A Ground Search Team Deployment Algorithm Based on Genetic Algorithm

Abstract

The present invention relates to a ground search team deployment algorithm based on a genetic algorithm. First, a deployment model of a team is established by modeling a search area and a search team. The deployment problem of the team is converted into a mathematical optimization problem with the shortest time for the team to arrive at a spacecraft as the goal and the grid in a safe area as the constraint condition for the team's initial deployment position. Finally, a deployment plan is obtained by solving the genetic algorithm and a multi-resolution method. The present invention provides a mathematical theoretical basis for the deployment problem of a ground search team, solves the problem of calculating the deployment plan by using a genetic algorithm, and improves the solution efficiency by using a multi-resolution grid technology, which is conducive to ensuring the safety of the team, improving the search efficiency of the team, and shortening the time required to complete the task.

Description

Technical Field

[0001] The invention relates to the field of aerospace search and rescue, in particular to a ground search team deployment algorithm based on a genetic algorithm. Background Art

[0002] With the deepening of my country's manned space program and lunar exploration program, the search and recovery missions such as manned spacecraft return capsules and lunar exploration returners are showing a trend of high frequency, normalization, and increasing emergency events. This requires the aerospace ground search team to be able to quickly reach the returner landing point and efficiently complete the search mission. The deployment of the ground search team refers to the deployment of each team's standby positions before the start of the aerospace search mission.

[0003] The existing ground search team deployment method is based on factors such as the returner aiming point, landing area, and search team vehicle performance, and is obtained by referring to previous mission experience. Although the deployment scheme based on experience can meet the needs of phased search missions, it has the following two problems:

[0004] 1. Due to the lack of systematic theoretical support, there are shortcomings in scientificity;

[0005] 2. Since the search performance cannot be quantified, it is impossible to quantitatively evaluate the pros and cons of the search solution.

[0006] Therefore, how to establish a mathematical model for team deployment and solve it to obtain a deployment plan is an important theoretical problem that needs to be solved urgently in the field of aerospace search. Summary of the invention

[0007] In view of the above problems, the present invention provides a ground search team deployment algorithm based on genetic algorithm, which ensures the safety of the team, improves the search efficiency of the team, and shortens the time required to complete the task.

[0008] The technical solution adopted by the present invention to solve the above technical problems is: a ground search team deployment algorithm based on genetic algorithm, comprising the steps of:

[0009] S1: Establishing the layout model

[0010] Firstly, the search area and search team are modeled, and then the shortest arrival time is used as an indicator to quantify the effectiveness of the deployment plan, and the deployment plan is solved through genetic algorithm.

[0011] S1.1: Search area modeling

[0012] S1.1.1: Use gridding techniques to divide the search area into a series of regular, sufficiently small rectangular grids;

[0013] S1.1.2: After completing the grid division, the grid needs to be assigned corresponding passable attributes to describe the regional environment. The passable attributes of the grid are indicators that comprehensively reflect the search team's ability to pass within the grid;

[0014] Assume that the maximum speed of the search team is V max , the maximum speed that can be achieved between grid i and grid j is V ij , then the pass factor is defined as γ ij =V ij / V max According to the definition, the traffic factor is a real number between 0 and 1. The closer it is to 1, the higher the traffic capacity is, and the faster the team can move forward in the area. Generally speaking, the trafficability attribute is related to the geographical environment of the grid area, the team attributes and the forward direction. The present invention uses a weighted directed graph method to intuitively describe the traffic factor. To this end, the center point of each grid area is abstracted as a vertex of the graph, and adjacent grids are connected by directed edges. Each directed edge is assigned a weight, which represents the traffic factor of the team from the starting point of the directed edge to the end point of the directed edge. The larger the traffic factor between grids, the better the vehicle traffic capacity is.

[0015] S1.2: Search Team Modeling

[0016] Model the ground search team as a point target with a position and a maximum travel speed. The specific steps are:

[0017] S1.2.1: Obtain the maximum speed that the search team can achieve when traveling between adjacent grids from the traffic factor directed graph;

[0018] S1.2.2: Divide the distance between the center points of the grids by the maximum speed obtained in S1.2.1 to obtain the travel time of the squad between adjacent grids;

[0019] S1.2.3: After obtaining the travel time of the team between each grid, use the Dijkstra algorithm to obtain the optimal path and the shortest time required between the team and the target grid.

