Material distribution method, system, electronic device and medium

Abstract

Embodiments of the present invention relate to the technical field of emergency material distribution, and provide a material distribution method, system, electronic device, and medium. The method includes: constructing a mixed-integer programming model with a plurality of constraint information as constraint conditions according to multiple disaster-stricken points in a target area and the data of isolation points and non-isolation points at each disaster-stricken point, wherein the mixed-integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model; taking minimizing the distribution time as the goal, solving the mixed-integer programming model based on a multi-heuristic initialization-based adaptive particle swarm optimization algorithm; determining the optimal distribution paths of the multi-truck and multi-ground unmanned vehicle based on the solution result of the mixed-integer programming model; and performing material distribution on multiple disaster-stricken points in the target area and the isolation points and non-isolation points at each disaster-stricken point based on the optimal distribution paths. Thereby, efficient path planning and resource optimization scheduling are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of emergency material distribution, and particularly to a method, a system, an electronic device and a medium for material distribution. Background Art

[0002] In recent years, with the increase of public health events worldwide, the field of emergency material distribution has received great attention and become the focus of attention from all sectors of society. In public health events, there are multiple disaster-stricken points in the city where emergency materials need to be distributed. In this context, as a new type of logistics distribution tool, the ground unmanned vehicle distribution has the characteristics of "contactless" distribution, with a large carrying capacity and a long driving distance. It is suitable for distributing emergency materials in a more secure and complex urban traffic environment, reducing the risk of virus transmission, and showing significant advantages in public health events.

[0003] However, in the collaborative distribution scenario of ground unmanned vehicles and trucks, how to efficiently schedule trucks carrying multiple ground unmanned vehicles to achieve optimal resource allocation has become a key challenge. In particular, problems such as the launch and recovery of ground unmanned vehicles at different positions, the task allocation and path optimization of multiple distribution points cannot be effectively solved by the existing technologies. Summary of the Invention

[0004] The present invention provides a method, a system, an electronic device and a medium for material distribution, which are used to solve the defect that trucks and ground unmanned vehicles cannot be efficiently scheduled in the existing technology, and to achieve efficient path planning and resource optimization scheduling.

[0005] The present invention provides a method for material distribution, including:

[0006] Constructing a mixed integer programming model with a plurality of preset constraint information as constraint conditions according to a plurality of disaster-stricken points in the target area and the data of isolation points and non-isolation points at each disaster-stricken point, wherein the mixed integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model;

[0007] Solving the mixed integer programming model based on a multi-heuristic initialization-based adaptive particle swarm optimization algorithm with the goal of minimizing the distribution time;

[0008] Determining the optimal distribution paths of the multi-truck and multi-ground unmanned vehicle based on the solution result of the mixed integer programming model;

[0009] Performing material distribution on a plurality of disaster-stricken points in the target area and the isolation points and non-isolation points at each disaster-stricken point based on the optimal distribution paths.

[0010] In a possible implementation manner, the method further includes:

[0011] The multiple constraint information includes: path constraint, load constraint, mileage constraint, and collaborative operation time constraint;

[0012] Define decision variables and objective variables according to multiple disaster-stricken points in the target area and the data of isolated and non-isolated points at each disaster-stricken point. Among them, the decision variables include path selection variables, release and recovery node selection variables, and the objective variables include the total time for each truck to complete the distribution task, and the total time includes its own distribution operation time and collaborative operation time;

[0013] Based on the decision variables and objective variables, and using the multiple constraint information as constraint conditions, construct a mixed-integer programming model.

[0014] In a possible implementation manner, the method further includes:

[0015] Use the greedy heuristic method, the nearest neighbor heuristic method, and the stochastic nearest neighbor heuristic method to generate initial solutions of the mixed-integer programming model respectively, and combine the initial solutions according to a preset ratio to generate an initial population;

[0016] Optimize the particle swarm optimization algorithm by dynamically adjusting the inertia weight, adaptive learning factor, tournament selection, and velocity pruning mechanism to obtain the multi-heuristic initialization-based adaptive particle swarm optimization algorithm;

[0017] Based on the multi-heuristic initialization-based adaptive particle swarm optimization algorithm, solve the mixed-integer programming model and output the global optimal solution.

