094] The sub-region division in this embodiment is described below with reference to specific examples.

[0095] Figure 3 An example of the division of a square area with a side length of 8km is given. Figure 3 (a) shows that the total area of ​​the reconnaissance area is 64km 2 In this area, square 0 represents the base. There are 5 vehicle-drone junctions, represented by triangles and numbered 1-5. The maximum endurance of the drone is 19.5 minutes, the flight speed is 80 km / h, the reserved range is 10 km, and the drone scanning width is 100 m. The maximum scanning area of ​​the drone is 4 km calculated by formula (15) 2 , and the drone takes 12 minutes to scan each grid. Therefore, the entire area can be divided into 16 grids, and the grids are numbered in sequence, such as Figure 3 (b) as shown. Figure 3 (c) is an example of sub-area division. According to the above sub-area division strategy, the sub-area division results are shown in Table 2: the vehicle-drone rendezvous points 1, 2, 3, and 4 are all divided into grids to form four corresponding sub-areas, and the vehicle-drone rendezvous point 5 is not divided into a grid. Therefore, when planning the vehicle path later, the vehicle-drone rendezvous points to be visited are 1, 2, 3, and 4, and the vehicle-drone rendezvous point 5 is not in the vehicle path.

[0096] Table 2 Sub-region division results

[0097]

[0098] After the sub-area division scheme is obtained, the meeting points of each vehicle and drone are also determined, so the vehicle path planning can be regarded as a TSP problem: the vehicle starts from the base, carries a drone to visit the vehicle and drone meeting points in each sub-area in turn, and finally returns to the base. There are many algorithms to solve the TSP problem, such as greedy algorithms, dynamic programming, and meta-heuristic algorithms. The scale of examples adapted to each algorithm, the speed of each solution, and the solution effect are not the same. Since the number of meeting points considered in this embodiment is relatively small, dynamic programming is used to quickly solve the vehicle driving path, and the order of the vehicle's visit to the vehicle and drone meeting points in each sub-area and the length of the vehicle path, that is, the vehicle path planning scheme, are obtained. Dynamic programming to solve the TSP problem is a conventional technical means in the field, so it will not be repeated in this embodiment.

[0099] In step 3, after obtaining the sub-area division scheme, the UAV task allocation in the sub-area is to divide the grid in the sub-area into a number equal to the number of UAVs as the task set of each UAV. The key is how to allocate so that the time for each UAV to perform the task is more evenly distributed, and try to avoid the situation where some UAVs complete the task early and need to return to the vehicle to wait for a long time, thereby shortening the mutual waiting time between UAVs and shortening the total time of the reconnaissance task in the entire area.

[0100] In the process of averaging the time it takes for each drone to complete the task set, the following two points need to be considered when calculating the time it takes for each grid to be covered: the round-trip time between the vehicle-drone junction and the grid and the scanning time of the drone within the grid. In this embodiment, when the drone performs grid scanning, the vertex closest to the vehicle-drone junction among the four vertices of the grid is selected as the scanning start point, and the vertex closest to the vehicle-drone junction except the scanning start point is selected as the scanning end point. The round-trip distance between the vehicle-drone junction and the grid is roughly represented by twice the distance between the vehicle-drone junction and the center of the grid. The round-trip distance divided by the drone speed is the round-trip time between the vehicle-drone junction and the grid. When the drone scans within the grid, it scans in a zigzag manner. Since the grid size is consistent, the time it takes for the drone to scan a grid is fixed.

[0101] Determine the mission set of each drone by dividing the total grid array into m sub-arrays (m is the total number of drones). Therefore, it is necessary to first calculate the time array corresponding to the coverage of each grid in the sub-area, and then create m empty arrays, fill the coverage time of each grid into these m empty arrays in turn, and finally obtain m new arrays. In the process of forming a new array, the sum of the m arrays should be as close as possible, which means that the time for each drone to perform the mission is roughly equal, reducing the waiting time between drones. The m grid number arrays corresponding to the m time arrays correspond to the mission sets of the m drones.

[0102] For example, Figure 4 An example of task allocation is given. Assume that a sub-area contains 9 grids, and the coverage time array of each grid is arranged in descending order as [9, 8, 7, 6, 5, 4, 3, 2, 1]. Three drones are used to complete the reconnaissance of the sub-area, and the average value of the three arrays is calculated to be 15. Then the task set determination process of the three drones in the sub-area is as follows: Figure 4 shown.

