Path planning method for truck and drone collaborative search and rescue in mountainous environments

Abstract

The present invention discloses a path planning method for truck and drone collaborative search and rescue in mountainous environments, and belongs to the field of path planning. By acquiring population and shelter data in the disaster-stricken area, the terrain and road network data of the area, and establishing a shelter site selection model based on the data parameters; using the site selection model to obtain the evacuation results of each population point, using the analytic hierarchy process to determine the rescue priority, and establishing a path planning model for truck and drone collaborative search and rescue, and obtaining a search and rescue plan corresponding to the site selection result. This method can establish a suitable shelter site selection model in a mountainous environment to optimize the site selection results, and at the same time optimize the search and rescue method under the collaboration of the carrier truck and the drone to improve the search and rescue efficiency, and is suitable for truck and drone collaborative search and rescue in mountainous environments.

Description

Technical Field

[0001] The present invention relates to the technical field of path planning, and in particular to a path planning method for coordinated search and rescue by trucks and unmanned aerial vehicles in mountainous environments. Background Art

[0002] Mountainous areas are more susceptible to natural disasters such as earthquakes, landslides, mudslides, and avalanches due to their complex terrain. A comprehensive assessment is required when considering the location of shelters in mountainous areas, combining multiple factors such as geology and altitude to ensure that shelters can play their greatest role when disasters occur. Reasonable shelter location can mitigate disaster risks, improve the safety of shelters, and protect the lives of residents to the greatest extent possible when disasters occur.

[0003] At the same time, with the rapid development of drone technology, the application of drones in disaster search and rescue has gradually become possible. Drones can provide aerial perspectives, high-resolution images and videos, and the ability to quickly cover large areas. The research on truck-drone collaborative search and rescue system comes from the urgent need to improve the efficiency of disaster management and emergency rescue. Optimizing the transportation path of vehicles and coordinating drones can better cope with complex terrain and road conditions, quickly search and locate in the first time after the disaster, and improve the efficiency of search and rescue.

[0004] Therefore, how to use trucks and drones to collaborate in complex mountainous environments and optimize the rescue routes under their collaboration to improve rescue efficiency has become an issue that needs further research. Summary of the invention

[0005] The present invention provides a path planning method for coordinated search and rescue by trucks and drones in mountainous environments, which can establish a suitable shelter site selection model in mountainous environments to optimize site selection results, and at the same time optimize the search and rescue method under the cooperation of carrier trucks and drones to improve the search and rescue efficiency.

[0006] An embodiment of the present invention provides a path planning method for coordinated search and rescue by trucks and drones in a mountainous environment, comprising the following steps:

[0007] Step S1, establishing a shelter site selection model based on the population of the disaster-stricken area, the geographical information of the shelter, and the evacuation safety conditions, and solving the shelter site selection model using a genetic algorithm to obtain all evacuation plans;

[0008] Step S2, according to the evacuation plan, obtain the population of the shelter, obtain the road network safety data, establish a hierarchical analysis model, and determine the rescue priority;

[0009] Step S3, establishing a path planning model for truck-UAV collaborative search and rescue based on the rescue priority, solving the path planning model using an improved genetic algorithm, and obtaining a path planning solution for truck-UAV collaborative search and rescue.

[0010] The path planning method for coordinated search and rescue by trucks and drones in mountainous environments according to the embodiment of the present invention has the following beneficial effects:

[0011] (1) In the past, evacuation in shelter site selection was too idealistic and could not adapt to the complex disaster situation in mountainous areas. This paper adopts a two-stage evacuation, which not only ensures that residents can evacuate safely and quickly, but also provides later guarantees and increases the feasibility of the evacuation plan.

[0012] (2) In view of the lack of application of previous site selection models in mountainous areas, a more realistic shelter allocation scheme is obtained by considering terrain constraints and combining the original distance and capacity constraints. The actual evacuation time considering road traffic is calculated by the travel time function. Compared with the previous SO model, the average evacuation distance and time are improved by 18.57% and 25.49% respectively.

[0013] (3) Different from the previous objective functions of minimizing time or cost, this paper takes the urgency and complexity of post-disaster rescue into consideration, takes obtaining the highest value within a fixed time as the objective function, assigns different importance scores to all shelters to be searched and rescued, and reserves time for search and rescue and information collection for each group of search and rescue missions, which is more in line with the real context.

[0014] (4) In view of the random generation of the initial population of the traditional genetic algorithm, which may lead to low algorithm efficiency, the initial population is generated through a greedy strategy, which ensures the quality of the initial solution and accelerates the algorithm startup. It converges faster than the traditional GA at different search and rescue times, improves the quality of the optimal solution by 7.77%, and shortens the average operation time by 23.28%.

[0015] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0017] Figure 1 A flowchart of a path planning method for collaborative search and rescue by trucks and drones in mountainous environments according to an embodiment of the present invention;

[0018] Figure 2 A flow chart of the shelter site selection model based on two-stage planning;

[0019] Figure 3 A model flow chart for truck-drone collaborative search and rescue based on rescue priority;

[0020] Figure 4 This is a schematic diagram of two-stage evacuation;

[0021] Figure 5 A schematic diagram of the shelter's service radius;

[0022] Figure 6 It is the travel time variation diagram;

[0023] Figure 7 It is a schematic diagram of the two-stage model;

[0024] Figure 8 Calculate 3D distances for the second stage;

[0025] Fig. 9 a is the result of the first stage evacuation, Fig. 9 b is the first stage algorithm iteration graph;

[0026] Fig.10 It is the evacuation distance of three zones;

[0027] Fig.11 The number of people that can be accommodated in shelters and the number of transfer vehicles for the first phase;

[0028] Fig.12 a is the second stage evacuation diagram, Fig.12 b is the second stage algorithm iteration graph;

[0029] Fig.13 a is the second stage shelter evacuation time, Fig.13 b is the maximum evacuation time of the shelter, Fig.13 c is the population accommodated by the second-stage shelters;

[0030] Fig.14 a is the first index η, Fig.14 b is the second index σ, Fig.14 c is the improvement chart of evacuation distance and evacuation time results;

[0031] Fig.15 AHP model for rescue priority;

[0032] Fig.16 This is a schematic diagram of the truck drone path planning model;

[0033] Fig.17 Number the two chromosomes;

[0034] Fig.18 It is the traditional cross method;

[0035] Fig.19 For the improved crossover method;

[0036] Fig. 20 a is the path planning diagram under 24-hour search and rescue time. Fig. 20 b is the 24-hour algorithm iteration graph; Fig. 20 c is the progress of the search and rescue work under 24 hours of time;

[0037] Fig.21 The curves showing the results of different search and rescue times;

[0038] Fig. 22 is the speed change curve of the UAV;

[0039] Fig.23 The curves of the results under different truck speeds are shown below;

[0040] Fig.24 It is the algorithm fitness comparison curve;

[0041] Fig.25 a is the GA result, Fig.25 b is the GSGA result, Fig.25 c is the comparison result between GA and GSGA at different search and rescue times;

[0042] Fig.26 a is 36 hours of GA operation time, Fig.26 b is 36 hours of GSGA operation time, Fig.26 The c is 36 hours of algorithm operation time comparison, Fig.26 d is the comparison result of different search and rescue time algorithms. DETAILED DESCRIPTION

[0043] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0044] The present invention provides a path planning method for truck-UAV collaborative search and rescue in mountainous environments. The method is used for rescue in complex mountainous environments. In this scenario, the search and rescue process is roughly as follows: by determining the population distribution and shelter locations in the disaster-stricken area, an evacuation plan is planned for all disaster victims based on the shelter site selection model of this example. Then, using the truck-UAV as a search and rescue tool, the best rescue plan is determined within a fixed time, and the truck-UAV starts from the starting point together and goes to different shelters respectively, and then converges at a certain shelter to complete information exchange or charging tasks; starting from the shelter, the search and rescue tasks are performed again separately, and so on, until the maximum search and rescue time is reached. The improvement idea is mainly to optimize the shelter site selection model and the path planning algorithm. The specific steps are as follows: Figure 1 , Figure 2 and Figure 3 shown.

