A post-disaster multi-vehicle relay rescue method considering multiple terrains

Abstract

The present invention discloses a post-disaster multiple-vehicle relay rescue method considering multiple landforms, belonging to the field of rescue technology. According to the different landforms, the difficulty of different vehicles to pass and the degree of harm to the wounded are given; the disaster area environment is divided according to the different landforms, and the number of divisions is counted; it is judged whether the rescue vehicle can directly reach a certain division, and the direct arrival matrix of the vehicle is given according to the specific landforms and division conditions; the rescue time, the number of rescued people, and the rescue cost are used as the objective function to construct a mathematical model with the number of rescued people as the main goal; the relay path of all vehicles is optimized by a particle swarm algorithm to obtain the divisions passed by each vehicle. The present invention determines the number of vehicles actually participating in the rescue according to the number of survivors and the minimum vehicle capacity, thereby determining the length and dimension of the particles, which can ensure that the robot rescues the most survivors in the shortest possible time and at the least possible cost.

Description

Technical Field

[0001] The invention belongs to the field of rescue technology, and in particular relates to a post-disaster multiple-vehicle relay rescue method taking multiple landforms into consideration. Background Art

[0002] From ancient times to the present, disasters caused by floods, earthquakes, fires and many other human factors have caused great losses to mankind, and the disasters caused by them may cause a series of chain reactions, such as landslides caused by earthquakes and dam breaches caused by floods, which have caused great losses to people's lives and production and life. More importantly, the above-mentioned chain disasters have also brought great inconvenience to rescue.

[0003] After a disaster occurs, the topographic environment will change, and the travel of some means of transportation will be extremely restricted. Therefore, in order to complete the rescue work faster, it is necessary to combine multiple means of transportation according to the topographic environment and complete the rescue through relaying of different means of transportation.

[0004] It is worth noting that different means of transportation are suitable for different terrains. For example, helicopters are suitable for all terrains. However, considering the cost factor, helicopters are more suitable for mountain rescue, large trucks are suitable for hill rescue, and ambulances are more suitable for driving on plains. It is easy to know that after a disaster, the environment changes badly and there are also certain casualties. Disaster rescue must race against time, and at the same time, it is necessary to ensure that the vital signs of the survivors are intact. These are all issues that need to be considered during the rescue process.

[0005] Generally, disaster rescue is divided into three processes: the first is life exploration, which uses a swarm of drones to traverse the entire environment to find survivors with vital signs and quickly locate their positions; the second is specific rescue, which uses the fastest way to rescue the survivors with the least loss based on the vital signs information; the third is to rescue the survivors from the disaster area and send them to the nearest rescue center or hospital for treatment or rehabilitation. In these three processes, for the collection of vital signs information, the current drone control technology has been developed very mature, and the existing technology can quickly and accurately explore vital signs; and there are currently many effective rescue strategies for the specific rescue process. For the last process, the present invention takes into account the multi-geographic environment after the disaster and provides a specific relay rescue strategy, in order to quickly send the survivors to the hospital, ensure that the number of rescued people is the largest, and at the same time spend less time and cost.

[0006] For specific rescue scenarios, if the terrain is mountainous, helicopters and other means of assistance are needed; however, if the survivors are sent to the nearest rescue center or hospital, considering the scarcity, cost and subsequent specific rescue situation of helicopters, ambulances and other means of transportation can be used to carry out the rescue; in addition, in hilly terrain, off-road vehicles, trucks, etc. can pass effectively, and the above-mentioned means of transportation are the most efficient for transporting the wounded. Different terrains require consideration of different means of transportation, and the relay and selection between multiple means of transportation have a greater impact on the rescue results, thereby affecting the efficiency of the entire rescue. Summary of the invention

[0007] Purpose of the invention: In order to solve the problems existing in the prior art, the present invention provides a post-disaster multiple-vehicle relay rescue method that takes into account multiple terrains, promotes the rapid and effective completion of rescue missions, and rescues as many survivors as possible.

[0008] A post-disaster multiple-vehicle relay rescue method considering multiple terrains comprises the following steps:

[0009] 1) Partition the post-disaster environment and provide a direct arrival matrix R;

[0010] 2) Establish a mathematical model of the number of rescuers, rescue time and rescue cost, and then obtain the objective function of the rescue problem; where the decision variable in the mathematical model is the path P of each vehicle, where P i ={p i1 , p i2 , p i3 ,…,p ij ,…,p i(U-1) , p iU} represents the travel path of the ith vehicle, that is, the terrain partition it passes through, where p ij ∈Z,i=1,2,…,VN,j=1,2,…,U,p ij Indicates the number of the geomorphic partitions that the vehicle passes through; U indicates the number of areas that the vehicle actually passes through, VN indicates the total number of vehicles; y i Indicates whether the i-th means of transport is in use. If it is 1, it means it is in use; if it is 0, it means it is not in use.

[0011] 3) The particle swarm optimization algorithm (PSO) is used to solve the problem, which includes the following steps:

[0012] Step 3.1: Initialize the particles, including the size of the particle group, the position and speed of the particles, etc.;

[0013] Step 3.2: Decode and evaluate each particle according to the objective function of the problem, and select the individual extreme value (pbest) and the global extreme value (gbest);

[0014] Step 3.3: For each particle, perform the following operations:

[0015] Step 3.31: Update the velocity and position of each particle;

[0016] Step 3.32: Evaluate the fitness value of each particle according to the established objective function;

[0017] Step 3.33: Update pbest and gbest;

[0018] Step 3.4: Determine whether the algorithm meets the termination condition. If so, the algorithm ends and outputs the global extreme value; otherwise, return to step 3.3.