[0103] Therefore, reference Figure 5 In this embodiment, the task allocation scheme is obtained by performing the following operations on each sub-region in the sub-region division scheme:

[0104] Step 3.1, calculate the time array required for each grid to be covered by the drone reconnaissance, and sort the time array in descending order to obtain a time array sequence, where each time array is obtained by adding the drone's scanning time in the grid and the drone's round-trip time between the vehicle-drone junction and the grid;

[0105] Step 3.2, get the drone list and calculate the average value avg of all time arrays in the time array sequence, which is:

[0106] avg=SUM / UN

[0107] Where SUM is the sum of all time arrays in the time array sequence, and UN is the number of drones in the drone list;

[0108] Step 3.3, obtain the value num1 of the time array SUM1 that is ranked first in the time array sequence, and determine whether num1<avg is true:

[0109] If yes, go to step 3.4;

[0110] Otherwise, assign the grid of the time array to any drone UN in the drone list x and will use drones x Delete from the drone list, delete the time array SUM1 from the time array sequence, and go to step 3.10;

[0111] Step 3.4, assign the grid of the time array SUM1 to any drone UN in the drone list x and will use drones x Delete it from the drone list, set delta = avg-num1 and go to step 3.5;

[0112] Step 3.5, traverse the time array sequence and determine whether there is a time array SUM k Satisfy delta=num k :

[0113] If so, then the time array SUM k The grid is assigned to the drone UN x , SUM the time array k After deleting from the time array sequence, go to step 3.10;

[0114] Otherwise, go to step 3.6;

[0115] Step 3.6, get the maximum value num in the current time array sequence max , judge delta≥num max Is it established:

[0116] If so, set num max The corresponding time array SUM max The grid is assigned to the drone UN x , SUM the time array max Delete from the time array sequence and set delta = delta-num max Then, return to step 3.5;

[0117] Otherwise, go to step 3.7;

[0118] Step 3.7, get the minimum value num in the current time array sequence min , judge delta≤num min Is it established:

[0119] If so, set num min The corresponding time array SUM min The grid is assigned to the drone UN x , SUM the time array min After deleting from the time array sequence, go to step 3.10;

[0120] Otherwise, go to step 3.8;

[0121] Step 3.8, get the adjacent and satisfying num in the current time array sequence a>delta>num b SUM of time array a 、SUM b , judge (delta-num b )>(num a -delta) is true:

[0122] If so, sum the time array b The grid is assigned to the drone UN x , SUM the time array b Delete from the time array sequence and set delta = delta-num b Then, return to step 3.5;

[0123] Otherwise, go to step 3.9;

[0124] Step 3.9, let distance = num a -delta, determine whether num exists in the current time array sequence x Satisfy num x <distance:

[0125] If so, set num x The corresponding time array SUM x The grid is assigned to the drone UN x , SUM the time array x Delete from the time array sequence and set delta = delta-num x Then, return to step 3.5;

[0126] Otherwise, the time array SUM a The grid is assigned to the drone UN x , SUM the time array a After deleting from the time array sequence, go to step 3.10;

[0127] Step 3.10, determine whether the current time array sequence is empty:

[0128] If yes, then output the task allocation plan for the current sub-area;

[0129] Otherwise, return to step 3.2.

[0130] By performing steps 3.1 to 3.10 for each sub-region, a task allocation solution can be obtained.

[0131] From the above steps 1-3, it can be seen that the sub-region division scheme obtained by region division is the basis for vehicle path planning and drone task allocation. How to divide the sub-region has a great impact on the total vehicle travel time and the total drone reconnaissance mission time, which is reflected in the following two aspects: (1) How to divide the sub-region affects the determination of the vehicle-drone rendezvous point to be visited, thereby affecting the vehicle path and vehicle travel time; (2) How to divide the sub-region affects the allocation of drone task sets within the sub-region, thereby affecting the drone reconnaissance mission completion time. Since the number of sub-regions is equal to the number of vehicle-drone rendezvous points to be visited, reducing the number of vehicle-drone rendezvous points to be visited, that is, reducing the number of sub-regions, is the key to shortening vehicle travel time. On the other hand, since multiple drones perform reconnaissance missions in coordination and drones may need to wait for each other after completing the mission, minimizing the waiting time between drones is the key to shortening the drone reconnaissance mission time. Therefore, the optimization direction of the initial feasible solution is to re-divide the sub-regions to reduce the vehicle travel time and the drone reconnaissance mission time.