[0020] S2: Establishing the layout problem

[0021] To characterize the importance of a grid, each grid needs to be given a weight between 0 and 1. The closer it is to 1, the more important the grid is. The weight is selected through the uniform weight method, the empirical estimation method, or the landing point probability method.

[0022] The present invention describes the optimal deployment problem of ground teams as follows: given the weight of each grid in the returner landing area, how to determine the standby position of each team at the initial time of the mission so that each team is on standby in a safe area and the weighted time to reach the returner landing grid is the shortest, where the time to reach the returner landing grid refers to the minimum time for at least one team to reach the returner landing point;

[0023] Assume that the grid area where the returner may land corresponds to the number set U D ={d1,…,d D}, where d i represents the i-th grid number, D is the total number of grids where the returner may land, and is the grid d i Assign a non-negative weight ω to measure the importance i Without loss of generality, we can assume

[0024]

[0025] The grid number set of the safe area is denoted as Ω. is the grid number set for the i-th team. When there are n search and rescue teams, the set S = (P1,…,P n ) represents the layout scheme, where P i ∈Ω i The standby grid number of the i-th search and rescue team;

[0026] Given a deployment plan S, if the i-th search and rescue team starts from the initial position P i Start and arrive at grid d along the shortest path k The time required is t ik , then at least one team arrives at d k The time is

[0027]

[0028] The weighted time to complete the task corresponding to the deployment plan S is

[0029]

[0030] Therefore, the ground team deployment problem can be described as solving an optimization problem

[0031]

[0032] S3: Solving the layout problem

[0033] When the safe area Ω contains m grids and the number of search and rescue teams is n, and the deployment area of ​​each team is Ω, there are m optional deployment schemes. nIn this paper, multi-resolution grid technology and genetic algorithm are used to solve the potential problem (4). The specific steps are as follows:

[0034] S3.1: Determine the coding method. The grid number set of the safe area is Ω. There are n search and rescue teams. Use the inverse of (3) As the fitness function G(S), assuming that the number of genetic algorithm populations is K, the solution S of the layout problem can be encoded as an n-dimensional vector S = (P1,…,P n ), the vector elements are selected from the safe area grid number set Ω;

[0035] S3.2: Determine the gene crossover operation, for two potential schemes S = (P1, ..., P n ) and S'=(P1',…,P n '), the gene crossover operation is defined as the two new solution vectors (P1, ..., P n / 2 ,P n / 2+1 ',…,P n ') and (P1',…,P n / 2 ',P n / 2+1 ,P n ) process;

[0036] S3.3: Determine the gene mutation operation, the potential scheme S = (P1, ..., P n ) is defined as randomly selecting an element of S and replacing it with a random element from the set Ω-S.

[0037] Preferably, the specific steps of solving the layout problem using the genetic algorithm are as follows:

[0038] 1) Randomly generate 2K solution vectors S of the potential problem k , forming the initial population of the genetic algorithm, and randomly generating an optimal solution S op ;

[0039] 2) Calculate the fitness function G(S) corresponding to each solution vector k ), and calculate the roulette probability function corresponding to each solution vector The solution with the highest fitness function is denoted as S' op , S' op and S op For comparison, if S' op The fitness is higher than S op , then update S op =S' op If not, do not update S op ;

[0040] 3) Use the roulette wheel method to independently and randomly select two solutions from the set of 2K solution vectors of the potential problem. The solution vector Sk The probability of being selected is P k , generate a random number ξ uniformly distributed between [0,1]. If ξ<0.9, the two solution vectors are cross-operated to obtain a new solution. If ξ≥0.9, the two solutions are directly used as the new solution.

[0041] 4) Repeat step 3) K times to obtain 2K populations consisting of new solutions;

[0042] 5) Perform mutation operations on the individuals in the population obtained in step 4) in sequence, and set the mutation probability to 0.005;

[0043] 6) Repeat steps 2) to 5) until the number of repetitions exceeds the set threshold, and then stop. op Output as the optimal solution.