[0045] like Figure 1 As shown, the path planning method for collaborative search and rescue by trucks and drones in mountainous environments includes the following steps:

[0046] Step S1, establishing a shelter site selection model based on the population of the disaster-stricken area, the geographical information of the shelter, and the evacuation safety conditions, and solving the shelter site selection model using a genetic algorithm to obtain all evacuation plans;

[0047] Step S2, according to the evacuation plan, obtain the population of the shelter, obtain the road network safety data, establish a hierarchical analysis model, and determine the rescue priority;

[0048] Step S3, establishing a path planning model for collaborative search and rescue by trucks and UAVs based on the rescue priority according to the rescue priority, solving the path planning model using an improved genetic algorithm, and obtaining a path planning solution for collaborative search and rescue by trucks and UAVs.

[0049] In step S3, the improved genetic algorithm is used to solve the path planning model. The genetic algorithm based on the greedy strategy generates a population of drone paths that meet the needs in the local space. Combined with the global search characteristics of the genetic algorithm, an optimized truck path is generated for each drone path.

[0050] In this example, after an earthquake disaster occurs in a mountainous environment, the evacuation of residents not only needs to consider timeliness, but also needs to consider the actual terrain due to the complexity of the mountainous environment. Therefore, the complex evacuation problem is divided into two stages. Figure 4 and Figure 5 shown.

[0051] Designate short-term shelters as first-stage shelters, which provide short-term, emergency shelter during a disaster or emergency. Plan to gradually move people from temporary short-term shelters to long-term shelters over time to ensure a smooth transition. Designate long-term shelters as second-stage shelters, which focus more on providing long-term shelter and support.

[0052] like Figure 6 As shown in FIG, the BPR function (Bureau of Public Roads) represents the relationship between the travel time of a certain road section and the road flow. The travel time only considers the actual flow of the road, but does not consider the flow of adjacent paths.

[0053] t(x)=t 0 [1+α(x / c 0 ) β ] (1)

[0054] Among them, t(x) is the time it takes to travel on a path with a specified flow rate of x, c is the theoretical capacity of the road, and t 0 is the travel time when the road flow is 0, also known as the "free flow travel time". Parameters α>0 and β>0 are adjustment parameters defined according to road characteristics, which are 0.15 and 4 respectively. This function will be used to calculate the actual evacuation time in the second stage.

[0055] Combination Figure 7 and Figure 8 As shown in Figure 2, the objective function of the first-stage model is formula (2), which calculates the evacuation distance of all residential points to minimize the total evacuation distance of residents:

[0056]

[0057] The model constraints are:

[0058]

[0059] Formula (3)d max represents the distance constraint of the model, ensuring that the evacuation distance of each settlement cannot exceed the maximum value; Equations (4) and (5) ensure that all settlements are evacuated and each settlement can only be assigned to an open shelter.

[0060]

[0061] Formula (6) is the capacity constraint of the model, which first ensures that the outflow of the population point is equal to the flow received by the shelter and does not exceed the capacity limit; Formula (7) is to calculate the total capacity of all population points accepted by each shelter. The result is used to count the number of people accommodated and is used in the second stage;

[0062]

[0063] Formula (8-9) is the decision variable that determines whether population point i is assigned shelter j; whether shelter j is open.

[0064] The second stage occurs after the disaster situation stabilizes, and the first stage shelter is transported to the appropriate second stage shelter by vehicles; it is necessary to count the population accommodated by the first stage shelter as shown in formula (7) and calculate the required vehicles. The actual evacuation time is determined by the BPR function. The second stage transfer work is completed by minimizing the maximum accommodation time of the second stage shelter k. Formula (10) is to minimize the maximum accommodation time of the second stage shelter:

[0065]

[0066] t(x)=t 0 [1+α(x / c 0 ) β ](12)

[0067] n truck =[q j / C trcuk ]+1(13)

[0068] Formula (11) calculates the maximum accommodation time of shelter k in the second stage, which means counting the maximum evacuation time of all evacuees to shelter k; Formula (12) calculates the actual travel time of the BPR function; Formula (13) calculates the number of vehicles required to transport disaster victims from shelter j in the first stage. To ensure that all victims are evacuated, the value is rounded and added to one.

[0069]

[0070] Formula (14) ensures that the time for each evacuation to the second-stage shelter does not exceed the maximum evacuation time; (15) and (16) ensure that all first-stage shelters are transferred and the second-stage shelters are open.

[0071]

[0072] Formula (17) ensures that the population flowing out of the first-stage shelter is equal to the population received by the second-stage shelter, and the total population accommodated is not greater than the capacity limit; Formula (18) is the terrain constraint of the model, M is a very large positive number, when x jk When it is equal to 1, it means that the first-stage shelter j is assigned to the second-stage shelter k, and the altitude difference between the two points is ensured not to exceed h 0 , when it is 0, it means that the second-stage shelter k is not selected, M is eliminated and the constraint does not work; Formula (19) is the distance constraint during the shelter transfer.

[0073]

[0074]

[0075] Formulas (20) and (21) are decision variables that determine whether shelter j in the first stage is assigned to shelter k in the second stage; and whether shelter k is open. The parameters of the shelter selection model in this example are shown in Tables 1 and 2:

[0076] Table 1 Variable table of the first stage

[0077]

[0078] Table 2 Variable table of the second stage

[0079]

[0080] This example first determines the distribution of people after the disaster, providing a target for subsequent emergency rescue work. Then, considering the complex operating environment in mountainous areas, drones carrying rescue equipment are used to collect information about the disaster-stricken areas in the first place, grasp the basic situation of the affected people, and collect information from shelter nodes and roads in the post-disaster transportation network.

[0081] The analytic hierarchy process is used to divide the rescue priority. In this example, seven indicators such as shelter population density, road damage, and disaster level index are selected as the basis for classification. The division steps are as follows:

[0082] (1) Problem decomposition: The problem is decomposed into three levels: target level, standard level and sub-standard level. The target is the overall goal of the entire decision-making problem, the standard is the high-level factor that affects the achievement of the target, and the sub-standard is a more specific factor. The rescue priority of the disaster-stricken area is the target level, divided into 4 standard levels and 7 sub-standard levels, as shown in Table 3:

[0083] Table 3 Target criteria for rescue priority

[0084]

[0085] Within each sub-criteria layer, specific evaluation indicators can be listed to measure and evaluate each sub-criteria. For example, under the casualty number sub-criteria, the evaluation indicators can include the number of casualties, the number of serious injuries, the number of critical injuries, etc.