[0132] Through the task allocation process in step 3, we can know the maximum difference in the task time of different drones in each sub-area. The difference in drone task time is due to the different task sets of each drone in the sub-area, which leads to different total time for each drone to complete all tasks. The drone that takes the least time to complete the task in the nth sub-area is Time is recorded as The drone that takes the longest time to complete the mission is Time is recorded as The biggest gap is So choose Sub-areas with a value greater than GT are optimized, and the grids in these sub-areas are redistributed to shorten the UAV mission time gap as much as possible. In this process, the driving time of the vehicle visiting the convergence point in all sub-areas and the time for the UAV to perform the reconnaissance mission are comprehensively considered to minimize the total time. Therefore, in order to improve the quality of the solution of the vehicle-mounted multi-UAV collaborative mission planning strategy, the following two removal operators and two reconstruction operators are designed in combination with different removal and reconstruction strategies in the optimization method. Any removal operator and a reconstruction operator are combined into a neighborhood operator, so there are four neighborhood operators in this embodiment:

[0133] In this embodiment, the removal operator set includes:

[0134] Random removal operator: randomly select a grid to remove in each sub-region in the sub-region partitioning scheme;

[0135] Optimal removal operator: In each sub-area in the sub-area division scheme, the grid farthest from the corresponding vehicle-UAV junction is selected for removal.

[0136] The set of reconstruction operators includes:

[0137] Random reconstruction operator: randomly add the removed grids to the sub-region where the vehicle-drone junction is located within the remaining range of the drone;

[0138] Greedy reconstruction operator: Add the removed meshes to the sub-region that can optimize the current solution.

[0139] In step 4, this embodiment designs an optimization method based on a mixture of simulated annealing and large-scale neighborhood search, with the goal of completing all reconnaissance tasks in the shortest time possible, and obtains a vehicle-mounted multi-UAV collaborative task planning strategy. Figure 6 ,The specific process of optimizing the vehicle-mounted multi-UAV collaborative task planning strategy is:

[0140] Step 4.1, obtain the initial temperature T s , termination temperature T e , annealing rate cr, maximum number of consecutive optimal solutions N max , remove the operator set, reconstruct the operator set and the initial solution s of the collaborative task planning;

[0141] Step 4.2, let s b =s, T c =T s , count = 0, where s b is the optimal solution of collaborative task planning, s′ is the current solution of patrol path planning strategy, Tc is the current temperature, and count is the number of consecutive optimal solutions;

[0142] Step 4.3, according to the roulette algorithm, select the removal operator and reconstruction operator with the largest weight in the removal operator set and the reconstruction operator set, and adjust the sub-area division scheme in the initial solution s according to the selected removal operator and reconstruction operator to obtain a new sub-area division scheme, and obtain a new vehicle path planning scheme and a new task allocation scheme based on the new sub-area division scheme, that is, obtain the current solution s of the collaborative task planning t ;

[0143] Step 4.4, calculate the current solution s respectively t , optimal solution b The task completion time t(s t )、t(s b ), judge t(s t )<t(s b ) is true:

[0144] If so, let s = s t 、s b =s t, after count=0, proceed to step 4.7;

[0145] Otherwise, set count = count + 1 and go to step 4.5;

[0146] Step 4.5, determine t(s t )>t(s b ) is true:

[0147] If yes, go to step 4.6;

[0148] Otherwise, t(s t )=t(s b ), let s = s t , proceed to step 4.7;

[0149] Step 4.6, obtain the random variable rand from 0 to 1, and determine whether rand < exp(t(s t )-t(s b )) / T c Is it established:

[0150] If so, let s = s t Then proceed to step 4.7;

[0151] Otherwise, proceed to step 4.7

[0152] Step 4.7, determine T c >T e And count<N max Is it established:

[0153] If so, let T c =cr×T c , after updating the weights of the removal operator and the reconstruction operator, return to step 4.3;

[0154] Otherwise, output the current s b , as a collaborative mission planning strategy for vehicle-mounted multi-UAVs.