[0018] In a possible implementation manner, the method further includes:

[0019] Use the greedy heuristic method based on the greedy strategy. In each step of path selection, select the customer point that is the nearest to the current node and satisfies the constraint conditions as the next access node. Starting from the distribution center, repeat this process until no new customer points can be added to obtain the first initial solution;

[0020] Use the nearest neighbor heuristic method to select the next access point according to multiple factors to obtain the second initial solution, where the multiple factors include distance factor, convenience of node connection, and potential influencing factors for subsequent path planning;

[0021] Use the stochastic nearest neighbor heuristic method to randomly select a candidate node from the set of adjacent nodes that satisfy the constraint conditions, and continue to expand the path based on the candidate node to obtain the third initial solution;

[0022] Combine the first initial solution, the second initial solution, and the third initial solution according to a preset ratio to generate an initial population.

[0023] In a possible implementation, the method further includes:

[0024] Dynamically adjusting the inertia weight by adopting a linearly decreasing weight strategy;

[0025] Increasing the change range of the learning probability by using an exponential function, determining the learning probability of each particle, and using the learning probability of each particle as the adaptive learning factor;

[0026] Randomly selecting two or more particles for comparison, and selecting the better particles as candidate particles for position update;

[0027] Randomly pruning some velocity components, controlling the velocity change of the particles within a preset range and maintaining the movement direction, to obtain the adaptive particle swarm optimization algorithm.

[0028] In a possible implementation, the method further includes:

[0029] Verifying the solution result of the mixed integer programming model based on the material distribution result.

[0030] The present invention also provides a material distribution system, including the following modules:

[0031] A model construction module, configured to construct a mixed integer programming model with a plurality of disaster-stricken points in a target area and the data of isolation points and non-isolation points of each disaster-stricken point as constraint conditions, where the mixed integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model;

[0032] A model solution module, configured to solve the mixed integer programming model based on the multi-heuristic initialization-based adaptive particle swarm optimization algorithm with the goal of minimizing the distribution time;

[0033] A path determination module, configured to determine the optimal distribution path of the multi-truck and multi-ground unmanned vehicle based on the solution result of the mixed integer programming model;

[0034] A material distribution module, configured to perform material distribution on the plurality of disaster-stricken points in the target area and the isolation points and non-isolation points of each disaster-stricken point based on the optimal distribution path.

[0035] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the material distribution method as described in any one of the above when executing the computer program.

[0036] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, and the computer program implements the material distribution method as described in any one of the above when executed by a processor.

[0037] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the material distribution method described in any one of the above.

[0038] For the material distribution method, system, electronic device and medium provided by the present invention, by using the data of multiple disaster-affected points in the target area and the isolated points and non-isolated points of each disaster-affected point, a mixed integer programming model is constructed with a plurality of preset constraint information as constraint conditions, wherein the mixed integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model; aiming at minimizing the distribution time, the mixed integer programming model is solved based on the multi-heuristic initialization adaptive particle swarm optimization algorithm; based on the solution result of the mixed integer programming model, the optimal distribution paths of the multi-truck and multi-ground unmanned vehicle are determined; and based on the optimal distribution paths, materials are distributed to the multiple disaster-affected points in the target area and the isolated points and non-isolated points of each disaster-affected point. Compared with the defect in the prior art that trucks and ground unmanned vehicles cannot be efficiently scheduled, in this solution, the accessibility of trucks and the flexibility of ground unmanned vehicles in the emergency scenario are comprehensively considered, a multi-truck and multi-ground unmanned vehicle collaborative distribution model is constructed with the goal of reducing the total distribution time of all trucks, and the optimal distribution path is obtained by solving the model based on the multi-heuristic initialization adaptive particle swarm optimization algorithm, realizing efficient path planning and resource optimal scheduling. Description of the Drawings

[0039] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0040] Figure 1 is a schematic flowchart of the material distribution method provided by the present invention.