[0086] (2) Establish a judgment matrix: At each level, the decision maker creates a judgment matrix to compare the relative importance of different criteria and sub-criteria, as shown in Table 4. The decision maker uses a comparison matrix to compare each criterion or sub-criteria with other criteria or sub-criteria and assign weights based on their relative importance.

[0087] Table 4 Judgment matrix scale description

[0088] Scale meaning 1 Indicates that two factors are equally important. 3 Indicates that compared with the two factors, the former is slightly more important than the latter 5 Indicates that compared with the two factors, the former is significantly more important than the latter 7 Indicates that compared with two factors, the former is more important than the latter 9 Indicates that compared with the two factors, the former is extremely more important than the latter 2,4,6,8 Indicates the middle value of adjacent judgments reciprocal <![CDATA[If the importance ratio of factor i to factor j is a ij , then the importance ratio of factor j to factor i is a ji >

[0089] (3) Hierarchical single sorting: Hierarchical single sorting means that all elements in this layer are compared with each other for an element in the previous layer, and hierarchical sorting is carried out to arrange the order of importance. The specific calculation can be carried out based on the judgment matrix A. During the calculation, it is ensured that it can meet the requirements. The characteristic root and eigenvector conditions of A. The largest characteristic root of A is λ max ,λ max The normalized eigenvector of

[0090] (4) Calculate weights: Calculate the weights of each criterion and sub-criteria by performing mathematical operations on the judgment matrix. This usually involves calculating eigenvectors and eigenvalues, and then normalizing the weights to ensure that they add up to 1.

[0091] (5) Consistency test: AHP introduces the consistency ratio to test the consistency of the judgment matrix. If the judgment matrix is ​​inconsistent, it needs to be adjusted to ensure that the decision maker's judgment is consistent.

[0092] (6) Construct a comprehensive evaluation: Combine the weight of each criterion and sub-criteria with its contribution to the goal to generate a comprehensive evaluation score for decision making.

[0093] This example defines two networks, namely the truck network and the drone network. The truck network is represented as G t =(V t ′,E t ), where V t ′=V t ∪{0}, is the union of truck nodes and vehicle depots, E t is the set of truck paths. The other network is a drone network, where the flight range of the drone is limited by the battery life T max . Drone Network G d =(V,E d ), the network consists of a complete set of shelter points V = V t ∪V d ∪{0}, and the set of drone paths E d composition.

[0094] According to the rescue priority, different rescue values ​​are assigned to the nodes to be rescued. In a fixed time, truck-UAV collaborative search and rescue is used to build a collaborative search and rescue path planning model based on rescue priority with the goal of maximizing the rescue value.

[0095]

[0096] The above formula is the objective function, which makes the total rescue value divided into two parts: the rescue value of the shelter searched by the truck and the rescue value of the shelter searched by the drone. The model constraints are as follows:

[0097]

[0098] Constraint (23) ensures that the total search and rescue time of the truck-UAV does not exceed the maximum search and rescue time. The former is the search and rescue time of all tasks. In this model, a fixed search and rescue time is reserved for each group of search and rescue tasks; the latter is the time taken for the task to travel. When the UAV and the truck perform search and rescue tasks separately, the time they take to complete the tasks is generally different, so the completion time of each search and rescue task should be the maximum value of the two task times. If the truck arrives at the rendezvous point first, the maximum value is the time taken by the UAV, which is:

[0099]

[0100] Formula (24) represents the time it takes for the UAV to perform the search and rescue mission. If the UAV arrives at the rendezvous point first, the maximum value is the truck time t uw x uw The sum of all maximum mission times is the total search and rescue time, and cannot exceed the total time D.

[0101]

[0102] Constraints (25)-(26) are truck path constraints, which are used to restrict the legality of the truck path. Formula (25) ensures that each node (shelter) is visited at most once while ensuring flow balance; Formula (26) ensures that the truck starts from the starting point and eventually returns to the initial starting point;

[0103]

[0104] Constraints (27)-(29) are the UAV path constraints. Formula (27) is the flow balance constraint of the UAV path, which ensures the legality of the path; Formula (28) ensures that the UAV trip ending at node (u, w) cannot exceed the maximum range of the UAV; Formula (29) ensures that for the UAV trip ending at node (u, w), it must take off at node u and land at node w, and this trip can only be selected once.

[0105]

[0106] Constraint (30) is to ensure the auxiliary constraint of the drone path. This constraint is used to eliminate illegal task nodes separated from the starting point of the flight, ensuring the continuity of each group of flight task nodes of the drone. In the drone network, for a given drone trip starting from node u and ending at node w, the drone node set Vd There are multiple subsets S, If and only if (31)

[0107]

[0108] That is, the UAV can only go to other nodes under the UAV path subset when it ends at the node (u, w).

[0109]

[0110] Constraints (32)-(33) are path node constraints, which transform the UAV path variables with a u and g m Association,ensuring that the starting node of the drone subpath is the same as the truck node.

[0111]

[0112]

[0113] The above constraints are used to ensure the coordination between the UAV and the truck; Equation (34) ensures that the UAV can perform the task on the path (u, w) if and only if the truck path contains the path (u, w); Equation (35) restricts that the UAV can start driving at the same truck node u if and only if the truck path contains the node u.

[0114]

[0115] The above defines the variables of the drone truck, constraint (36) is the path decision variable of the truck and the drone; constraint (37) indicates whether the shelter node is visited. According to the constraints, the parameters of the model are shown in Table 5:

[0116] Table 5 Model variables

[0117]

[0118] In view of the problem that the previous evacuation process was too idealistic and lacked application in mountainous areas, this example improved the site selection model by adding terrain constraints and travel time considerations; at the same time, it optimized the path planning algorithm for truck-drone collaboration and improved the basic theories and key technologies of transportation planning and management. The main advantages are:

[0119] 1. Adding mountain terrain constraints to further increase the feasibility of the evacuation plan, a mountain shelter site selection model based on two-stage planning was established. Combined with the BPR function, the actual evacuation time considering the passage traffic flow was calculated. Compared with previous studies, the average evacuation distance and time increased by 18.57% and 25.49% respectively, making the plan more realistic.

[0120] 2. Different from the previous objective functions of shortest time or lowest cost, this example comprehensively considers the urgency and complexity of post-disaster rescue, takes the highest rescue value as the objective function, divides all shelters to be rescued into different importance levels, and establishes a path planning model for truck-UAV collaborative rescue based on rescue priority. The proposed GSGA converges faster than traditional GA at different search and rescue times, and the quality of the optimal solution is improved by 7.77%, and the average operation time is shortened by 23.28%.

[0121] During post-disaster rescue, the situation may change at any time. By flexibly adjusting the rescue priorities, the search and rescue team can better adapt to disaster situations at different stages and improve their ability to respond to changes.

[0122] In a specific embodiment, the site selection model involves a large number of optional points and multiple constraints. A genetic algorithm is used to solve the model. First, multiple safety constraints need to be considered in the site selection problem. Second, the algorithm encodes the universality of feasible solutions. Finally, the optimal site selection scheme in stages is found. The variables are shown in Table 6.