[0155] In step 4.7, update the weights of the removal operator and the reconstruction operator, specifically:

[0156]

[0157] In the formula, ω d To remove the weight of an operator or reconstruct the weight of an operator, u d is the number of times the operator is removed or reconstructed, ρ is the weight update coefficient, and s d is the score of the removal operator or reconstruction operator;

[0158] At the beginning of the iteration, the weights and scores of each removal operator and reconstruction operator are the same. During each iteration, the scores are increased step by step according to the performance of the removal operator and / or reconstruction operator, as follows:

[0159] If a new global optimal solution is obtained after destruction / reconstruction, the corresponding removal operator and reconstruction operator will be increased by 1.5 points;

[0160] If no new global optimal solution is obtained after destruction / reconstruction, but it has not been accepted and is better than the current solution, the corresponding removal operator and reconstruction operator will increase by 1.2 points;

[0161] If the solution obtained after destruction / reconstruction is not accepted and is worse than the current solution, but is updated to the current solution, the corresponding removal operator and reconstruction operator will increase by 0.8 points;

[0162] If the solution obtained after destruction / reconstruction is not accepted before and is worse than the current solution, and has not been updated to the current solution, the corresponding removal operator and reconstruction operator will increase by 0.6 points.

[0163] The following four simulation cases and one actual case (a reconnaissance case of some areas in Changsha’s main urban area) are used to test the advantages of the single-vehicle multi-machine mode in large-area reconnaissance and the effectiveness and performance of the adaptive large-area search method that integrates the annealing mechanism.

[0164] The experimental parameter settings are shown in Table 3. The average speed of the vehicle is set to 40 km / h, and the average speed of the drone is set to 80 km / h. The scanning distance of the drone is set to 100 m. The drone endurance time is 19.5 mins, and the reserved range is 10 km. Therefore, the maximum coverage area SD of each drone can be obtained. max =4km 2 , it takes 12 minutes for a drone to scan a grid. The initial temperature in the adaptive large-area search method with annealing mechanism is 1000°, the termination temperature is 10°, the cooling rate is 0.97, the number of times the optimal solution does not change during continuous iterations is 100, and the initial weight of each removal and reconstruction operator is 1.

[0165] Table 3 Parameter settings

[0166]

[0167]

[0168] Generate a 12km×12km square area on the plane. The entire area can be divided into 36 grids by the area division method, such as Figure 7As shown, the 36 grids are numbered from left to right and from bottom to top, 1-36. Assuming that the center point of each grid in the area is on the road network in the area, 6 vehicle-drone convergence points are randomly generated on the road network to ensure that the distance from the center of each grid in the area to its nearest convergence point is less than or equal to half of the remaining range of the drone (so that each grid is within the range of the drone and can be detected). The center of one side of the square area is set as the base.

[0169] Example 1 Figure 7 As shown in the figure, the randomly generated vehicle-drone rendezvous points are numbered clockwise from 1 to 6, represented by black triangles. The red rectangle represents the base, numbered 0. The vehicle carrying the drone sets out from the base to scan the area composed of 36 grids, and returns to the base after completing all scanning tasks.

[0170] Based on the sub-area division in step 1, each vehicle-UAV convergence point is initially divided into its corresponding sub-area. The sub-area division results are shown in Table 4 and Figure 8 shown.

[0171] Table 4 Results of initial sub-region division in Example 1

[0172]

[0173]

[0174] Then, based on the sub-area division results, the vehicle path planning method (step 2) is used to plan the vehicle path, and the visit order of the 6 vehicle-drone junctions is 0-1-2-3-5-4-6-0. The total length of the vehicle driving path is 46 km, and the time taken is 4140 s.

[0175] Then, based on the sub-area division results, the task allocation method (step 3) is used to assign tasks to each drone in each sub-area, obtain the task set of each drone in each sub-area and calculate the time required to complete the drone reconnaissance in each sub-area. In this experiment, the cases of using 1, 2, 3, 4, 5, 6, and 7 drones are analyzed in turn. The TSA method obtains the vehicle driving time, drone reconnaissance time, and total mission time as shown in Table 5. At the same time, the maximum difference in the reconnaissance time of different drones in each sub-area is recorded. like Fig. 9 As shown in the figure, it can be seen that in some sub-areas, the drones have a long waiting time (greater than 450s), and there is room for further optimization.

[0176] Table 5 Total mission time for different numbers of drones in Example 1

[0177]

[0178] Taking the case of using three drones as an example, Table 6 shows the drone task allocation table and drone reconnaissance time in each sub-area under this situation. The drone path, vehicle path and drone task allocation results obtained by the three-stage method are shown in Figure 6. Fig.10 As shown, the total time to complete the reconnaissance mission is 16676s.