[0044] Preferably, the algorithm of the present invention is accelerated using multi-resolution grid technology, and the specific steps are as follows:

[0045] 1) First, use a large grid to divide the area, and use a genetic algorithm to solve the problem to obtain a preliminary layout plan;

[0046] 2) Then divide the area into finer grids, and restrict the deployment range of each team to the large grid given by the preliminary deployment plan, and further solve the deployment plan corresponding to the fine grid through genetic algorithm;

[0047] Assuming that the side length of a given large grid is k times the side length of a fine grid, each given large grid contains k 2 There are m fine grids in the safe area Ω, and the number of search and rescue teams is n. If the fine grid is directly used for the layout algorithm calculation, the total layout scheme is When multi-resolution grid technology is used, the potential layout schemes corresponding to the large grid are n m There are 2 types of fine grid layout schemes: Therefore, the total solution that needs to be searched is It can be seen that by adopting multi-resolution grid technology, the total layout scheme can be greatly reduced, thereby reducing the amount of calculation and improving the solution efficiency.

[0048] Preferably, the uniform weight method refers to assigning the same weight to each grid, that is, assuming that the probability of the returner landing on each grid is the same; the empirical estimation method is to estimate the landing probability of the returner based on previous search and rescue experience, and set the weight of the grid to the estimated probability; the landing point probability method is to use the Monte Carlo shooting method, according to the returner motion equation, use integral extrapolation to obtain a series of predicted landing points of the returner, obtain the landing point probability of each grid through statistics, and use it as the grid weight.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] 1. The present invention first models the search area, and with the help of geographic information grid technology, a series of regular grids are used to characterize the search area, and attributes such as grid passability, safety and importance are given to the grid. The passability attribute of the grid is an indicator that comprehensively reflects the ability of the search team to pass within the grid; then the search team is modeled, and the team is described using particles with position and velocity attributes, so as to establish a mathematical model of the team's deployment. With the shortest time for the team to arrive at the spacecraft as the goal, and the grid in the safe area as the constraint condition for the team's initial deployment position, the team's deployment problem is converted into a mathematical optimization problem. Finally, the genetic algorithm and multi-resolution grid method are used to solve the deployment plan. The present invention provides a mathematical theoretical basis for the deployment problem of the ground search team, solves the problem of deployment plan calculation by using a genetic algorithm, and improves the solution efficiency by using multi-resolution grid technology, which is conducive to ensuring the safety of the team, improving the search efficiency of the team, and shortening the time required to complete the task. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is the directed graph corresponding to the traffic factors of the four grid points;

[0052] Figure 2 This is a schematic diagram of the multi-resolution grid potential algorithm;

[0053] Figure 3 It is a uniform weight distribution scheme. DETAILED DESCRIPTION

[0054] The following will be combined Figure 1-3 The present invention is described in detail. The exemplary embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0055] A ground search team deployment algorithm based on genetic algorithm, comprising the following steps:

[0056] S1: Establishing the layout model

[0057] The geographical factors of the search area and the attributes of the ground team largely determine the deployment plan. The present invention first models the search area and the search team, then uses the shortest arrival time as an indicator to quantify the effectiveness of the deployment plan, and solves the deployment plan through a genetic algorithm.

[0058] S1.1: Search area modeling

[0059] The geographical factors of the search area directly determine the search team's speed and route, and have a great impact on the deployment plan. Therefore, it is necessary to first model the geographical environment within the search area.

[0060] In order to model the search area, the present invention uses gridding technology, first dividing the search area into a series of regular, sufficiently small rectangular grids. After the grid division is completed, the grids need to be assigned corresponding attributes to describe the regional environment. For the deployment problem, the main regional attribute of concern is passability, and different geographical factors can ultimately be abstracted as the impact on the team's traffic capacity. Using the traffic capacity index to characterize the environment not only fully considers the factors that affect the deployment plan, but also avoids the tediousness of listing various geographical types. The passability attribute of the grid is an indicator that comprehensively reflects the search team's traffic capacity within the grid. Assume that the maximum speed of the search team is V max , the maximum speed that can be achieved between grid i and grid j is V ij , then the pass factor is defined as γ ij =V ij / V max . According to the definition, the traffic factor is a real number between 0 and 1. The closer it is to 1, the higher the traffic capacity is and the faster the team can move forward in the area. Generally speaking, the trafficability attribute is related to the geographical environment of the grid area, the team attributes and the direction of travel. The maximum travel speed that the same vehicle can reach in different terrains is different, and different types of vehicles also have different travel speeds in the same terrain. At the same time, the traffic attribute is also related to the direction of the team's travel. For example, the maximum speed that can be achieved when going uphill is less than the maximum speed when going downhill.