[0019] Furthermore, the specific partitioning method is as follows:

[0020] Step 1: Use the disaster area and the hospital as diagonal vertices to construct a rectangular rescue map;

[0021] Step 2: Connect the other two vertices of the rectangle to obtain another diagonal line of the rectangle, which is the first-level partition, recorded as partition line l1;

[0022] Step 3: Taking the disaster area and the hospital as the starting points, continue to use straight lines to divide according to the similarities and differences of the landform environment, and the straight line obtained is the partition line l2; further determine whether partitioning is required. If necessary, go to step 4; otherwise, go to step 6;

[0023] Step 4: In the polygon formed by the two partition lines l2 and l1, according to the similarities and differences of the landform environment, use straight lines to further divide, and obtain the partition line l3; further determine whether it is necessary to divide, if necessary, go to step five; otherwise, go to step six;

[0024] Step 5: Continue partitioning the polygon formed by partition lines l2 and l3 to obtain a new partition line l4. If no further partitioning is required, go to step 6;

[0025] Step 6: After the partitioning is completed, the number of partitions ZN is counted.

[0026] 3. A post-disaster multiple-vehicle relay rescue method considering multiple terrains according to claim 1, characterized in that in step 1), in the divided partitions, the direct reach matrix R between two partitions (N×N) , expressed as shown in formula (1):

[0027]

[0028] Among them, r ij is a Boolean variable, rij 0 means it cannot be reached directly; r ij If it is 1, it means direct arrival.

[0029] Furthermore, in step 2), the rescue time is the maximum value of the travel time of all vehicles, that is, the total travel time of different vehicles passing through U landforms, as shown in formula (2):

[0030]

[0031] Where, T represents the rescue time; v ij represents the speed of the i-th vehicle in the j-th terrain environment, A ij represents the difficulty of the i-th vehicle traveling in the j-th terrain; Length(p ij ) represents the path length of the i-th vehicle in the j-th terrain, which can be specifically expressed as |p ij p i(j-1) | represents the distance between two partitions.

[0032] Furthermore, in step 2), the number of rescuers S is as shown in formula (3):

[0033]

[0034] Among them, ST m represents the initial life span of the mth survivor, Δt represents the time of handing over survivors between two different landforms, represents all the time of carrying survivor m, max represents the maximum value function, sgn(x) represents the sign function,

[0035] Furthermore, in step 2), assuming that the i-th vehicle is running normally, its unit length cost is C i , the cost of driving on the jth terrain is C i / A j , then the rescue cost is expressed as formula (4):

[0036]

[0037] Among them, Weg i Represents the number of survivors on the i-th vehicle.

[0038] Furthermore, in step 2), the objective function is as shown in formula (5):

[0039] As shown in formula (5):

[0040]

[0041] y i ∈{0,1} (8)

[0042] Weg i ≤Capacity (9)

[0043] Where F represents the objective function, represents the normalization operation; Formula (6) represents the path p ij and path p ij′ There is no common rescue mission; Formula (7) indicates that the two geomorphic partitions through which the vehicle passes can be directly reached, indicating that it is passable; Formula (8) indicates the value range of the variable; Formula (9) indicates that the load of each vehicle cannot exceed the rated number of people, here we take Capacity = min{Capaticy1, Capaticy2, ..., Capaticy U}, which is the minimum load capacity of all transportation vehicles.

[0044] Furthermore, in step 2), for the rescue of M survivors, it is first necessary to allocate transportation vehicles to them and determine the number of survivors Weg in each vehicle. The specific allocation strategy is as follows:

[0045] Step 2.1: Determine the number of vehicles actually involved in the rescue VN R ,Right now, indivual;

[0046] Step 2.2: Randomly select VN R survivors are assigned to each mode of transport;

[0047] Step 2.3: Calculate the number of remaining survivors and determine whether the allocation is completed. If it is 0, proceed to step 2.5; if the remaining number is greater than VN R , then go to step 2.2; if less than, go to step 2.4;

[0048] Step 2.4: For the remaining survivors, randomly select the means of transportation corresponding to the number of remaining survivors for allocation;

[0049] Step 2.5: Count the number of survivors Weg on each vehicle i , ending the allocation.

[0050] Furthermore, in the algorithm design in step 3), first, we need to clarify the relationship between the solution to the problem (the vehicle's driving path) and the particles. Secondly, we need to determine the length and dimension of the particles. The specific steps are as follows:

[0051] Step 3.1: Classify and count the geomorphic areas on both sides of the partition line l1. The statistical direction is from the starting point to the end point, so that two sets can be obtained;

[0052] Step 3.2: For the two sets obtained in step 3.1, classify and count the terrain areas on both sides of the partition line l2;

[0053] Step 3.3: Use the same method to classify and count the terrain areas on both sides of the partition line l3;

[0054] Step 3.4: Using the above method, the total number of cycles is max{l 左 , l 右}-1, where l 左 Indicates the maximum number of final partition lines to the left of partition line l1; after the loop is completed, the terrain set obtained by statistics is rotated 900, that is, the transposed form ZR, and the number of subsets obtained is the length of the particle, that is, U. In this step, it is also necessary to count the number of integers (that is, terrain areas) in each set, recorded as {Num1, Num2, Num3, ..., Num U};

[0055] The particle code is in integer form, and the kth particle is represented as: in, Decode the particles, assuming X k Any element in That is, the encoding for the jth region can be expressed as: The number of types of transportation in each area is: TN = {TN1, TN2, TN3, ..., TN N}, the number of integers in each set obtained in step 3.4 above is {Num1, Num2, Num3, ..., Num U}, the specific decoding steps are as follows:

[0056] Step 1: From Extract elements from Execute it remainder operation;

[0057] Step 2: According to the remainder result obtained in step 1, an integer corresponding to the remainder result is taken from the first rotation set ZR, thereby obtaining the number of an area passed by the vehicle;

[0058] Step 3: Then The remaining elements in perform the operations of step 1 and step 2, so as to obtain the path of the vehicle, that is, the U areas that the vehicle needs to pass through;

[0059] Step 4: According to the vehicle travel area number obtained in step 3, the number of vehicle types in the corresponding area is taken from TN and recorded as TN q ={TN q1 , T.N. q2 , T.N. q3 ,…,TN qU};

[0060] Step 5: From Extract elements from Execute it The result of the modulo operation is the model selected in the area. The number of corresponding models minus 1, the corresponding y i Set to 1; if it is 0, TN q1 -1.