[0179] Table 6 Task allocation table for example 1 using 3 drones

[0180]

[0181] Based on the three-stage algorithm (TSA) of steps 1-3, the initial feasible solution obtained by area division, vehicle path planning, and UAV task allocation, the vehicle path planning method and the task allocation method are used to optimize the initial feasible solution using the adaptive large neighborhood search algorithmbased on annealing mechanism (ALNSAM). The optimized vehicle driving time, UAV reconnaissance time and total mission time are shown in Table 7 after optimizing the initial feasible solutions of 2, 3, 4, 5, 6, and 7 UAVs in Example 1. The experimental results in each case are the average of 5 experiments. At the same time, the maximum difference in the reconnaissance time of different UAVs in each sub-area after optimization is recorded. like Fig.11 shown.

[0182] Table 7 Total mission time for different numbers of drones in Example 1

[0183]

[0184] The effect of ALNSAM on TSA mission efficiency improvement before and after optimization is shown in Table 8. When 2, 3, 4, 5, 6, and 7 drones are used, the total mission time is shortened by 1.32%, 7.75%, 8.62%, 9.66%, 11.89%, and 15.67%, respectively. It can be seen that there is a certain optimization effect. The maximum difference in mission time of different drones in each sub-area after optimization It can be seen that the maximum waiting time of the drones in the sub-area to be optimized has been greatly reduced, which is also the main source of the reduction in total mission time. Compared with the single-vehicle single-machine mode, the change in total mission time in the single-vehicle multi-machine mode is as follows: Fig.12 shown.

[0185] Table 8 Example 1 The improvement effect of ALNSAM method on TSA method under different numbers of drones

[0186]

[0187]

[0188] Taking the case of using three drones as an example, Table 9 shows the sub-area division results based on the ALNSAM method in this case. Table 10 shows the drone task allocation table and drone reconnaissance time in each sub-area. The drone path, vehicle path and drone task allocation results obtained by the ALNSAM method are shown in Figure 1. Fig.13 As shown, the total mission time to complete the reconnaissance mission is 15310s.

[0189] Table 9 Sub-area division results of Example 1 using 3 UAVs based on the ALNSAM method

[0190]

[0191] Table 10 Task allocation table of Example 1 using 3 UAVs based on ALNSAM method

[0192]

[0193] The experimental results show that when the vehicle-mounted single-UAV mode is used to cover the area of ​​the problem, the total mission time calculated for Example 1 is 36200s. The time required for using 2, 3, 4, 5, 6, and 7 UAVs is reduced by 42.97%, 57.50%, 65.28%, 69.35%, 70.28%, and 71.98% respectively. It can be seen that in the execution of multiple large-area coverage tasks, the use of a single-vehicle multi-machine mode can significantly reduce the mission time compared with a single-vehicle single-machine mode, and efficiently complete large-area reconnaissance tasks; the ALNSAM method and the TSA method can effectively solve the complex path planning problems caused by vehicle-machine collaboration and meet the actual application needs. In Example 1, the ALNSAM method can shorten the total planning time of large-area reconnaissance tasks by about 9.15% based on the TSA method.

[0194] Based on the above processing, three more cases are generated, numbered as Case 2, Case 3, and Case 4. They are all square areas with an area of ​​196 km. 2 、256km 2 、324km 2. First, the TSA method is used to solve them respectively to obtain the initial feasible solution. The total mission time and basic information of each example under the TSA method when using 3, 4, 5, 6, and 7 drones are shown in Table 11. Based on the initial feasible solution, the ALNSAM method is used for optimization. The total mission time of each example under different numbers of drones is shown in Table 12. The improvement effect of the ALNSAM method on the TSA method in each example is shown in Table 13. On the basis of 3 drones, the experimental results are improved with each increase in the number of drones as shown in Table 14.

[0195] Table 11 Basic information of each example and total task time of initial feasible solution

[0196]

[0197] Table 12 Total mission time for each example based on the ALNSAM method with different numbers of UAVs

[0198]

[0199] Table 13 Improvement effect of ALNSAM method on TSA method in each case

[0200]

[0201] Table 14 Schematic diagram of efficiency improvement brought by increasing the number of drones in each case

[0202]

[0203] The experimental results show that the ALNSAM method has a better optimization effect in examples of different sizes. Compared with the TSA method, the ALNSAM method can shorten the total mission time by an average of 10.02%. The total mission time decreases with the increase in the number of UAVs. In the ALNSAM method, the average improvement of the experimental results of 4 UAVs compared with 3 UAVs, 5 UAVs compared with 4 UAVs, 6 UAVs compared with 5 UAVs, and 7 UAVs compared with 6 UAVs are 17.68%, 11.44%, 5.58%, and 6.57%, respectively.