[0061] The present invention uses a weighted directed graph method to intuitively describe the traffic factor. To this end, the center point of each grid area is abstracted as a vertex of the graph, and adjacent grids are connected using directed edges. Each directed edge is assigned a weight, which represents the traffic factor of the team from the starting point of the directed edge to the end point of the directed edge. Figure 1 The figure shows the directed graph corresponding to the traffic factors of the four grid points. The left side of the figure shows the actual grid area. There are roads between grid 4 and grid 3, and between grid 3 and grid 2 (bold lines in the figure). The right side shows the corresponding directed graph. It can be seen from the figure that the traffic factors between grids connected by road networks are larger, which means that the vehicle traffic capacity is better.

[0062] S1.2: Search Team Modeling

[0063] When determining the deployment plan, the team position and travel speed are of primary concern. Among them, the team position and maximum speed mainly affect the planning of the search path and the calculation of the time required to complete the search. For this reason, the present invention models the ground search team as a point target with a position and a maximum travel speed.

[0064] In order to obtain the deployment plan, it is necessary to calculate the completion time corresponding to the deployment plan. The travel path of each team is the basis for calculating the completion time. In order to obtain the time required for the team to reach the target grid from the initial position, first obtain the maximum speed that the search team can reach when traveling between adjacent grids from the traffic factor directed graph, and then divide the distance between the grid center points by the speed to obtain the travel time of the team between adjacent grids. After obtaining the travel time of the team between each grid, the Dijkstra algorithm can be used to obtain the optimal path and the shortest time required between the team and the target grid.

[0065] S2: Establishing the layout problem

[0066] In order to characterize the importance of the grid, each grid needs to be assigned a weight. The weight is a number between 0 and 1, and the closer it is to 1, the more important the grid is. For example, if the information of the returner's aiming point is known, a higher weight can be assigned to the grid where the aiming point is located, indicating that the grid should be given extra attention when solving the deployment plan. The selection of weights is obtained through uniform weight method, empirical estimation method and landing point probability method. Among them, the uniform weight method refers to assigning the same weight to each grid, that is, assuming that the probability of the returner landing on each grid is the same; the empirical estimation method is to estimate the landing probability of the returner based on previous search and rescue experience, and set the weight of the grid to the estimated probability; the landing point probability method uses the Monte Carlo shooting method, according to the returner's motion equation, and uses integral extrapolation to obtain a series of predicted landing points of the returner, and obtains the landing point probability of each grid through statistics, and uses it as the grid weight.

[0067] The present invention describes the optimal deployment problem of ground teams as follows: given the weight of each grid in the returner landing area, how to determine the standby position of each team at the initial moment of the mission so that each team is on standby in a safe area and the weighted time to reach the returner landing grid is the shortest. The time to reach the returner landing grid refers to the minimum time for at least one team to reach the returner landing point.

[0068] Assume that the grid area where the returner may land corresponds to the number set U D ={d1,…,d D}, where d i represents the i-th grid number, and D is the total number of grids where the returner may land. i Assign a non-negative weight ω to measure the importance i Without loss of generality, we can assume

[0069]

[0070] The grid number set of the safe area is recorded as is the grid number set where the i-th team can be deployed. When there are n search and rescue teams, the set S = (P1,…,P n ) represents the layout scheme, where P i ∈Ω i is the standby grid number of the i-th search and rescue team.

[0071] Given a deployment plan S, if the i-th search and rescue team starts from the initial position P i Start and arrive at grid d along the shortest path k The time required is t ik , then at least one team arrives at d k The time is

[0072]

[0073] The weighted time to complete the task corresponding to the deployment plan S is

[0074]

[0075] Therefore, the ground team deployment problem can be described as solving an optimization problem

[0076]

[0077] S3: Solving the layout problem

[0078] When the safe area Ω contains m grids and the number of search and rescue teams is n, and the deployment area of ​​each team is Ω, there are m optional deployment schemes. n Therefore, the amount of calculation required to solve the layout problem described by equation (4) grows exponentially. When the number of grids and teams is large, it is impossible to obtain the optimal layout solution by listing them one by one.

[0079] To overcome the above difficulties, the present invention uses multi-resolution grid technology and genetic algorithm to solve the layout problem (4).