[0061] Step 6: Select Take out the elements and execute step 5 to get the car models used in each area; determine whether the car models of all areas have been selected. If yes, go to step 7, otherwise, continue to execute the same operation as step 5;

[0062] Step 7: Decoding is completed and the decoded result is output.

[0063] Correct the infeasible solution generated by directly reaching the matrix R; assume that an element of the k-th particle That is, for the encoding of the jth area, there are two areas that cannot be reached directly. Assuming that they are decoded, the driving path obtained is represented by P j ={p j1 , p j2 ,…,p ju ,…,p jU}, assuming that the vehicle travels to area p ju After that, the transportation cannot directly reach the next area p j(u+1) , the solution is an infeasible solution, then the correction method is as follows:

[0064] Step 1: Find the pth node in the matrix R ju row, find the region p from this row ju The area that can be directly reached, that is, the area number with all 1s, is denoted as P juR ={p juR1 , p juR2 ,…,p jRk}, Rk represents all regions p ju The number of areas reached directly;

[0065] Step 2: For the pth ju +1 set, recorded as Search With step 1 P juR the same elements;

[0066] Step 3: Determine whether the element in step 2 is the Uth element. If so, end the correction operation and output the corrected result; if not, j Perform the same operation as step 1 for the next element.

[0067] Particle velocity and position update formula

[0068] The present invention adopts integer coding because the particle position and velocity update formulas are as shown in formulas (10) and (11):

[0069] v kd (t+1)=wv kd (t)+c1r1(p kd (t)-x kd (t))+c2r2(p gd (t)-x kd (t)) (10)

[0070]

[0071] Among them, v kd represents the velocity of the d-dimensional element of particle k, xk d represents the position of the d-dimensional element of particle k, c1 and c2 are constants, r1 and r2 are random numbers between [0, 1], and p kd is the d-th dimension element of the individual extreme value of particle k, indicating the best position (fitness value) experienced by the individual, p gd It is the d-th dimension element of the global extremum of all particles, indicating the best position experienced by the entire group.

[0072] Pbest and gbest update methods

[0073] The individual extreme value pbest and the global extreme value gbest guide the evolution direction of the particles. As the algorithm runs, the particles are constantly evolving according to formulas (10) and (11). Then the individual extreme value and the global extreme value of the particles will also be continuously updated during the evolution, thereby better guiding the particles to search for better solutions. Therefore, the selection of the individual extreme value pbest and the global extreme value gbest is very important. gbest is selected from pbest. Therefore, the selection method of pbest is given first.

[0074] The idea of ​​updating individual extreme values ​​is: if the difference between the fitness value F(Curpar) of the current particle and the fitness value F(pbest) of pbest is greater than a certain threshold Δ f , that is, F(Curpar)-F(pbest)>Δf , then pbest is updated to the current particle; if |F(Curpar)-F(pbest)|≤Δ f , then compare the number of rescuers of the two. If S(Curpar)>S(pbest), pbest is updated to the current particle; otherwise, compare the rescue time of the two. If T(Curpar)>T(pbest), pbest is updated to the current particle; otherwise, pbest remains unchanged.

[0075] After obtaining pbest, the selection strategy of gbest is given: if the fitness value F(Curpar) of the current particle is greater than the fitness value F(gbest) of gbest, if F(Curpar)>F(gbest), gbest is updated to the current particle; otherwise, gbest remains unchanged.

[0076] Beneficial effects: Compared with the prior art, the post-disaster multiple-vehicle relay rescue method of the present invention, which takes multiple landforms into consideration, can rescue as many survivors as possible with less rescue time and rescue cost, and effectively complete rescue tasks in multiple landform environments. Considering the complex and changeable environment of the disaster area, the environment is divided according to the similarities and differences of the landforms, and the direct arrival matrix of the transportation tools is given; considering that different landforms only allow one or some transportation tools to pass, the present invention adopts multiple transportation tools for relay rescue, such relay ensures the smooth progress of the rescue and improves the rescue efficiency; according to the difficulty of passing multiple transportation tools in a certain landform and the degree of harm to the survivors, the three goals of rescue time, number of rescuers and rescue cost are considered, and the above three goals are weighted according to the importance of the goals to construct a single objective function, which simplifies the problem and shortens the running time of the algorithm, which is of great significance for rescue; the particle swarm algorithm is used to solve the problem, and the solution of a problem is for one particle. Through the decoding method of the particle, repeated rescue will not occur, and the occurrence of conflicts is avoided. More importantly, the relay rescue of multiple transportation tools can effectively save costs, speed up the completion of rescue tasks, and rescue as many survivors as possible. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 This is a schematic diagram of the robot rescue environment;

[0078] Figure 2 It is a schematic diagram of the partitioning method;

[0079] Figure 3 This is the flowchart for solving the PSO algorithm. DETAILED DESCRIPTION

[0080] The present invention is further described below in conjunction with the accompanying drawings and specific implementation examples. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0081] The present invention considers the above-mentioned multi-terrain and multi-vehicle relay rescue problem, and hopes to provide an optimal or suboptimal rescue route to ensure that the rescue mission is completed quickly and effectively to rescue as many survivors as possible. It can be seen that the above-mentioned problem is a multi-terrain, multi-vehicle, and multi-route planning problem, which is an NP-hard problem. The solution of this problem is generally solved by evolutionary optimization methods. Considering that particle swarm evolutionary optimization methods are widely used in path planning problems, the present invention intends to use particle swarm algorithm to solve the above problem.