[0041] Figure 2 is a schematic diagram of the truck-ground unmanned vehicle collaborative distribution mode provided by the present invention.

[0042] Figure 3 is a schematic diagram of the path generation process provided by the present invention.

[0043] Figure 4 is a schematic diagram of the influence of the learning factor on the heuristic algorithm provided by the present invention.

[0044] Figure 5 is a flowchart of the adaptive particle swarm optimization algorithm initialized based on multiple heuristic algorithms provided by the present invention.

[0045] Figure 6 It is the collaborative distribution diagram of the truck and the ground unmanned vehicle provided by the present invention.

[0046] Figure 7 It is the schematic diagram of the result analysis of different truck load capacities provided by the present invention.

[0047] Figure 8 It is the schematic diagram of the result analysis of different ground unmanned vehicle load capacities provided by the present invention.

[0048] Figure 9 It is the schematic diagram of the result analysis of different ground unmanned vehicle mileage provided by the present invention.

[0049] Figure 10 It is the schematic structural diagram of the material distribution system provided by the present invention.

[0050] Figure 11 It is the schematic structural diagram of the electronic device provided by the present invention. Detailed implementation manners

[0051] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0052] To facilitate the understanding of the embodiments of the present invention, the following will further explain and illustrate with specific embodiments in conjunction with the accompanying drawings. The embodiments do not constitute a limitation to the embodiments of the present invention.

[0053] Figure 1 It is the schematic flowchart of the material distribution method provided by the present invention. As Figure 1 shown, the method includes the following:

[0054] S11. According to multiple disaster-stricken points in the target area and the data of isolation points and non-isolation points of each disaster-stricken point, a mixed integer programming model is constructed with a plurality of preset constraint information as constraint conditions.

[0055] In the embodiments of the present invention, for example, there are multiple disaster-stricken points in a certain city where emergency supplies need to be distributed. There is an emergency supply distribution center in this city, and there are multiple human-driven trucks and driverless ground unmanned vehicles in the distribution center for material distribution. This method mainly studies how to plan the paths of distribution vehicles so that they can meet the material distribution needs in the disaster-stricken area and achieve the goal of the shortest total distribution time of all trucks.

[0056] First of all, it is necessary to construct a mixed integer programming model, which is a multi-truck and multi-ground unmanned vehicle collaborative distribution model.

[0057] Specifically, the affected points in the city are allocated by both trucks and ground unmanned vehicles. The trucks can also serve as mobile warehouses for the launch and recovery of ground unmanned vehicles. The affected points are divided into isolated points and non-isolated points. The former are distributed by ground unmanned vehicles, and the latter are distributed by trucks. During the distribution operation, the truck carries multiple ground unmanned vehicles from the distribution center and sequentially travels to the affected points or temporary parking points according to the established route, delivers supplies to the affected points, and launches or recovers ground unmanned vehicles. As Figure 2 shown 1 The truck sequentially travels to non-isolated points C1 - C5 - C8 according to the predetermined route and launches ground unmanned vehicles D1 and D2 at C1. After the ground unmanned vehicles are launched, they sequentially travel to the affected points near C1 for material replenishment according to the established routes (C1 - C2 - C3 and C1 - C4). After the distribution task is completed, they meet the truck at the next non-isolated point for material loading and battery replacement, or directly return to the distribution center (such as the route of ground unmanned vehicle D4 from C11 - C13 - distribution center). Each ground unmanned vehicle can only execute one distribution task, but each distribution task can serve multiple affected points. When the ground unmanned vehicle returns to the distribution center, no more material distribution is carried out.

[0058] Considering the technical characteristics and operating conditions of ground unmanned vehicles, the following assumptions are made:

[0059] (1) Trucks and ground unmanned vehicles can access multiple disaster areas, and each disaster area is served by one or only one truck or ground unmanned vehicle for distribution. To reduce human contact and the risk of virus transmission, trucks can only serve non-isolated points, and ground unmanned vehicles can only serve isolated points.