[0204] The following is an analysis of the example:

[0205] The effectiveness and performance of the model and method of the present invention are verified based on the real road network data of Changsha. Fig.14As shown in the figure, a 20km×20km area in Changsha city is intercepted on the map and divided into 100 2km×2km grids, which are numbered 1-100. A base is set on the road network in the area, which is numbered 0, and 15 vehicle-drone rendezvous points for vehicles to park and launch and recover drones are numbered 1-15. The actual distance between each vehicle-drone rendezvous point on the road network is extracted to form a distance matrix.

[0206] Using the TSA method, the high-quality initial feasible solution of the example is obtained when 3, 4, 5, 6, and 7 drones are used for regional reconnaissance, and then the ALNSAM method is used to optimize the initial feasible solution. Taking the case of using 4 drones as an example, the total mission time of the initial feasible solution obtained by the TSA method is 38636s, and the final regional reconnaissance mission planning solution obtained by the ALNSAM method is as follows: The sub-region division results are shown in Table 15, where node 2 is not assigned to the grid, so it does not need to be visited by the vehicle. The vehicle visits 14 vehicle-drone junctions in the area, the vehicle path is 0-1-3-4-5-6-7-8-9-10-11-12-13-14-15-0, and the vehicle driving time is 9900s; the drone task allocation results and reconnaissance time in each sub-region are shown in Table 16, and the total drone mission time is 25008s; the total reconnaissance mission time of the entire area is the sum of the vehicle driving time and the drone total mission time, which is 34908s. Compared with the TSA method, the ALNSAM method shortens the total time of regional reconnaissance mission by 9.65%.

[0207] Table 15 Sub-area division results based on ALNSAM in the example

[0208]

[0209] Table 16 Results of drone task allocation in each sub-region based on ALNSAM in the example

[0210]

[0211] Based on the TSA method and the ALNSAM method, experiments were conducted. When 3, 4, 5, 6, and 7 drones were used, the comparison of the drone task completion time of the two methods and the improvement of the ALNSAM method on the initial solution were shown in Table 17. The experimental results show that when 3, 4, 5, 6, and 7 drones were used in this example, the ALNSAM method can shorten the total task time by an average of 3.63%, 9.65%, 10.84%, 10.35%, and 18.62% respectively compared with the TSA method. It can be seen that the ALNSAM method has a good optimization effect. On the basis of 3 drones, the improvement of the experimental results brought about by each increase in the number of drones is shown in Table 18. It can be seen that in the ALNSAM method, the improvement of the experimental results of 4 drones compared to 3 drones, 5 drones compared to 4 drones, 6 drones compared to 5 drones, and 7 drones compared to 6 drones are 17.76%, 10.40%, 7.32%, and 5.00% respectively, showing a marginal diminishing effect.

[0212] Table 17 Comparison of total task time between TSA method and ALNSAM method in examples

[0213]

[0214] Table 18 Schematic diagram of efficiency improvement brought by increasing the number of drones in the example

[0215]

[0216] Through the above examples and examples, it can be seen that the present invention is aimed at the problem of regional reconnaissance mission planning for vehicles and multiple UAVs. Taking minimizing the total mission time as the objective function, a corresponding mathematical model is constructed, and a three-stage method (TSA) is designed to obtain a high-quality initial solution. On this basis, an adaptive large neighborhood search method (ALNSAM) with annealing mechanism is designed to optimize the initial solution. It has the following advantages:

[0217] (1) Vehicle-machine cooperative regional reconnaissance system In the vehicle-machine cooperative mode with vehicles as auxiliary and drones as main, the car serves as a mobile platform and charging station for drones, carrying drones to perform large-scale maneuvers in the area, which can effectively expand the mission range of small drones and improve the efficiency of regional reconnaissance. At the same time, the single-vehicle multi-machine mode can significantly shorten the mission time compared with the single-vehicle single-machine mode, and the mission completion time decreases with the increase in the number of drones. Experiments show that the total mission time required for the single-vehicle multi-machine mode can be reduced by 42.97%-71.98% compared with the single-vehicle single-machine mode; in the experiments using 3, 4, 5, 6, and 7 drones respectively, the total mission time can be shortened by 17.70%, 11.24%, 5.92%, and 6.26% respectively on the basis of the previous experiment.