[0082] 1. Problem and Environment Description

[0083] 1.1 Environment Description

[0084] In order to effectively solve the above problem without loss of generality, the following assumptions are made:

[0085] 1. If a subsequent secondary disaster occurs, it will not significantly affect the landforms in and outside the disaster area;

[0086] 2. The power or fuel of all vehicles is sufficient for the entire rescue operation;

[0087] 3. The speed of each vehicle is constant during its travel;

[0088] 4. The same type of transportation has the same distribution location and difficulty level in the same terrain, and the same type of transportation causes the same damage to survivors in the same terrain;

[0089] Based on the above assumptions, the problem to be solved by the present invention is as follows: In the post-disaster environment, there are N different landforms, V different types of vehicles, and the number of the j-th type of vehicle in the i-th landform is , i = 1, 2, ..., N, j = 1, 2, ..., V, the positions of each vehicle are known. In the whole environment, at a certain moment, there are M survivors (target points) waiting for relay rescue, among which, Represents the total number of all vehicles waiting for rescue in the i-th area. From the above analysis, we can see that each survivor suffers different degrees of injury, so their survival time is limited. It is required to use the above-mentioned different types of vehicles to implement relay rescue and send the survivors to the nearest rescue center or hospital as soon as possible, so as to ensure that under the linkage of multiple vehicles, the number of rescued people is as large as possible, the damage to the survivors is minimized, and the rescue cost is the lowest.

[0090] 1.2 Environment Description

[0091] Figure 1 A schematic diagram of the rescue using multiple means of transportation. It can be seen from the figure that the survivors are distributed in different locations in the disaster area. The same type of vehicles are distributed in the same location. For example, trucks are distributed on the road, and can also be distributed on hills or grasslands; ambulances are distributed on the road; and helicopters can be distributed on all terrains, etc. Therefore, in the process of carrying out the rescue, we must not only consider the purpose of various means of transportation, but also consider different terrains. At the same time, the survival time of the survivors is also a factor that needs to be considered. Only in this way can we achieve the premise of rescuing as many survivors as possible while minimizing the rescue cost and time.

[0092] However, different means of transportation have different difficulties in passing on different landforms, and the degree of harm to survivors is also different. For example, helicopters are equipped with professional rescue equipment, which can effectively extend the life span, so they have a positive protective effect on survivors; while trucks driving on bumpy grass or hills, and without more professional rescue equipment in the car, will accelerate the consumption of life span, and have a reverse harm effect on survivors; in addition, it is more difficult for trucks to drive in post-disaster landforms than on flat roads. Based on the above analysis, Table 1 gives the difficulty of passing various means of transportation under different landforms, as well as the harm index to survivors. Among them, in A / B, A represents the difficulty of passing, the larger the A, the greater the difficulty of passing, 0 means unobstructed, and 1 means impassable; B represents the degree of harm to survivors when vehicles travel on different landforms, the larger the B, the greater the harm, B>0 means harm; and B<0 means getting help, which enhances the strength of life.

[0093] Table 1 Passage difficulty and injury index

[0094]

[0095] 1.3 Terrain processing

[0096] from Figure 1 It can be seen that the landform division is relatively obvious. In order to simplify the subsequent calculation, the terrain needs to be processed first. On the basis of using the existing and mature over-segmentation method for the map, the present invention proposes a partitioning method. For the post-disaster environment, considering the actual rescue landform environment, the partitioning is performed up to four times. The specific method is as follows:

[0097] Step 1: Use the disaster area (starting point) and the hospital (end point) as diagonal vertices to construct a rectangular rescue map. Figure 1 For example, the constructed rectangular map is as follows Figure 2 As shown, the hollow circles represent disaster areas, and the hollow squares represent hospitals;

[0098] Step 2: Connect the other two vertices of the rectangle to obtain another diagonal line of the rectangle, which is the first-level partition, denoted as partition line l1. Figure 1 For example, Figure 2 The other diagonal line in is the partition line l1;

[0099] Step 3: Taking the disaster area and the hospital as the starting points, continue to use straight lines to divide according to the landform environment, and the straight line obtained is the partition line l2. Figure 1 Environment, according to the specific landform conditions, use straight lines to divide, the straight line obtained is Figure 2 The partition line l2 in the process; further determine whether partitioning is required. If so, go to step 4; otherwise, go to step 6.

[0100] Step 4: In the polygon formed by the two partition lines l2 and l1, according to the landform environment, use straight lines to further divide and obtain the partition line l3. Then, for Figure 1 The rescue environment is further divided into each area formed by the partition lines l2 and l1, and the Figure 2 Partition line l3 in the process; further determine whether partitioning is required. If so, go to step five; otherwise, go to step six.

[0101] Step 5: Continue partitioning the polygon formed by partition lines l2 and l3 to obtain a new partition line l4, which is Figure 2 If partitioning is not required, go to step 6.

[0102] Step 6: After the partitioning is completed, the number of partitions ZN is counted.

[0103] like Figure 2 As shown, Figure 1 The rescue environment is divided into 13 landforms, from Figure 2 It can be seen that, except for helicopters, other means of transportation cannot directly reach partition 11 from partition 1. In addition, vehicles cannot directly reach partition 4 from partition 8. The possibility of direct arrival of two partitions is 0. Based on the above analysis, the present invention gives the direct arrival matrix R (N×N) , expressed as shown in formula (1):

[0104]

[0105] Among them, r ij is a Boolean variable, r ij A value of 0 indicates that the destination cannot be reached directly; a value of 1 indicates that the destination can be reached directly.

[0106] In order to better illustrate the construction method of directly reaching the matrix, Figure 2As an example, it can be seen that the same area can be reached directly, but in order to prevent the vehicle from stagnating in a certain area, the present invention sets r ii =0, then Figure 2 The direct access matrix R for the environment shown is as follows:

[0107]

[0108] 2. Model Construction

[0109] Considering the particularity of the rescue problem, the objectives to be optimized in the present invention are the number of rescued people, rescue time and rescue cost, requiring the rescue mission to be completed with the least possible cost and the shortest possible time under the premise of rescuing the most survivors. Using the above information, a mathematical model of the above three objectives is given.

[0110] The decision variable in the present invention is the path P of each vehicle, where P i ={p i1 , p i2 , p i3 ,…,p ij ,…,p i(U-1) , p iU} represents the driving path of the i-th vehicle, where p ij ∈Z,i=1,2,…,VN,j=1,2,…,U,p ij Indicates the number of the landform that the vehicle passes through, such as Figure 2 The landform divisions 1 to 13 in the figure; U represents the number of areas actually passed by the vehicle, and VN represents the total number of vehicles. i Indicates whether the i-th means of transportation is in use. If it is 1, it means it is in use; if it is 0, it is not in use.