[0060] (2) The locations and demands of each affected point are known.

[0061] (3) The driving speed of the truck and the number of ground unmanned vehicles it can carry are known. The truck has no driving distance limit but has a load limit.

[0062] (4) The driving speed, load limit, and mileage limit of the ground unmanned vehicle are known.

[0063] (5) Ground unmanned vehicles can be launched from trucks or start directly from the warehouse to serve the disaster area directly.

[0064] (6) The time for the truck and the ground unmanned vehicle to deliver materials after arriving at the affected location is known, and the time for the ground unmanned vehicle to deliver and recover operations is known. The truck can simultaneously recover and launch multiple ground unmanned vehicles.

[0065] (7) Each ground unmanned vehicle consumes a certain amount of energy for each trip. When the ground unmanned vehicle starts a new journey, it will be replaced by a fully charged battery.

[0066] (8) When the truck serves non-isolated points, ground unmanned vehicles can be launched simultaneously, and the launch time is less than the service time.

[0067] (9) Understand the urban road traffic conditions.

[0068] Further, define decision variables and objective variables according to multiple disaster-stricken points in the target area and the data of isolation points and non-isolation points for each disaster-stricken point. Among them, the decision variables include path selection variables, release and recovery node selection variables, and the objective variable includes the total time for each truck to complete the distribution task. The total time includes its own distribution operation time and collaborative operation time. Based on the decision variables and objective variables, and using multiple constraint information as constraint conditions, a mixed-integer programming model is constructed.

[0069] The names, meanings, and types of each variable are shown in Table 1:

[0070]

[0071] Table 1 Variable names, meanings, and types

[0072]

[0073] Table 1 Variable names, meanings, and types

[0074] The urban emergency material distribution network is represented by a graph where is the set of all nodes, is the set of all arcs, . The starting node is and the ending node is representing the same distribution center. represents the set of all disaster-stricken points, where the set of disaster-stricken points allowed for truck distribution is . Since the truck can reach limited customer points due to actual traffic conditions and infection risk restrictions, . represents the set of nodes where unmanned vehicles may be released, represents the set of nodes where unmanned vehicles may be recovered, , .

[0075] represents the set of all trucks, represents the set of all unmanned vehicles. The times for the truck and the unmanned vehicle to serve the disaster-stricken point are and respectively. Each release and recovery of the unmanned vehicle requires a certain operation time, which are respectively recorded as and WThe time point when the truck arrives is recorded as , and the time point when the driverless vehicle arrives at is . The departure time of the truck from is recorded as , and the departure time of the driverless vehicle from is recorded as . .

[0076] Model construction includes: the complete operation time of each truck which consists of two parts: the time to complete its own distribution task and the time to cooperate with the driverless vehicle. The specific definitions of the operation time are as follows:

[0077] 1) Definition of distribution operation time: Denote as the cumulative time for the truck to complete its own distribution task, including the driving time between disaster-stricken nodes and the service time at disaster-stricken points.

[0078] (1)

[0079] 2) Definition of truck-driverless vehicle cooperation operation time: Denote as the cooperation operation time generated by the truck and the driverless vehicle at point due to mutual waiting, release, and recovery. At any point , the cooperation between the truck and the driverless vehicle is divided into two parts: the waiting time required to recover the previously dispatched driverless vehicle and the time required to dispatch the current driverless vehicle.

[0080] Recovery waiting time: For the driverless vehicle paired with the truck d , if the truck arrives at point with the driverless vehicle on board, the recovery waiting time is recorded as = 0. If the driverless vehicle d and the truck do not arrive at point together, then it is divided into two cases: (1) The truck arrives at the recovery point first, that is . Then, after the truck completes its distribution task at point, it waits until the driverless vehicle arrives and then performs the recovery operation. The time consumed in this process is recorded as ; (2) The driverless vehicle arrives at the recovery point first, that is . Then the time consumed is the recovery operation time of the driverless vehicle, .