[0080] Genetic algorithm is a widely used heuristic algorithm with the characteristic of being able to approach the global optimal solution. The main idea of ​​genetic algorithm is to simulate the process of continuous reproduction, gene crossover, mutation, and survival of the fittest of the initial population to obtain the offspring of the population suitable for survival, based on the principle of "survival of the fittest" in the inheritance and evolution of organisms in the natural environment. It is an adaptive algorithm that can gradually obtain the optimal solution.

[0081] In order to use genetic algorithm to solve the problem of ground team deployment, we first need to determine the encoding method. The grid number set of the safe area is Ω. There are n search and rescue teams. Use the inverse of (3) As the fitness function G(S). Assuming that the number of genetic algorithm populations is K, the solution S of the layout problem can be encoded as an n-dimensional vector S = (P1,…,P n ), the vector elements are selected from the safe area grid number set Ω. Secondly, the gene crossover operation needs to be determined. For the two layout schemes S = (P1,…,P n ) and S'=(P1',…,P n '), the gene crossover operation is defined as the two new solution vectors (P1, ..., P n / 2 ,P n / 2+1 ',…,P n ') and (P1',…,P n / 2 ',P n / 2+1 ,P n ) process. Finally, it is necessary to determine the gene mutation operation, the potential scheme S = (P1,…,P n ) is defined as randomly selecting an element of S and replacing it with a random element from the set Ω-S. The process of using genetic algorithms to solve the potential problem is as follows:

[0082] 1) Randomly generate 2K solution vectors S of the potential problem k , forming the initial population of the genetic algorithm, and randomly generating an optimal solution S op ;

[0083] 2) Calculate the fitness function G(S) corresponding to each solution vector k ), and calculate the roulette probability function corresponding to each solution vector The solution with the highest fitness function is denoted as S' op , S' op and S op For comparison, if S' op The fitness is higher than S op , then update S op =S' op If not, do not update S op ;

[0084] 3) Use the roulette wheel method to independently and randomly select two solutions from the set of 2K solution vectors of the potential problem. The solution vector S k The probability of being selected is P k , generate a random number ξ uniformly distributed between [0,1]. If ξ<0.9, the two solution vectors are cross-operated to obtain a new solution. If ξ≥0.9, the two solutions are directly used as the new solution.

[0085] 4) Repeat step 3) K times to obtain 2K populations consisting of new solutions;

[0086] 5) Perform mutation operations on the individuals in the population obtained in step 4) in sequence, and set the mutation probability to 0.005;

[0087] 6) Repeat steps 2) to 5) until the number of repetitions exceeds the set threshold, and then stop. op Output as the optimal solution.

[0088] When the search and rescue area is large and the grid is small, it still takes a long time to use the genetic algorithm. In order to further shorten the solution time, the algorithm of the present invention can be accelerated using multi-resolution grid technology. The basic idea of ​​solving the deployment plan with a multi-resolution grid is to first divide the area using a larger grid, obtain a preliminary deployment plan through a genetic algorithm, and then divide the area into finer grids. By limiting the deployment range of each team to the large grid given by the preliminary deployment plan, the deployment plan corresponding to the fine grid is further solved. Figure 2 As shown in the figure, when a larger grid is used for deployment, the standby position of the i-th team is located in a given large grid, that is, the dark grid in the figure. In order to further obtain the accurate deployment position, the dark grid is further divided into several fine grids, and the deployment position of the team is restricted to the grid. The team deployment plan based on fine grids is obtained by solving the genetic algorithm.

[0089] By using multi-resolution grid technology, the solution efficiency can be improved. Assuming that the side length of a given large grid is k times the side length of a fine grid, each given large grid contains k 2 There are m fine grids in the safe area Ω, and the number of search and rescue teams is n. If the fine grid is directly used for the layout algorithm calculation, the total layout scheme is When using the multi-resolution grid method, there are n potential layout schemes corresponding to the large grid. m There are 2 types of fine grid layout schemes: Therefore, the total solution that needs to be searched is It can be seen that by adopting multi-level grid technology, the total layout plan can be greatly reduced, thereby reducing the amount of calculation and improving the solution efficiency.