[0111] 2.1 Rescue Time

[0112] Based on the above analysis, the travel time required for different vehicles to pass through U landforms is given. Since the rescue work of each vehicle is carried out in parallel, the rescue time of the present invention is the maximum value of the travel time of all vehicles. The specific calculation formula is shown in formula (2).

[0113]

[0114] Where, T represents the rescue time; v ij represents the speed of the i-th vehicle in the j-th terrain environment, A ij represents the difficulty of the i-th vehicle traveling in the j-th terrain; Length(p ij ) represents the path length of the i-th vehicle in the j-th terrain, which can be specifically expressed as |p ij p i(j-1) | represents the distance between two landforms.

[0115] 2.2 Number of rescuers

[0116] The number of rescued people is the most important indicator in the rescue process, which is closely related to the life span of the survivor. If the survivor is sent to the hospital and is still within his life span, it means that he is successfully rescued. At this time, the number of rescued people increases by 1; if he exceeds his life span, it means that the rescue fails. At this time, the number of rescued people does not change. Based on the above analysis, the mathematical expression of the number of rescued people is given, as shown in formula (3).

[0117]

[0118] Among them, ST m represents the initial life span of the mth survivor, Δt represents the time of handing over survivors between two different landforms, represents all the time of carrying survivor m, max represents the maximum value function, sgn(x) represents the sign function,

[0119] 2.3 Rescue costs

[0120] In the present invention, the rescue cost is mainly reflected in the driving cost of the vehicle. If the vehicle is put into use, the cost needs to be calculated. If it is not put into use, the cost is not calculated. Assuming that the i-th vehicle is running normally, its unit length cost is C i , the cost of driving on the jth terrain is C i / A j , then the total cost of rescue is expressed as formula (4).

[0121]

[0122] Among them, Weg i represents the number of survivors on the i-th vehicle.

[0123] 2.4 Mathematical model

[0124] From the above analysis, we can know that rescuing the largest number of people is the most important goal of rescue, and rescue time and rescue cost are secondary goals. Based on this, the objective function of the rescue problem studied in this invention is given, as listed in formula (5):

[0125]

[0126] y i ∈{0,1} (8)

[0127] Wegi ≤Capacity (9)

[0128] Wherein, F represents the objective function to be solved by the present invention, represents the normalization operation; Formula (6) represents the path p ij and path p ij′ There is no common rescue mission (survivors); (7) indicates that the vehicle can directly reach between two consecutive partitions, which means that it is passable; (8) indicates the value range of the variable; (9) indicates that the load of each vehicle cannot exceed the rated number of people, here we take Capacity = min{Capaticy1, Capaticy2, ..., Capaticy U} to ensure the timeliness and convenience of transportation in relay rescue.

[0129] 3. Algorithm Solution

[0130] The present invention intends to use particle swarm optimization algorithm (PSO) to solve the problem. The solution process of the PSO algorithm is as follows: Figure 3 shown.

[0131] Specifically:

[0132] Step 1: Initialize the particles, including the size of the particle group, the position and speed of the particles, etc.

[0133] Step 2: Decode and evaluate each particle according to the established problem objective function, and select the individual extreme value (pbest) and the global extreme value (gbest);

[0134] Step 3: For each particle, perform the following operations:

[0135] Step 3.1: Update the velocity and position of each particle;

[0136] Step 3.2: Evaluate the fitness value of each particle according to the established objective function;

[0137] Step 3.3: Update pbest and gbest;

[0138] Step 4: Determine whether the algorithm meets the end condition. If so, end and output the global extreme value; otherwise, return to step 3.

[0139] 3.1 Personnel allocation

[0140] For the rescue of M survivors, we first need to allocate transportation to them. The specific allocation strategy is as follows:

[0141] Step 1: Determine the number of vehicles actually involved in the rescue VN R ,Right now, indivual;

[0142] Step 2: Randomly select VN R survivors are assigned to each mode of transport;

[0143] Step 3: Calculate the number of remaining survivors to determine whether the allocation is complete. If it is 0, proceed to step 5. If the remaining number is greater than VN R , then go to step 2; if less than, go to step 4;

[0144] Step 4: For the remaining survivors, randomly select the means of transportation corresponding to the number of remaining survivors for allocation;

[0145] Step 5: Count the number of survivors Weg on each vehicle i , ending the allocation.

[0146] 3.2 Determination of particle length and dimension

[0147] Before using the particle swarm algorithm to solve the above problem, we must first clarify the relationship between the solution to the problem (the vehicle's driving path) and the particles. Secondly, we need to determine the length and dimension of the particles.

[0148] Regarding the relationship between the solution to the problem and the particles, in the present invention, a particle represents the path of the vehicle put into use, that is, it represents a solution to the problem, wherein the dimension of the particle represents the area number passed by the vehicle, and the length represents the number of vehicles put into use. Specifically, the dimension and length of the particle need to be determined according to the specific area division, so how to determine the dimension and length of the specific particle?

[0149] According to Section 3.1, we need to determine the travel path of each vehicle. Then, the dimension of the particle is the number of vehicles participating in the rescue in the origin area VN R .

[0150] According to the terrain processing method in Section 1.3, the areas with the same partition line level are classified to determine the length of the particles. The specific steps are as follows:

[0151] Step 1: Classify and count the landforms on both sides of the partition line l1, and the direction of the statistics is from the starting point to the end point. Figure 2 For example, the left side is {1, 2, 3, 4, 5, 6, 7}, and the right side is {11, 12, 13, 8, 9, 10}.

[0152] Step 2: For the two sets obtained in step 2, classify and count the terrain areas on both sides of the partition line l2. Figure 2 For example, the left set can be divided into {1, 2, 3} and {4, 5, 6, 7}; the right set can be divided into {11, 12, 13} and {8, 9, 10}.