[0123] Table 6 Algorithm variable table

[0124]

[0125] Using genetic algorithms to solve the two-stage planning problem requires setting different fitness functions and setting new terrain constraints and evacuation time constraints in the second stage, but the remaining steps are basically the same. The specific steps are as follows:

[0126] (1) Chromosome definition and coding

[0127] First, each feasible solution is encoded into a "chromosome", and a list with a length equal to the number of shelters is generated as an individual. Each element in the candidate shelter set is assigned a value and the initial population is generated randomly.

[0128] (2) Definition of fitness

[0129] Since the data points in this example are longitude and latitude data, the Euclidean distance cannot be used to calculate the distance between two sample points. Therefore, the Haversine formula is considered to calculate the distance between sample points.

[0130]

[0131] Where: d ij represents the horizontal distance between point i and point j, r represents the radius of the earth 6371 kilometers, lon i ,lat i Respectively represent the longitude and latitude of point i, lon j ,latj Represent the longitude and latitude of point j respectively. Since the altitude is not considered in the first stage, this distance is the evacuation distance, and the distance calculation method is inaccurate. Therefore, the evacuation distance and evacuation time are calculated by the following method.

[0132]

[0133] The smaller the distance or time corresponding to the chromosome, the higher the fitness. The final fitness score is the sum of the values ​​calculated for all settlements, and the fitness is defined as:

[0134]

[0135] (3) Selection, crossover, and mutation

[0136] ① Selection: Through random selection, the probability of each chromosome being selected is p s Equal. The parent individuals and the offspring individuals are combined to form a larger population, and then individuals with higher fitness are selected from them to form the next generation population. ② Crossover: Define a crossover probability p c , represents the probability of crossover for each pair of parent individuals. If the generated random number is less than the crossover probability, the crossover operation is performed; otherwise, the parent individuals are directly passed to the next generation without crossover. ③ Mutation: Determine that if the probability of each individual is less than the set mutation probability p m , a mutation will occur; it is to change 0 to 1, or 1 to 0, ensuring that the total number of genes with a gene value of 1 in the chromosome remains unchanged before and after the mutation.

[0137] (4) Termination condition judgment

[0138] Determine whether the number of iterations meets the maximum number of iterations N. If i ≥ N, the algorithm stops, and if i < N, it returns to recalculate the fitness function. Finally, the individual with the best fitness and its fitness are returned, that is, the individual with the best fitness among all individuals in the entire iteration process and its fitness, which is the best evacuation plan.

[0139] The path planning model is a variant of TSP. This problem is a combinatorial optimization problem, and its complexity grows exponentially with the increase of the problem scale. The genetic algorithm can find potential solutions in a large-scale search space by performing random search and optimization in the search space, and it is not easy to fall into the local optimal solution. The variables are shown in Table 7.

[0140] Table 7 Algorithm variable table

[0141]

[0142] Therefore, a genetic algorithm based on greedy strategy (GSGA) is designed. The core idea of ​​the algorithm is to select the local best solution at each step without considering the future situation, in the hope of eventually getting the global best solution. The algorithm steps are as follows:

[0143] Step 1: Create an initial population

[0144] In order to improve the efficiency of the algorithm, a greedy strategy is used to generate the initial population, and the shelter number is used instead of the binary code. Since the path of this problem consists of drone paths and truck paths, the initial path generation process is described in two parts.

[0145] (1) UAV path generation part:

[0146] Step 1: Select the initial node set, which is defined as the edge points in the road network, to determine the start and end nodes of the search and rescue mission, and start generating the first set of drone search and rescue paths from this point set as the current node.

[0147] Step 2: When selecting the next mission node, based on the greedy strategy, under the premise of satisfying the constraints of UAV range, energy consumption, etc., the node with the highest rescue value is selected as the next possible mission point, and so on.

[0148] Step 3: When the length of the first set of task paths is greater than d min When , determine the search and rescue time t of this group of tasks rescue and the flight time t mn Does the sum exceed the maximum energy consumption limit of the drone? If it does, fly back to the end point directly, otherwise return to step 2 and continue to find the next node.

[0149] Step 4: Generate truck paths under the group of drone paths as shown in (2), and calculate the fitness function F of the group.

[0150] Step 5: Get a complete individual (an individual is a list, including the drone path, the total value of the drone search and rescue, the truck path, the total time, and the total value). The pseudo code for generating the drone path using the greedy strategy is shown in Table 8.

[0151] Table 8 Greedy strategy generates pseudo code for drone paths

[0152]

[0153]

[0154] (2) Truck route generation part:

[0155] Based on the generated drone path, in order to maximize the model's objective function (4-1) and reduce the waiting time on the road, the drone needs to travel as many nodes and as far as possible at a time. The specific steps are as follows:

[0156] Step 1: While ensuring that the starting point of the truck is the same as the starting point of the drone, check whether the four nodes behind the current node of the drone (n max ) (if it is the end point, go directly to) as the confluence node. If the node is within the energy consumption range, jump to step 5. If it cannot be reached, abandon the proposed fourth node and search for other nodes.

[0157] Step 2: Check whether the three nodes behind the current node of the drone (if it is the end point, go directly to it) can be used as the confluence node. If the node is within the energy consumption range, jump to step 5. If it cannot be reached, abandon the proposed third node and search for other nodes.

[0158] Step 3: Check whether the second node behind the current node of the drone (if it is the end point, go directly to it) can be used as the convergence node. If the node is within the energy consumption range, jump to step 5. If it cannot be reached, abandon the proposed second node and search for other nodes.

[0159] Step 4: Check whether the node behind the current node of the drone is the end point. If it is not the end point, this point will be used as the convergence node and continue to step 5. The search and rescue mission is not over. If it is the end point, the truck and the drone will go directly to the end point to converge and the mission will be over.

[0160] Step 5: When the truck is on its way from the current node to the next confluence node, if it can reach the new node before the drone arrives at the confluence node, the truck can go to other unfinished task nodes by itself during this period; otherwise, the truck will go directly to the confluence node to complete the charging and information recovery of the drone. The pseudo code for truck path generation is shown in Table 9.

[0161] Table 9 Truck route generation pseudo code

[0162]

[0163]

[0164] Step 2: Fitness calculation

[0165] In order to evaluate the quality of each chromosome, it is necessary to define the fitness function F. The fitness of the algorithm consists of two parts. The fitness of the chromosome is the rescue value obtained by each group of tasks. The more value the task obtains, the higher the fitness and the better the individual. The fitness calculation formula of the algorithm is:

[0166]

[0167] Step 3: Select elite individuals

[0168] According to the fitness value of the individual, a part of the individuals are selected as parents. The roulette wheel selection method is adopted, and the individual with higher fitness value has a greater probability of being selected.

[0169]

[0170] The selection operator of this algorithm selects the path planning results with higher fitness from all search and rescue schemes, eliminates those with lower fitness, and then performs the next step of crossover and mutation operations.

[0171] Step 4: Cross

[0172] The crossover of traditional genetic algorithms is to select a pair of individuals from the parent generation, cross their genes by changing the gene coding, and generate new offspring. If this crossover operation is used in the coding method of this algorithm, it may cause duplicate path nodes in the newly generated offspring, affecting the algorithm iteration. For this reason, the crossover rules need to be redesigned.