[0090] In the specific implementation, simulation experiments are used to verify the rationality of the layout algorithm proposed by the algorithm of the present invention. In the simulation experiment, the landing area of ​​the return capsule is a rectangle of 48km×36km, and it is assumed that the grid side length is selected as 3km. Figure 3 As shown, the landing area is divided into 16×12 grid areas, and the thick lines in the area represent roads.

[0091] Assume that eight search teams need to be deployed. Assume that teams 1, 2, and 3 travel at a speed of 75 km / h on the road and 50 km / h in roadless areas. The remaining teams travel at a speed of 60 km / h on the road and 40 km / h when there is no road. To ensure safety, the standby area of ​​each team is required to be located at the edge grid of the landing area. The weights corresponding to each grid are equal. The deployment plan in this case is as follows Figure 2 As shown in the figure, the weighted arrival time corresponding to the scheme is 9.97 minutes, and the grid with the longest arrival time is located in the sixth row and fifth column, corresponding to a time of 21.39 minutes. This deployment scheme deploys the fastest teams 1, 2, and 3 in the area with the most convenient transportation, and the remaining teams are evenly distributed around the entire deployment area, thereby shortening the time required for the teams to arrive at the return capsule.

[0092] The technical solutions provided by the embodiments of the present invention are introduced in detail above. Specific examples are used herein to illustrate the principles and implementation methods of the embodiments of the present invention. The description of the above embodiments is only applicable to help understand the principles of the embodiments of the present invention. At the same time, for those skilled in the art, according to the embodiments of the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A ground search team deployment algorithm based on genetic algorithm, characterized by: Includes steps: S1: Establishing the layout model Firstly, the search area and search team are modeled, and then the shortest arrival time is used as an indicator to quantify the effectiveness of the deployment plan, and the deployment plan is solved through genetic algorithm. S1.1: Search area modeling S1.1.1: Use gridding techniques to divide the search area into a series of regular, sufficiently small rectangular grids; S1.1.2: After completing the grid division, the grid needs to be assigned corresponding passable attributes to describe the regional environment. The passable attributes of the grid are indicators that comprehensively reflect the search team's ability to pass within the grid; Assume that the maximum speed of the search team is V max , the maximum speed that can be achieved between grid i and grid j is V ij , then the pass factor is defined as γ ij =V ij / V max According to the definition, the traffic factor is a real number between 0 and 1. The closer it is to 1, the higher the traffic capacity is and the faster the team can move forward in the area. Generally speaking, the trafficability attribute is related to the geographical environment of the grid area, the team attributes and the forward direction. The weighted directed graph method is used to intuitively describe the traffic factor. To this end, the center point of each grid area is abstracted as a vertex of the graph, and adjacent grids are connected by directed edges. Each directed edge is assigned a weight, which represents the traffic factor of the team from the starting point of the directed edge to the end point of the directed edge. The larger the traffic factor between grids, the better the vehicle traffic capacity. S1.2: Search Team Modeling Model the ground search team as a point target with a position and a maximum travel speed. The specific steps are: S1.2.1: Obtain the maximum speed that the search team can achieve when traveling between adjacent grids from the traffic factor directed graph; S1.2.2: Divide the distance between the center points of the grids by the maximum speed obtained in S1.2.1 to obtain the travel time of the squad between adjacent grids; S1.2.3: After obtaining the travel time of the team between each grid, use the Dijkstra algorithm to obtain the optimal path and the shortest time required between the team and the target grid; S2: Establishing the layout problem To characterize the importance of a grid, each grid needs to be given a weight between 0 and 1. The closer it is to 1, the more important the grid is. The weight is selected through the uniform weight method, the empirical estimation method, or the landing point probability method. The optimal deployment problem of ground teams is described as follows: given the weight of each grid in the returner landing area, how to determine the standby position of each team at the initial moment of the mission so that each team is on standby in a safe area and the weighted time to reach the returner landing grid is the shortest, where the time to reach the returner landing grid refers to the minimum time for at least one team to reach the returner landing point; Assume that the grid area where the returner may land corresponds to the number set U D ={d1,…,d D }, where d i represents the i-th grid number, D is the total number of grids where the returner may land, and is the grid d i Assign a non-negative weight ω to measure the importance i Without loss of generality, we can assume The grid number set of the safe area is denoted as Ω. is the grid number set for the i-th team. When there are n search and rescue teams, the set S = (P1,…,P n ) represents the layout scheme, where P i ∈Ω i The standby grid number of the i-th search and rescue team; Given a deployment plan S, if the i-th search and rescue team starts from the initial position P i Start and arrive at grid d along the shortest path k The time required is t ik , then at least one team arrives at d k The time is The weighted time to complete the task corresponding to the deployment plan S is Therefore, the ground team deployment problem can be described as solving an optimization problem S3: Solving the layout problem When the safe area Ω contains m grids and the number of search and rescue teams is n, and the deployment area of ​​each team is Ω, there are m optional deployment schemes. n In this paper, multi-resolution grid technology and genetic algorithm are used to solve the potential problem (4). The specific steps are as follows: S3.1: Determine the coding method. The grid number set of the safe area is Ω. There are n search and rescue teams. Use the inverse of (3) As the fitness function G(S), assuming that the number of genetic algorithm populations is K, the solution S of the layout problem can be encoded as an n-dimensional vector S = (P1,…,P n ), the vector elements are selected from the safe area grid number set Ω; S3.2: Determine the gene crossover operation, for two potential schemes S = (P1, ..., P n ) and S'=(P1',…,P n '), the gene crossover operation is defined as the two new solution vectors (P1, ..., P n / 2 ,P n / 2+1 ',…,P n ') and (P1',…,P n / 2 ',P n / 2+1 …,P n ) process; S3.3: Determine the gene mutation operation, the potential scheme S = (P1, ..., P n ) is defined as randomly selecting an element of S and replacing it with a random element from the set Ω-S.