[0153] Step 3: Use the same method to classify and count the terrain areas on both sides of the partition line l3. Figure 2 The left side can be divided into {1, 2, 3}, {4, {5, 6}, 7}; the right side remains unchanged.

[0154] Step 4: Using the above method, the total number of cycles is max{l 左 , l 右}-1, where l 左 Indicates the maximum number of final partition lines to the left of partition line l1, Figure 2 For example, the left partition line is l4, then l 左 is 4; and l 右 Then it is the maximum number of final partition lines to the right of partition line l1, Figure 2 Middle 右 is 3. After the loop is finished, the terrain set obtained by counting is rotated 900 degrees, that is, the transposed form ZR, and the number of subsets obtained is the length of the particle, that is, U. In this step, it is also necessary to count the number of integers (that is, terrain areas) in each set, recorded as {Num1, Num2, Num3, ..., Num U}.by Figure 2 For example, the set after rotation is: That is, there are 4 parts, so for Figure 1 In the rescue scenario, the encoding length of the particles is 4, and the number of integers in each set is 3, 4, 3, and 3 respectively.

[0155] Note: When a set on one side does not need to be classified, in order to ensure the consistency of the set level, the remaining set classification is divided internally and no longer split into a single set, see step 3. In addition, when rotating 900, the internally split sets remain unchanged, see step 4.

[0156] 3.3 Particle Encoding and Decoding

[0157] Since the problem is to find the number of the landform partitions that the vehicle passes through, the particle encoding of the present invention is in the form of integers, and the kth particle is represented by: in, So, how to get the vehicle's driving path from the particles (U integers)?

[0158] To solve the above problem, the particles need to be decoded. Assume that the k-th dimension of the k-th particle can be expressed as: The number of types of transportation in each area is: TN = {TN1, TN2, TN3, ..., TN N}, the number of integers in each set obtained in step 4 of section 3.2 above is {Num1, Num2, Num3, …, Num U}, the specific decoding steps are as follows:

[0159] Step 1: From Extract elements from Execute it remainder operation;

[0160] Step 2: According to the remainder result obtained in step 1, an integer corresponding to the remainder result is taken from the set ZR, thereby obtaining the number of an area passed by the vehicle;

[0161] Step 3: Then The remaining elements in perform the operations of step 1 and step 2, so as to obtain the path of the vehicle, that is, the U areas that the vehicle needs to pass through.

[0162] Step 4: According to the vehicle travel area number obtained in step 3, the number of vehicle types in the corresponding area is taken from TN and recorded as TN q ={TN q1 , T.N. q2 , T.N. q3 ,…,TN qU};

[0163] Step 5: From Extract elements from Execute it The result of the modulo operation is the model selected in the area. The number of corresponding models minus 1, the corresponding y i Set to 1; if it is 0, TN q1 -1.

[0164] Step 6: Select Take out the elements and execute step 5 to get the model used in each area. Determine whether the model of all areas has been selected. If yes, go to step 7. Otherwise, continue to execute the same operation as step 5.

[0165] Step 7: Decoding is completed and the decoded result is output.

[0166] Note: Taking into account the continuous back and forth transportation of rescue vehicles, the number of each element in TN is changing. When a vehicle is dispatched for transportation in a certain terrain area, the number of corresponding vehicles is reduced by 1. If it is reduced to 0, the corresponding type of vehicle is reduced by 1; and when the vehicle is transported back, the number of corresponding vehicles is increased by 1, and the number of types of vehicles remains unchanged.

[0167] 3.4 Correction Operation

[0168] Step 2: According to the remainder result obtained in step 1, an integer corresponding to the remainder result is taken from the set ZR, thereby obtaining the number of an area passed by the vehicle;

[0169] Step 3: Then The remaining elements in perform the operations of step 1 and step 2, so as to obtain the path of the vehicle, that is, the U areas that the vehicle needs to pass through.

[0170] Step 4: According to the vehicle travel area number obtained in step 3, the number of vehicle types in the corresponding area is taken from TN and recorded as TN q ={TN q1 , T.N. q2 , T.N. q3 ,…,TN qU};

[0171] Step 5: From Extract elements from Execute it The result of the modulo operation is the type of transportation selected in the area. The number of corresponding types minus 1, the corresponding y i Set to 1; if it is 0, TN q1 -1.

[0172] Step 6: Select Take out the element and execute step 5 to get the type of transportation used in each area. Determine whether all the types of areas are selected. If yes, go to step 7. Otherwise, continue to execute the same operation as step 5.

[0173] Step 7: Decoding is completed and the decoded result is output.

[0174] Note: Taking into account the continuous transport back and forth of rescue vehicles, the number of each element in TN is changing. When a vehicle is dispatched for transportation in a certain terrain area, the number of corresponding vehicles is reduced by 1. If it is reduced to 0, the corresponding type of vehicle is reduced by 1. When the vehicle is transported back, the number of corresponding vehicles is increased by 1, and the number of types of vehicles remains unchanged.

[0175] 3.4 Correction Operation

[0176] The decoded particles may have infeasible solutions. The infeasibility is mainly reflected in the fact that after the particles are decoded, two adjacent areas in the driving path of one or several vehicles cannot be directly reached. Considering the existence of the above infeasible solutions, the present invention corrects the generated infeasible solutions based on the direct arrival matrix R. Assume There are two areas that cannot be reached directly. Decode them and assume that the driving path obtained is P j ={p j1 , p j2 ,…,p ju ,…,p jU}, assuming that the vehicle travels to area p ju After that, the transportation cannot directly reach the next area p j(u+1) , the solution is an infeasible solution, then the specific correction method is as follows:

[0177] Step 1: Find the pth node in the matrix R ju row, find the region p from this row ju The area that can be directly reached, that is, the area number with all 1s, is denoted as P juR ={p juR1 , p juR2 ,…,p jRk}, Rk represents all regions p ju The number of areas reached directly;

[0178] Step 2: For the pth ju +1 set, recorded as Search With step 1 P juR the same elements;

[0179] Step 3: Determine whether the element taken in step 2 is the Uth element. If so, end the correction operation and output the corrected result. If not, perform the correction on P. j Perform the same operation as step 1 for the next element.