[0173] In order to ensure that a node is only passed once, a crossover operation needs to be performed on the same chromosome. Because the generation process of a chromosome must satisfy the constraints, after the node position of the chromosome is changed, the constraints must also be satisfied. After the exchange, the truck path is calculated based on the new drone path and returned.

[0174] Step 5: Mutation

[0175] Perform adaptive mutation operations on the newly generated offspring, and dynamically adjust the mutation probability according to the fitness value of the individual. Individuals with higher fitness may require a smaller probability, while individuals with lower fitness may require a larger probability.

[0176]

[0177] Guarantee 0<k 1 <k 2 <1, after mutation, check whether the newly generated drone path is legal. If legal, calculate the truck path based on the new drone path; otherwise, recalculate. Mutation helps increase the diversity of the population and prevents falling into a local optimal solution.

[0178] Step 6: Termination Condition

[0179] The maximum number of iterations is set to determine whether to end the loop. If the termination condition is met, the algorithm ends, otherwise it returns 2. The specific parameters are shown in Tables 10 and 11.

[0180] Table 10 Geographic information data of Yibin City

[0181] Serial number name longitude latitude Population (thousands of people) Altitude(m) 1 Xijiao Subdistrict, Cuiping District 104.599213 28.749523 91 339 2 Anbu Subdistrict, Cuiping District 104.609593 28.784182 92 319 ... 137 Qilin Miao Township, Xingwen County 105.191391 28.221815 21 379 138 Xianfeng Miao Township, Xingwen County 105.052788 28.23101 9 1349

[0182] Table 11 Model algorithm parameter value table

[0183] parameter Value parameter Value Initial population size 50 Selection probability 0.2 Maximum number of iterations 200 Crossover probability 0.2 Mutation probability 0.6 Vehicle speed 40(km / h) Maximum evacuation distance 10(km) Maximum evacuation time 1h Theoretical capacity 2200(vehicles / h) Allowable altitude difference 200(m) Vehicle capacity 80(people)

[0184] (1) Results of the first phase of evacuation

[0185] according to Figure 9-Figure 26 In the first phase, 81 emergency shelters were selected based on the principles of site selection and safety, and 138 residential assembly points were divided for the three districts and eight counties of Yibin City. The sum of the evacuation distances of each residential point was 474.0382 km, and the average evacuation distance per person was 3.4350 km. A total of 3.87 million people were evacuated, and the evacuation distance of all people was 1.3293×10 7 km.

[0186] Taking the three districts of Yibin City (Cuiping, Nanxi, and Xuzhou) as an example, the specific allocation results of each shelter under the district and county, the evacuation distance, the number of settlements and the total population accommodated by each shelter are analyzed to provide a basis for the second stage of transfer work. The evacuation results of some areas are shown in Table 12:

[0187] Table 12 Statistics of evacuation results in some areas

[0188] Cuiping District Nanxi District Xuzhou District All districts total Number of population points 20 11 17 138 Number of shelters 11 9 8 81 Average distance (km) 2.17 3.19 4.61 3.43 Maximum number of population points that can be accommodated in a single shelter 3 4 3 5 Minimum number of population points that can be accommodated in a single shelter 1 1 1 0 Maximum evacuation distance (km) 6.83 6.33 7.66 8.39 Minimum evacuation distance (km) 0.51 0.45 1.88 1.24

[0189] (2) Results of the second phase of evacuation

[0190] The final total evacuation time is 21.4074h, which is the maximum evacuation time of the second stage shelter. Emergency rescue work is carried out on the residents in the shelter. The number of people accommodated in each shelter (second stage) is obtained through the evacuation results, and the result will be used to determine the rescue priority.

[0191] In the two-stage shelter site selection model of this example, constraints are used to ensure that the altitude difference between two points during the evacuation process does not exceed a certain value, ensuring the rationality of the selection of shelters and evacuation routes, so that the site selection results can better simulate the real environment. The advantages and disadvantages of this model compared with previous studies are shown in Table 13 below.

[0192] Table 13 Comparison of advantages and disadvantages of site selection models

[0193] This model Previous Model Evacuation methods Two-stage evacuation Two-stage evacuation Objective Function Shortest evacuation distance, minimize the maximum evacuation time of each shelter Minimum evacuation distance Model constraints Distance constraint; capacity constraint; terrain constraint; evacuation time constraint Distance constraint; Capacity constraint Special features Considering the impact of road traffic on evacuation time ——

[0194] Based on the evacuation plan results, take Shelter 1-20 as an example. Use the Amap API and the final evacuation plan to obtain the actual evacuation distance and time. Analyze the difference between the distance solved by the previous model and the distance solved by this example model and the actual distance to prove the effectiveness of the model.

[0195] (1) Variable description, as shown in Table 14.

[0196] Table 14 Evaluation index variable table

[0197]

[0198]

[0199] (2) Evaluation indicators

[0200] The first indicator η is the ratio of the evacuee’s evacuation distance to the actual evacuation route distance in formula (45). This indicator measures the effectiveness of the model in this example.

[0201]

[0202] The second indicator σ: Formula (46) is the ratio of the evacuee’s evacuation time to the actual evacuation time. This indicator will measure the effectiveness of this example model. Formula (47) is the evacuation time calculated by the second stage model; Formula (48) is the evacuation time in the real environment, obtained from the Amap API v truck Is a fixed value.

[0203]

[0204] (3) Comparison results

[0205] In the second stage, the model can theoretically obtain more realistic results by obtaining three-dimensional distance and BPR function. The evacuation results of the previous shelter model, the site selection model of this example, and the real value obtained by Amap API are compared, and the correctness of the theory is proved by the evaluation indicators σ and η.

[0206] By considering the three-dimensional distance and terrain constraints, the evacuation process has a significant impact; the evacuation distance is improved by 0.83%-51.72% compared with the original study; the evacuation time is improved by 8.33%-56.36%. In the first 20 groups of shelters, the average distance is increased by 18.57%, and the average time is close to 25.49%. Moreover, the farther the distance between the two points, the more accurate the results of the model are than the previous models, proving the rationality of the example model. The model optimization results are shown in Table 15.

[0207] Table 15 Model optimization results

[0208] Serial number Distance improvement (%) Time improvement (%) 1 0.83 8.33 2 12.24 19.04 3 25.54 46.15 … 19 2.37 20.06 20 21.55 30.30

[0209] Carry out post-disaster emergency rescue and search and rescue work. The location of the shelter node is known and determined, and the post-disaster search and rescue work is now carried out by drones equipped with search and rescue equipment.