[0218] (2) The idea of ​​combining polygon decomposition with gridding can discretize tasks when conducting large-area reconnaissance, facilitate task allocation, and improve task completion efficiency. The multi-UAV task allocation method that considers the average task set can minimize the mutual waiting time between UAVs and thus shorten the total task time.

[0219] (3) Both the three-stage method designed for the problem and the adaptive large-domain search method with annealing mechanism can solve the problem. The adaptive large-domain search method with annealing mechanism is based on the three-stage method and combines the neighborhood search strategy and adaptive criteria in ALNS with the Metropolis strategy in SA to optimize the initial solution. Experiments show that the ALNSAM method can shorten the total task time of the initial solution by about 10% on average.

[0220] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A vehicle-mounted multi-UAV collaborative mission planning method for large-scale regional reconnaissance, characterized in that: The steps include: Step 1: divide the area to be surveyed into a number of grids that can be surveyed by a single drone according to the range of the drone, and divide the area to be surveyed into a number of sub-areas, each of which has a vehicle-drone junction and a number of grids, to obtain a sub-area division scheme; Step 2: Based on the sub-area division results, the vehicle's driving path is planned to obtain the order in which the vehicle visits the vehicle-UAV junction points in all sub-areas, i.e., the vehicle path planning scheme; Step 3: Based on the sub-area division results, plan the task set for each UAV in each sub-area to obtain a task allocation plan, which specifically includes performing the following operations on each sub-area in the sub-area division plan: Step 3.1, calculate the time array required for each grid to be covered by the drone reconnaissance, and sort the time array in descending order to obtain a time array sequence, where each time array is obtained by adding the drone's scanning time in the grid and the drone's round-trip time between the vehicle-drone junction and the grid; Step 3.2, get the drone list and calculate the average of all time arrays in the time array sequence ,for: In the formula, is the sum of all time arrays in the time array sequence, is the number of drones in the drone list; Step 3.3, get the time array that is sorted first in the time array sequence Value ,judge Is it established: If yes, go to step 3.4; Otherwise, assign the grid of the time array to any drone in the drone list. , and the drone Delete from the drone list and set the time array After deleting from the time array sequence, go to step 3.10; Step 3.4: Set the time array Assign the grid to any drone in the drone list , and the drone Remove from the drone list and make Then proceed to step 3.5; Step 3.5, traverse the time array sequence to determine whether there is a time array satisfy : If so, then the time array Grids are assigned to drones , the time array After deleting from the time array sequence, go to step 3.10; Otherwise, go to step 3.6; Step 3.6, get the maximum value in the current time array sequence ,judge Is it established: If so, The corresponding time array Grids are assigned to drones , the time array Remove from the time array sequence and let Then, return to step 3.5; Otherwise, go to step 3.7; Step 3.7, get the minimum value in the current time array sequence ,judge Is it established: If so, The corresponding time array Grids are assigned to drones , the time array After deleting from the time array sequence, go to step 3.10; Otherwise, go to step 3.8; Step 3.8, get the adjacent and satisfying Time array , ,judge Is it established: If so, set the time array Grids are assigned to drones , the time array Remove from the time array sequence and let Then, return to step 3.5; Otherwise, go to step 3.9; Step 3.9, let , to determine whether the current time array sequence exists satisfy : If so, The corresponding time array Grids are assigned to drones , the time array Remove from the time array sequence and let Then, return to step 3.5; Otherwise, the time array Grids are assigned to drones , the time array After deleting from the time array sequence, go to step 3.10; Step 3.10, determine whether the current time array sequence is empty: If yes, then output the task allocation plan for the current sub-area; Otherwise, return to step 3.2; Step 4: The sub-area division scheme, vehicle path planning scheme and task allocation scheme are integrated to obtain the initial solution of collaborative task planning. An optimization method based on a mixture of simulated annealing and large-scale neighborhood search is used to obtain a vehicle-mounted multi-UAV collaborative task planning strategy with the goal of completing all reconnaissance tasks in the shortest time.