[0180] by Figure 1 Take the rescue scenario as an example, the path set is The number of particles in the set is 3, 4, 3, 3 respectively. Assuming that the particles are {20, 6, 11, 99}, the decoding method in Section 4.2 is used to decode them into {3, 6, 11, 10}. From the reach matrix, we can see that the third row, that is, the area that can be directly reached from area 3 is {2, 6, 7}. Area 6 can be directly reached from area 3, but area 6 cannot directly reach area 11. Therefore, this solution is an infeasible solution. It is infeasible to go from area 6 to area 11. Then, find the matrix that can be directly reached from area 6 from the direct reach matrix R, which is {2, 3, 5, 7, 13}. By comparing it with the third subset in the set ZR, Compare and select the same element, that is, 13; then find the direct area of ​​area 13 from the arrival matrix, which is {6, 7, 10, 12}, and Compare and select the same element, which is 10. At this point, the entire correction operation is completed, and the vehicle's driving path is {3, 6, 13, 10} 3.5 Particle speed and position update formula

[0181] The present invention adopts integer coding because the particle position and velocity update formulas are as shown in formulas (10) and (11):

[0182] v kd (t+1)=wv kd (t)+c1r1(p kd (t)-x kd (t))+c2r2(p gd (t)-x kd (t)) (10)

[0183]

[0184] Among them, v kd represents the velocity of the d-dimensional element of particle k, x kd represents the position of the d-dimensional element of particle k, c1 and c2 are constants, r1 and r2 are random numbers between [0, 1], and p kd is the individual extreme value of particle k, indicating the best position (fitness value) experienced by the individual, p gd is the global extreme value of the particle, indicating the best position experienced by the entire group.

[0185] 3.6 Pbest and gbest update methods

[0186] The individual extreme value pbest and the global extreme value gbest guide the evolution direction of the particles. As the algorithm runs, the particles are constantly evolving according to formulas (10) and (11). Then the individual extreme value and the global extreme value of the particles will also be continuously updated during the evolution, thereby better guiding the particles to search for better solutions. Therefore, the selection of the individual extreme value pbest and the global extreme value gbest is very important. gbest is selected from pbest. Therefore, the selection method of pbest is given first.

[0187] The idea of ​​updating individual extreme values ​​is: if the difference between the fitness value F(Curpar) of the current particle and the fitness value F(pbest) of pbest is greater than a certain threshold Δ f , that is, F(Curpar)-F(pbest)>Δ f , then pbest is updated to the current particle; if |F(Curpar)-F(pbest)|≤Δ f, then compare the number of rescuers of the two. If S(Curpar)>S(pbest), pbest is updated to the current particle; otherwise, compare the rescue time of the two. If T(Curpar)>T(pbest), pbest is updated to the current particle; otherwise, pbest remains unchanged.

[0188] After obtaining pbest, the selection strategy of gbest is given: if the fitness value F(Curpar) of the current particle is greater than the fitness value F(gbest) of gbest, if F(Curpar)>F(gbest), gbest is updated to the current particle; otherwise, gbest remains unchanged.

Claims

1. A post-disaster multiple-vehicle relay rescue method considering multiple terrains, characterized in that: The steps include: 1) Partition the post-disaster environment and provide a direct arrival matrix R; 2) Establish a mathematical model of the number of rescuers, rescue time and rescue cost, and then obtain the objective function of the rescue problem; where the decision variable in the mathematical model is the path P of each vehicle, where P i ={p i1 , p i2 , p i3 ,…,p ij ,…,p i(U-1) , p iU } represents the driving path of the i-th vehicle, where p ij ∈Z,i=1,2,…,VN,j=1,2,…,U,p ij Indicates the number of the geomorphic partitions that the vehicle passes through; U indicates the number of areas that the vehicle actually passes through, VN indicates the total number of vehicles; y i Indicates whether the i-th means of transport is in use. If it is 1, it means it is in use, and if it is 0, it is not in use. 3) The particle swarm optimization algorithm PSO is used to solve the problem, which includes the following steps: Step 3.1: Initialize the particles, including the size of the particle group, the position and velocity of the particles; Step 3.2: Decode and evaluate each particle according to the objective function of the problem, and select the individual extreme value pbest and the global extreme value gbest; Step 3.3: For each particle, perform the following operations: Step 3.31: Update the velocity and position of each particle; Step 3.32: Evaluate the fitness value of each particle according to the established objective function; Step 3.33: Update pbest and gbest; Step 3.4: Determine whether the algorithm meets the end condition. If so, the algorithm ends and outputs the global extreme value; otherwise, return to step 3.

3. The rescue time is the maximum value of the travel time of all vehicles, that is, the total travel time of different vehicles passing through U landforms, as shown in formula (2): Where T represents the rescue time; v ij represents the speed of the ith vehicle in the jth zone, A ij represents the difficulty of the i-th vehicle traveling in the j-th area; Length(p ij ) represents the path length of the i-th vehicle in the j-th area, which is specifically expressed as |p ij p i(j-1) | represents the distance between two geomorphic partitions; In step 2), the number of rescuers S is as shown in formula (3): Among them, ST m represents the initial life span of the mth survivor, Δt represents the time required to hand over a survivor between two different landforms, represents the time required to transport the survivor m, max represents the maximum value function, sgn(x) represents the sign function, In step 2), assuming that the i-th vehicle is running normally, its unit length cost is C i , the cost of driving on the jth terrain is C i / A j , then the rescue cost is expressed as formula (4): Among them, Weg i represents the number of survivors on the i-th vehicle; In step 2), the objective function is as shown in formula (5): st y i ∈{0，1} (8) Road i ≤Capacity (9) Where F represents the objective function, represents the normalization operation; Formula (6) represents the path p ij and path p ij′ There is no common rescue mission; Formula (7) indicates that the two geomorphic partitions through which the vehicle passes can be directly reached, indicating that it is passable; Formula (8) indicates the value range of the variable; Formula (9) indicates that the load of each vehicle cannot exceed the rated number of people, here we take Capacity = min{Capaticy1, Capaticy2, ..., Capaticy U }, which is the minimum load capacity of all transportation vehicles.