[0210] Determination of the priority of target areas. In the post-disaster search and rescue work, the population density of different areas results in different numbers of refugees accommodated in shelters. At the same time, the closer to the disaster, the more serious the damage may be, and the more priority rescue is needed. The shelters are divided into five levels through the hierarchical analysis method. The significance of each level is shown in Table 16:

[0211] Table 16 Meaning of rescue levels

[0212]

[0213]

[0214] The rescue level represents the rescue value of the shelter. The truck-drone can obtain the rescue value by continuously searching the shelter. The rescue plan with the highest rescue value. According to the post-disaster rescue documents, the rescue rules under multiple indicators are determined, namely, the population density of the shelter σ 1 , the distance from the disaster center is σ 2 , disaster level σ 3 , the terrain conditions of the shelter σ 4 And the surrounding traffic conditions 5 These five indicators. Among them, population density σ 1 The calculation rules are as follows:

[0215]

[0216] Distance from the disaster center σ 2 The calculation rules are:

[0217]

[0218] The earthquake situation in the same area is the same, so the disaster level σ 3 Taking the same value, the terrain situation of the shelter σ 4 is the altitude of the point, and the surrounding traffic conditions σ are already in the previous table 5 Damage to nearby roads:

[0219]

[0220] Two of the more important indicators are (distance from the disaster center, disaster level). In this example, these two indicators are given higher priority to ensure that after an earthquake disaster, priority is given to rescuing areas that may be more severely damaged. When using trucks and drones for search and rescue work, the priority of rescue will be determined based on the above five indicators. The specific information is shown in Table 17:

[0221] Table 17 Data of each shelter hierarchical analysis criteria layer

[0222]

[0223] The rescue priorities of various shelters determined by the hierarchical analysis method are shown in Table 18:

[0224] Table 18 Rescue priorities for each shelter

[0225]

[0226] The rescue priority determined by the hierarchical analysis method is converted into the search and rescue value of the nodes to be searched and rescued, the importance of rescue in each shelter is determined, and the truck-drone collaborative system is used to carry out search and rescue work in the target area.

[0227] Truck-UAV collaborative search and rescue results:

[0228] The total search and rescue time is set to 24 hours. At the same time, to ensure that complete rescue information can be obtained, an additional 0.4 hours of search and rescue time is allocated to each group of drone missions, which is used to detect the target shelter in detail. Finally, the model is solved by GSGA, and some parameter settings are shown in Table 19:

[0229] Table 19 Algorithm parameter values

[0230] Model parameters Parameter Value Algorithm parameters Parameter Value Drone speed 70(km / h) Population size 50 Truck speed 25(km / h) Maximum number of iterations 500 Total search and rescue time 24h Mission search and rescue time 0.4h Visit Shelter Cap 4 Maximum range of drone 80km Crossover probability 0.6

[0231] In the total search and rescue time of 24 hours, the truck-drone collaborative search and rescue system completed 16 groups of tasks in the target area, and completed a total of 151 rescue-value tasks; among them, the drone alone searched and rescued 103 shelters worth of shelters; the truck alone searched and rescued 14 shelters worth of shelters; and the two searched and rescued 34 shelters worth of shelters together. The drone visited 62 shelters and the truck visited 23 shelters, and the actual time consumed was 23.95 hours, as shown in Table 20.

[0232] Table 20 Coordinated search and rescue mission table (24 hours)

[0233] Task Force Time for each task (s) Drone routes Truck Route Mission rescue value 1 3897.37 28-46-49-19-38 28-38 9 2 2871.82 38-58-15-56-54 38-54 8 3 3730.20 54-0-29-57-52 54-52 7 4 4045.96 52-35-6-75-26 52-42-26 15 5 3998.74 26-69-76-67-1 26-1 7 6 4125.23 1-21-4-70-53 1-53 11 7 4345.56 53-41-31-59-47 53-47 14 8 3787.88 47-11-33-22-3 47-61-3 13 9 3778.31 3-43-14-20-73 3-73 10 10 4004.89 73-18-32-17-16 73-72-16 11 11 4056.65 16-74-65-9-71 16-7-71 14 12 2649.27 71-77-44-30-78 71-24-78 6 13 3288.97 78-25-23-36-63 78-39-63 11 14 3226.39 63-62-13-50 63-50 7 15 4429.54 50-66-27-68-10 50-10 6 16 6953.88 10-8-28 10-28 1

[0234] The parameters initially set in this example are based on experience and may have low reference value. For this reason, some parameters in the model are adjusted to provide more detailed guidance for post-disaster emergency rescue strategies. The adjusted parameters are:

[0235] (1) Total search and rescue time

[0236] All search and rescue tasks cannot be completed within the original 24h time, so the total search and rescue time is changed to solve the UAV-truck search and rescue results under different search and rescue times. The rescue value obtained by the UAV visiting the node alone, the rescue value obtained by the truck visiting the node alone, and the value obtained by the two visiting the node together are calculated. The results are shown in Table 21:

[0237] Table 21 Search and rescue time sensitivity results

[0238]

[0239] With other parameters fixed, as the total search and rescue time increases, the rescue value increases, that is, more targets can be rescued, although the marginal value decreases. Therefore, in practice, rescue organizations should weigh the increased search and rescue value and the total search and rescue time delay, and determine an optimal search and rescue time when the post-disaster time is limited.

[0240] (2) Drone speed

[0241] In addition to the total search and rescue time, the speed of the drone and the truck is more influential and easy to change. These two indicators affect the search and rescue time of each group of tasks. By adjusting different values, the model results under different speeds are compared. When the truck speed is fixed, the drone speed is gradually increased from 40 km / h to 100 km / h. The results are shown in Table 22:

[0242] Table 22 UAV speed sensitivity results

[0243]

[0244] As the speed of drones decreases, their rescue value and number of rescues also decrease, and they are getting closer to the result of trucks, and the total rescue value is also shrinking; and as the speed of drones increases, the number of shelters that trucks can rescue gradually decreases, which means that trucks will only be able to undertake auxiliary tasks, and most of the search and rescue tasks will be completed by drones. However, even if the speed of drones is increased to 100 kilometers per hour, they still cannot complete all the search and rescue work.

[0245] When the speed of the drone increases from 50 to 80 km / h, the rescue value changes rapidly; but after 80-100 km / h, the growth slows down and the marginal effect weakens. There are two reasons for this: first, the speed of the drone increases, but the total energy time is limited, and because each group of tasks requires additional search and rescue time, the search and rescue range of the drone is limited and it cannot rescue shelters farther away; second, it is limited by the speed of the truck, resulting in imperfect coordination between the two.

[0246] (3) Truck speed

[0247] The speed of the drone will greatly affect the search and rescue results of the model. In addition, densely populated nodes and severely damaged nodes should be given priority to ensure that more mission targets can be rescued in a short period of time. Then fix the drone speed to 70 km / h and adjust the truck speed from 20-60 km / h. The results are shown in Table 23:

[0248] Table 23 Truck speed sensitivity results

[0249]

[0250] When the truck speed is low, fewer shelters can be rescued. The increase in speed between 20-30 km / h has a greater impact on the rescue results. Above 30 km / h, there is a certain bottleneck, and the rescue results do not change with the increase in speed. However, as the speed increases further, the truck will be able to undertake more and more search and rescue missions and complete all search and rescue missions within 24 hours.

[0251] The results show that when the speed is 20-30 km / h, the impact is greater. The increase in truck speed can directly rescue more shelters, and the total rescue value is greatly improved; the change in truck speed at 30-45 km / h has little effect on the total search and rescue results, but the truck will be able to undertake more rescue tasks; when the truck speed is 45-60 km / h, the total rescue value is further increased, and the traversal rescue task can be completed in 24 hours. This is because the speed of the truck directly affects the rescue efficiency, and the rescue value it undertakes is increasing, while the UAV is affected by the range and energy consumption, and the rescue value it can undertake has a certain upper limit. The sensitivity analysis of the truck speed shows that:

[0252] (1) Both have a greater impact on the results at lower speeds, and the rescue efficiency is lower.