2. The method for vehicle-mounted multi-UAV collaborative mission planning for large-scale regional reconnaissance according to claim 1 is characterized in that: Step 1 specifically includes: Step 1.1, based on a reserved maneuvering range, calculate the maximum scanning area of ​​the drone ,for: In the formula, is the maximum flight time of the drone, is the flight speed of the drone, Reserve range for drones. is the sweep width of the drone; Step 1.2: Divide the area to be surveyed into several grids, and the area of ​​each grid is proportional to the maximum scanning area of ​​the drone. same; Step 1.3, calculate the distance between each grid center point and each vehicle-UAV rendezvous point in the road network, and assign each grid to the vehicle-UAV rendezvous point closest to it, and use the vehicle-UAV rendezvous point assigned with the grid as the vehicle to-be-visited point, thus obtaining the sub-area division scheme.

3. The method for vehicle-mounted multi-UAV collaborative mission planning for large-scale regional reconnaissance according to claim 2 is characterized in that: In step 2, dynamic programming is used to obtain the vehicle path planning solution.

4. The method for vehicle-mounted multi-UAV collaborative mission planning for large-scale regional reconnaissance according to any one of claims 1 to 3, characterized in that: In step 4, the optimization process of the vehicle-mounted multi-UAV collaborative mission planning strategy is: Step 4.1, get the initial temperature , termination temperature , annealing rate , the maximum number of consecutive optimal solutions , remove the operator set, reconstruct the operator set and the initial solution of collaborative task planning ; Step 4.2, let , , ,in, The optimal solution for collaborative task planning, is the current solution of the patrol path planning strategy, is the current temperature, is the number of consecutive optimal solutions; Step 4.3, select the removal operator and reconstruction operator with the largest weight in the removal operator set and reconstruction operator set according to the roulette algorithm, and perform the initial solution according to the selected removal operator and reconstruction operator. The sub-area division scheme in the algorithm is adjusted to obtain a new sub-area division scheme, and a new vehicle path planning scheme and a new task allocation scheme are obtained based on the new sub-area division scheme, that is, the current solution of collaborative task planning is obtained. ; Step 4.4, calculate the current solution respectively , optimal solution Task completion time , ,judge Is it established: If so, then , , Then proceed to step 4.7; Otherwise, let Then, proceed to step 4.5; Step 4.5, judge Is it established: If yes, go to step 4.6; Otherwise, let , proceed to step 4.7; Step 4.6, get a random variable between 0 and 1 ,judge Is it established: If so, then Then proceed to step 4.7; Otherwise, proceed to step 4.7 Step 4.7, judge and Is it established: If so, then , after updating the weights of the removal operator and the reconstruction operator, return to step 4.3; Otherwise, output the current , as a collaborative mission planning strategy for vehicle-mounted multi-UAVs.

5. The method for vehicle-mounted multi-UAV collaborative mission planning for large-scale regional reconnaissance according to claim 4 is characterized in that: The removal operator set includes: Random removal operator: randomly select a grid to remove in each sub-region in the sub-region partitioning scheme; Optimal removal operator: In each sub-area in the sub-area division scheme, the grid farthest from the corresponding vehicle-UAV junction is selected for removal.

6. The method for vehicle-mounted multi-UAV collaborative mission planning for large-scale regional reconnaissance according to claim 5 is characterized in that: The reconstruction operator set includes: Random reconstruction operator: randomly add the removed grids to the sub-region where the vehicle-drone junction is located within the remaining range of the drone; Greedy reconstruction operator: Add the removed meshes to the sub-region that can optimize the current solution.

7. The method for vehicle-mounted multi-UAV collaborative mission planning for large-scale regional reconnaissance according to claim 4 is characterized in that: In step 4.7, the weights of the update removal operator and the reconstruction operator are specifically: In the formula, To remove an operator or reconstruct the weight of an operator, To remove an operator or reconstruct the number of times an operator is used, is the weight update coefficient, is the score of the removal operator or reconstruction operator; At the beginning of the iteration, the weights and scores of each removal operator and reconstruction operator are the same. During each iteration, the scores are increased step by step according to the performance of the removal operator and / or reconstruction operator, as follows: If a new global optimal solution is obtained after destruction / reconstruction, the corresponding removal operator and reconstruction operator will be increased by 1.5 points; If no new global optimal solution is obtained after destruction / reconstruction, but it has not been accepted and is better than the current solution, the corresponding removal operator and reconstruction operator will increase by 1.2 points; If the solution obtained after destruction / reconstruction is not accepted and is worse than the current solution, but is updated to the current solution, the corresponding removal operator and reconstruction operator will increase by 0.8 points; If the solution obtained after destruction / reconstruction is not accepted before and is worse than the current solution, and has not been updated to the current solution, the corresponding removal operator and reconstruction operator will increase by 0.6 points.

Images