2. A post-disaster multiple-vehicle relay rescue method considering multiple terrains according to claim 1, characterized in that: In step 1), the specific method of partitioning is as follows: Step 1: Use the disaster area and the hospital as diagonal vertices to construct a rectangular rescue map; Step 2: Connect the other two vertices of the rectangle to obtain another diagonal line of the rectangle, which is the first-level partition, recorded as partition line l1; Step 3: Taking the disaster area and the hospital as the starting points, continue to use straight lines to divide according to the similarities and differences of the landform environment, and the straight line obtained is the partition line l2; further determine whether partitioning is required. If necessary, go to step 4; otherwise, go to step 6; Step 4: In the polygon formed by the two partition lines l2 and l1, according to the similarities and differences of the landform environment, use straight lines to further divide, and obtain the partition line l3; further determine whether it is necessary to divide, if necessary, go to step five; otherwise, go to step six; Step 5: Continue partitioning the polygon formed by partition lines l2 and l3 to obtain a new partition line l4. If no further partitioning is required, go to step 6; Step 6: After the partitioning is completed, the number of partitions ZN is counted.

3. A post-disaster multiple-vehicle relay rescue method considering multiple terrains according to claim 1, characterized in that: In step 1), in the partition, the direct access matrix R between two partitions (N×N) , expressed as shown in formula (1): Among them, r ij is a Boolean variable, r ij 0 means it cannot be reached directly; r ij If it is 1, it means direct arrival.

4. The post-disaster multiple transportation means relay rescue method considering multiple terrains according to claim 1, characterized in that: For the rescue of M survivors, we first need to allocate transportation to them and determine the number of survivors Weg in each vehicle. The specific allocation strategy is as follows: Step 2.1: Determine the number of vehicles actually involved in the rescue VN R ,Right now, indivual; Step 2.2: Randomly select the number of survivors VN equal to the number of vehicles actually participating in the rescue R Assigned to each mode of transport; Step 2.3: Calculate the number of remaining survivors and determine whether the allocation is completed. If it is 0, proceed to step 2.5; if the remaining number is greater than VN R , then go to step 2.2; if less than, go to step 2.4; Step 2.4: For the remaining survivors, randomly select the means of transportation corresponding to the number of remaining survivors for allocation; Step 2.5: Count the number of survivors Weg on each vehicle i , ending the allocation.

5. The post-disaster multiple transportation means relay rescue method considering multiple terrains according to claim 1, characterized in that: In step 3), the specific steps of designing the initialization particles are as follows: Step 3.1: Classify and count the geomorphic areas on both sides of the partition line l1, and the direction of the statistics is from the starting point to the end point; Step 3.2: For the two sets obtained in step 3.1, classify and count the terrain areas on both sides of the partition line l2; Step 3.3: Use the same method to classify and count the terrain areas on both sides of the partition line l3; Step 3.4: Using the above method, the total number of cycles is max{l 左 , l 右 }-1, where l 左 represents the maximum number of final partition lines on the left of partition line l1; after the loop is completed, the terrain set obtained by statistics is rotated 90°, that is, the transposed form ZR, and the number of subsets obtained is the length of the particle, that is, U. In this step, it is also necessary to count the number of integers in each set, recorded as {Num1, Num2, Num3, ..., Num U }; The particle code is in integer form, and the kth particle is represented as: in, Decode the particles, assuming X k Any element in That is, the encoding for the jth region can be expressed as: The number of types of transportation in each area is: TN = {TN1, TN2, TN3, ..., TN N }, the number of integers in each set obtained in step 3.4 above is {Num1, Num2, Num3, ..., Num U }, the specific decoding steps are as follows: Step 4.1: From Extract elements from Execute it remainder operation; Step 4.2: According to the remainder result obtained in step 4.1, an integer corresponding to the remainder result is taken from the first rotation set ZR, thereby obtaining the number of an area passed by the vehicle; Step 4.3: Next The remaining elements in perform the operations of step 4.1 and step 4.2, so as to obtain the path of the vehicle, that is, the U areas that the vehicle needs to pass through; Step 4.4: According to the area where the vehicle has traveled obtained in step 4.3, the number of types of vehicles in the corresponding area is taken from TN and recorded as TN q ={TN q1 , T.N. q2 , T.N. q3 ,…,TN qU }; Step 4.5: From Extract elements from Execute it The result of the modulo operation is the model selected in the area. The number of corresponding models minus 1, the corresponding y i Set to 1; if it is 0, TN q1 -1; Step 4.6: From Take out the elements and execute step 4.5 to get the model used in each area; determine whether the models of all partitions are selected. If yes, go to step 4.7, otherwise, continue to execute the same operation as step 4.5; Step 4.7: Decoding is completed and the decoded result is output; Correct the infeasible solution generated by directly reaching the matrix R; assume that an element of the k-th particle That is, for the encoding of the jth area, there are two areas that cannot be reached directly. Assuming that they are decoded, the driving path obtained is represented by P j ={p j1 , p j2 ,…,p ju ,…,p jU }, assuming that the vehicle travels to area p ju After that, the transportation cannot directly reach the next area p j(u+1) , the solution is an infeasible solution, then the correction method is as follows: Step 5.1: Find the pth point in the matrix R ju row, find the region p from this row ju The area that can be directly reached, that is, the area number with all 1s, is denoted as P juR ={p juR1 , p juR2 ,…,p jRk }, Rk represents all regions p ju The number of areas reached directly; Step 5.2: For the pth ju +1 set, recorded as Search Same as in step 5.1 juR the same elements; Step 5.3: Determine whether the element taken in step 5.2 is the Uth element. If so, end the correction operation and output the corrected result. If not, perform the correction on P. j Perform the same operation as step 5.1 for the next element.

Images