[0253] (2) The increase in the speed of the UAV has a more significant impact on the rescue results within a certain range; the increase in the speed of the truck has less significant impact than the UAV within a certain range.

[0254] (3) As the speed increases further, the marginal benefit of the drone decreases and the speed effect will no longer be significant; when the truck speed is above 45 km / h, the impact on the results is again greater.

[0255] From the sensitivity analysis of the above three parameters, it can be seen that in the context of post-disaster rescue, rescue factors need to be considered from multiple aspects; it is necessary to rescue more shelters without affecting the overall search and rescue time, which requires specific analysis based on actual conditions.

[0256] In S5, the algorithm of this example is compared with previous algorithms in two aspects:

[0257] (1) Convergence speed and optimal solution quality

[0258] Genetic algorithms are often used to solve traditional TSP, while the objective function of this example model is to rescue more people within a fixed time; traditional GA solutions often fall into local optimality and have low computational efficiency. To prove the effectiveness of the greedy strategy-based genetic algorithm (GSGA), the comparison results of the two algorithms under different search and rescue times are shown in Table 24:

[0259] Table 24 Results of different search and rescue time for two algorithms

[0260] Search and rescue time GA Iterations GSGA Iterations promote 24 131 396 151 213 15.27％ 25 137 279 153 227 11.68％ 26 137 267 158 254 15.33％ 27 142 358 160 179 12.68％ 28 143 412 160 168 11.89％ 29 152 517 162 179 6.58％ 30 154 430 162 153 5.19％ 31 160 372 166 196 3.75％ 32 160 331 169 231 5.62％ 33 160 303 169 167 5.62％ 34 162 391 169 137 4.32％ 35 164 354 169 135 3.05％ 36 169 521 169 103 0％

[0261] The model solution space of this example is large and complex. GSGA can more reliably approach the optimal solution in the calculation and reduce the risk of falling into the local optimal solution. As can be seen from the above table, when the total search and rescue time is 24-36 hours, GSGA is better than the optimal solution of traditional GA, and the final result is 3.05%-15.33% higher, and the optimal solution is improved by 7.77% on average.

[0262] As the total search and rescue time increases, GSGA often requires fewer iterations. This is because the greedy strategy generates high-quality initial solutions, which not only improves the quality of the optimal solution and is not prone to falling into the local optimal solution, but also speeds up the convergence of the algorithm. This shows that the particularity of this problem makes the local optimal solution of each step have a significant impact on the global optimal solution, and these local optimal solutions can quickly approach the global optimal solution when the problem scale is reduced, so GSGA will accelerate convergence.

[0263] (2) Comparison of operation time

[0264] The algorithm's operation time is a key performance indicator. Faster convergence does not necessarily mean shorter operation time. By comparing the results of different search and rescue times between the GSGA and GA algorithms, we can compare the execution speed under different input scales. The total search and rescue time was increased from 24 to 36 hours, and the results under 1000 iterations were compared, as shown in Table 25:

[0265] Table 25 Comparison of GA and GSGA operation time

[0266]

[0267]

[0268] The operation time of GA is basically stable at around 3300-4200s under different search and rescue time, and is not greatly affected by D; while the operation time of GSGA is basically the same in 24-hour rain. As the search and rescue time increases, the operation speed gradually decreases, 24 hours requires 3319.57s, and the operation time of 36 hours is shortened by 37.11%; the performance of GSGA will become more excellent as time increases.

[0269] As time goes by, the difference between the two algorithms in terms of operation time becomes more and more obvious, with an average time reduction of 23.28%. Compared with GA under the same time, the operation time of GSGA is reduced by 2.10-44.94%. Under 24 hours, the operation time of the two algorithms is basically the same. But under 36 hours, almost half of the operation time is saved. This also shows that as the scale of the problem increases, GSGA has a better operation efficiency.

[0270] This example is based on post-disaster emergency rescue. In response to the problem that the evacuation in previous shelter site selection was too idealistic and could not adapt to the complex disaster situation in mountainous areas, a two-stage evacuation was adopted to ensure that residents could evacuate safely and quickly, and to provide subsequent guarantees, increasing the feasibility of the evacuation plan.

[0271] At the same time, in the context of this example, with the goal of solving the search and rescue efficiency of truck-UAV collaboration, the shelter site selection problem and the truck-UAV collaboration path planning problem were solved respectively. Specifically, in view of the problem of idealization of the previous site selection model and the problem of low search and rescue efficiency in collaborative search and rescue, a shelter site selection model based on two-stage planning and a truck-UAV collaboration path planning model were established. In order to comprehensively consider the urgency and complexity of post-disaster rescue, the objective function was to obtain the highest value within a fixed time, and different importance scores were divided for all shelters to be searched and rescued. At the same time, search and rescue and information collection time were reserved for each group of search and rescue tasks, which was more in line with the reality. The solution algorithm was optimized. The example analysis of Yibin, Sichuan showed that compared with the random generation of the initial population of the traditional genetic algorithm, which may lead to problems such as slow algorithm efficiency, the initial population was generated by the greedy strategy, which ensured the quality of the initial trial solution and accelerated the algorithm startup. It converged faster than the traditional GA at different search and rescue times, and the quality of the optimal solution was improved by 7.77%, and the average operation time was shortened by 23.28%.