2. The genetic algorithm-based ground search team deployment algorithm according to claim 1 is characterized by: The specific steps of solving the layout problem using the genetic algorithm are as follows: 1) Randomly generate 2K solution vectors S of the potential problem k , forming the initial population of the genetic algorithm, and randomly generating an optimal solution S op ; 2) Calculate the fitness function G(S) corresponding to each solution vector k ), and calculate the roulette probability function corresponding to each solution vector The solution with the highest fitness function is denoted as S' op , S' op and S op For comparison, if S' op The fitness is higher than S op , then update S op =S' op If not, do not update S op ; 3) Use the roulette wheel method to independently and randomly select two solutions from the set of 2K solution vectors of the potential problem. The solution vector S k The probability of being selected is P k , generate a random number ξ uniformly distributed between [0,1]. If ξ<0.9, the two solution vectors are cross-operated to obtain a new solution. If ξ≥0.9, the two solutions are directly used as the new solution. 4) Repeat step 3) K times to obtain 2K populations consisting of new solutions; 5) Perform mutation operations on the individuals in the population obtained in step 4) in sequence, and set the mutation probability to 0.005; 6) Repeat steps 2) to 5) until the number of repetitions exceeds the set threshold, and then stop. op Output as the optimal solution.

3. The genetic algorithm-based ground search team deployment algorithm according to claim 2 is characterized in that: Use multi-resolution grid technology to accelerate the algorithm. The specific steps are as follows: 1) First, use a large grid to divide the area, and use a genetic algorithm to solve the problem to obtain a preliminary layout plan; 2) Then divide the area into finer grids, and restrict the deployment range of each team to the large grid given by the preliminary deployment plan, and further solve the deployment plan corresponding to the fine grid through genetic algorithm; Assuming that the side length of a given large grid is k times the side length of a fine grid, each given large grid contains k 2 When there are m fine grids in the safe area Ω and the number of search and rescue teams is n, if the fine grid is directly used for the layout algorithm calculation, the total layout scheme is When multi-resolution grid technology is used, the potential layout schemes corresponding to the large grid are n m There are 2 types of fine grid layout schemes: Therefore, the total solution that needs to be searched is It can be seen that by adopting multi-resolution grid technology, the total layout scheme can be greatly reduced, thereby reducing the amount of calculation and improving the solution efficiency.

4. The ground search team deployment algorithm based on genetic algorithm according to claim 1 is characterized by: The uniform weight method is to assign the same weight to each grid, that is, to assume that the probability of the returner landing on each grid is the same; the empirical estimation method is to estimate the landing probability of the returner based on previous search and rescue experience, and set the weight of the grid to the estimated probability; The landing point probability method uses the Monte Carlo shooting method. According to the returner's motion equation, it uses integral extrapolation to obtain a series of predicted landing points of the returner. The landing point probability of each grid is obtained through statistics and used as the grid weight.

Images