[0272] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

Claims

1. A path planning method for coordinated search and rescue by trucks and drones in mountainous environments, characterized in that: The following steps are involved: Step S1, establishing a shelter site selection model based on the population of the disaster-stricken area, the geographical information of the shelter, and the evacuation safety conditions, and solving the shelter site selection model using a genetic algorithm to obtain all evacuation plans; Step S2, according to the evacuation plan, obtain the population of the shelter, obtain the road network safety data, establish a hierarchical analysis model, and determine the rescue priority; Step S3, establishing a path planning model for truck-UAV collaborative search and rescue based on the rescue priority according to the rescue priority, solving the path planning model using an improved genetic algorithm, and obtaining a path planning scheme for truck-UAV collaborative search and rescue; In step S1, the objective function of the shelter site selection model calculates the evacuation distances of all settlements to minimize the total evacuation distance of residents: Among them, (I, J) are the population points and the first stage shelter set, q i is the population of settlement i, d ij is the evacuation distance from settlement i to first-stage shelter j, x ij Whether settlement i is assigned to shelter j in the first stage, with a value of 0 or 1; The constraints of the shelter site selection model are: Among them, d max is the maximum evacuation value, V is the set of all node locations, y j is whether shelter j is selected in the first stage, and takes the value of 0 or 1. Formula (3) represents the distance constraint of the shelter selection model, indicating that the evacuation distance of each settlement cannot exceed the maximum evacuation value; Formulas (4) and (5) ensure that all settlements are evacuated, and each settlement can only be assigned to an open shelter; Among them, C j is the upper limit of the capacity of shelter j in the first stage, q j is the population of shelter j, and formula (6) is the capacity constraint of the shelter location model. The constraint of formula (6) first ensures that the outflow of the population point is equal to the flow received by the shelter and does not exceed the capacity limit; formula (7) is to calculate the total capacity of all population points accepted by each shelter. The result of formula (7) is used to count the number of people accommodated and is used in the second stage: Formula (8) and Formula (9) are decision variables. Formula (8) determines whether settlement i is assigned shelter j; Formula (9) determines whether shelter j is open; In step S2, the population accommodated in the first stage shelter is counted and the required vehicles are calculated, and the actual evacuation time is determined by the BPR function. The second stage transfer work is completed by minimizing the maximum accommodation time of the second stage shelter k. The objective function of the hierarchical analysis model is: t(x)=t0[1+α(x / c0) β ] (12) n truck =[q j / C trcuk ]+1 (13) Among them, K is the set of shelters in the second stage, t jk is the maximum evacuation time of the second stage shelter, C trcuk is the vehicle capacity, n truck is the number of vehicles, t0 is the travel time when the road flow is 0, parameters α>0 and β>0 are adjustment parameters defined according to road characteristics, and c0 is the theoretical road capacity; Formula (11) calculates the maximum accommodation time of shelter k in the second stage, which means the maximum evacuation time of all evacuees to shelter k; Formula (12) calculates the actual road travel time of the BPR function; Formula (13) calculates the number of vehicles required for shelter j to transport disaster victims in the first stage, and rounds the value to ensure that all vehicles are evacuated and adds one; The constraints of the hierarchical analysis model are: Among them, T e-max is the maximum evacuation time of the second-stage shelter. Formula (14) ensures that the time for each evacuation to the second-stage shelter does not exceed the maximum evacuation time. Formulas (15) and (16) ensure that all first-stage shelters are transferred and the second-stage shelters are open. Among them, q j is the population of shelter j, q k is the population of shelter k, Q k is the population capacity of the kth shelter, h k is the altitude of shelter k, h j is the altitude of shelter j, h0 is the upper limit of the elevation difference between the two points, d jk is the evacuation distance from the first stage shelter j to the second stage shelter k, d max is the maximum evacuation distance, formula (17) ensures that the population flowing out of the first-stage shelter is equal to the population received by the second-stage shelter, and the total population accommodated is not greater than the capacity limit; formula (18) is the terrain constraint of the model, M is a very large positive number, when x jk When it is equal to 1, it means that the first-stage shelter j is assigned to the second-stage shelter k, and it is ensured that the altitude difference between the two points does not exceed h0. When it is 0, it means that the second-stage shelter k is not selected, and M is eliminated and the constraint does not work. Formula (19) is the distance constraint during the shelter transfer period; Formula (20) and Formula (21) are decision variables. Formula (20) determines whether shelter j in the first stage is assigned to shelter k in the second stage, and Formula (21) determines whether shelter k is open. In step S3, the rescue priority is converted into the rescue value of the node to be rescued, the importance of each shelter rescue is determined, and the truck-UAV collaborative search and rescue is used within a fixed time. The path planning model of collaborative search and rescue based on rescue priority is constructed with the maximum rescue value as the goal. The objective function is: Among them, e u is the rescue node value u∈V of the truck node t , o m The value of the node rescued by the drone node m∈V d , g m is a binary variable, and the drone node m∈V d Whether it has been visited, if it has been visited, it is equal to 1, otherwise it is 0; a u Represents a truck node u∈V t Whether it has been accessed, if it has been accessed, it is equal to 1, otherwise it is 0, V t is the set of shelter nodes that the truck searches for; V d The set of shelter nodes for drone search and rescue; Formula (22) indicates that the total rescue value of the shelter rescue value of truck search and rescue and the shelter rescue value of drone search and rescue is the highest; Constraints include: Where N is the number of search and rescue groups, T rescue is the search and rescue mission time for each group, t uw For a truck on the path (u, w)∈E t The driving time, x uw If the truck is on the path (u, w)∈E t If the vehicle is moving, it is 1; otherwise, it is 0. mn For the UAV on the path (m, n)∈E d The driving time, If the UAV is at u∈V t Take off, to w∈V t Landing, and (m, n)∈E d If it is 1, it is 0 otherwise. D is the path of the drone. Constraint (23) ensures that the total search and rescue time of the truck-UAV does not exceed the maximum search and rescue time limit, N × T rescue For the search and rescue time of all tasks, a fixed search and rescue time is reserved for each group of search and rescue tasks in the model; The time taken for the mission is different when the UAV and the truck perform the search and rescue mission respectively. The completion time of each search and rescue mission should be the maximum of the two mission times. If the truck arrives at the rendezvous point first, the maximum value is the time taken for the UAV, and its value is: Formula (24) represents the time it takes for the UAV to perform the search and rescue mission. If the UAV arrives at the rendezvous point first, the maximum value is the truck time t uw x uw , the sum of all maximum mission times is the total search and rescue time, and cannot exceed the total time D; Constraints (25)-(26) are truck path constraints, which are used to limit the legality of the truck path. Formula (25) ensures that each shelter node is visited at most once while ensuring flow balance; Formula (26) ensures that the truck starts from the starting point and eventually returns to the initial starting point; Constraints (27)-(29) are the UAV path constraints. Formula (27) is the flow balance constraint of the UAV path, which ensures the legality of the path; Formula (28) ensures that the UAV trip ending at node (u, w) cannot exceed the maximum range of the UAV; Formula (29) ensures that for the UAV trip ending at node (u, w), it must take off at node u and land at node w, and this trip can only be selected once: Constraint (30) is to ensure the auxiliary constraint of the drone path, which is used to eliminate the illegal task nodes separated from the starting point of the flight, and ensure the continuity of each group of flight task nodes of the drone. In the drone network, for a given drone trip starting from node u and ending at node w, the drone node set V d There are multiple subsets S, (m', n'), If and only if (31) That is, when the UAV ends at the node (u, w), it can go to other nodes under the UAV path subset; Constraints (32)-(33) are path node constraints, which transform the UAV path variables with a u and g m Association, ensuring that the starting node of the drone subpath is the same as the truck node: Among them, V t ′ is the union of the truck node and the vehicle segment. The constraints of equations (34) and (35) are used to ensure the coordination between the UAV and the truck. Equation (34) ensures that the UAV can perform the task on the path (u, w) only when the truck path contains the path (u, w); Equation (35) restricts that the UAV can start driving at the same truck node u only when the truck path contains the node u. Among them, E d is the UAV path set, E t is the truck path set, equations (36) and (37) define the variables of the drone truck, constraint (36) is the path decision variable of the truck and the drone; constraint (37) indicates whether the shelter node is visited.

2. The method according to claim 1, characterized in that In step 2, obtaining road network safety data includes: The BPR function represents the relationship between the travel time of a certain road section and the road flow. The travel time only considers the actual flow of the road, and does not consider the flow of adjacent paths. The actual travel time of the road is calculated using the BPR function: t(x)=t0[1+α(x / c0) β ] (1) Where t(x) represents the path travel time with path flow x.

3. The method according to claim 1, characterized in that In step S3, the improved genetic algorithm is used to solve the path planning model. The genetic algorithm based on the greedy strategy generates a population of drone paths that meet the needs in the local space. Combined with the global search characteristics of the genetic algorithm, an optimized truck path is generated for each drone path.

Images