[0081] Then the truck is at The waiting time for recycling all unmanned vehicles is: .

[0082] Release waiting time: Trucks in When the drone needs to be released at a certain point, the time consumed is the total time required to release the drone at that release point. .

[0083] Therefore, the truck The total collaborative operation time of the points is: .

[0084] For any truck The time from the warehouse to the warehouse is recorded as , is the sum of the time required for the truck to complete its own delivery task and the time required for the unmanned vehicle to work together. Therefore, the operating time of each truck in this problem is defined as

[0085] (2)

[0086] The constraints include the following:

[0087] (3)

[0088] (4)

[0089] (5)

[0090] (6)

[0091] (7)

[0092] (8)

[0093] (9)

[0094] (10)

[0095] (11)

[0096] (12)

[0097] (13)

[0098] (14)

[0099] (15)

[0100] (16)

[0101] (17)

[0102] If , then (18)

[0103] (19)

[0104] (20)

[0105] (21)

[0106] (22)

[0107] (23)

[0108] (24)

[0109] (25)

[0110] (26)

[0111] (27)

[0112] The objective function (3) minimizes the maximum delay time for trucks or drones to return to the warehouse. Constraint (4) stipulates that all truck trips must start and end at the warehouse. Constraints (5) and (6) stipulate that each disaster-stricken point can only be served by a truck or a drone alone once. Constraints (7) and (8) indicate that trucks and drones must leave a node after entering it during the distribution process. Constraint (9) indicates that drones must depart from the release node and return to the recovery node, and each drone is released only once.

[0113] Constraint (10) indicates that if a truck serves a disaster-stricken point , then the truck must enter that point once. Constraint (11) indicates that if a truck serves a disaster-stricken point , then the truck must leave that point once. Similarly, Constraint (12) indicates that if a drone d serves a point, then there is and only one drone d arriving at the point, passing through the arc ([[]] , ). Constraint (13) indicates that if a drone d serves point, then there is exactly one driverless vehicle d leaving arc point, passing through arc( , ). Constraint (14) means point is selected as the release point of the driverless vehicle carried by the truck d , the driverless vehicle d starts from point. Constraint (15) means point is selected as the recovery point of the driverless vehicle carried by the truck d , the driverless vehicle d has to return to point. Constraints (16) and (17) mean that the release and recovery points of the driverless vehicle are on the corresponding truck path nodes.

[0114] Constraint (18) restricts that the truck d carrying the driverless vehicle should first reach the release node of the driverless vehicle d , and then reach the recovery node of this driverless vehicle. Constraint (19) represents the recovery waiting time of the truck at node . Constraint (20) represents the time relationship between the arrival and departure of the driverless vehicle at the disaster - affected node. If the driverless vehicle passes through arc( , ), then it must first reach and then reach , and the time difference must be greater than or equal to the time for the driverless vehicle to serve point and the time for passing through arc( , ). Similarly, constraint (21) represents the time relationship between the arrival and departure of the truck at the node. If the truck passes through arc( , ), then it must first reach and then reach , and the time difference must be greater than or equal to the time for the truck to serve point and the time for passing through arc( , ) Also, add the waiting time for the truck to recycle and release the driverless vehicle. Constraint (22) represents the time relationship between the arrival time of the driverless vehicle at the node after release and the arrival time of the truck at the release node. Constraint (23) represents the load limit of the driverless vehicle. Constraints (24) and (25) represent the load limit of the truck and the mileage limit of the driverless vehicle. Constraints (26) and (27) represent the value range limits of the parameters and variables.

[0115] S12. Aiming at minimizing the distribution time, solve the mixed integer programming model by using an adaptive particle swarm optimization algorithm based on multi - heuristic initialization.

[0116] Use the greedy heuristic method, the nearest neighbor heuristic method and the stochastic nearest neighbor heuristic method to generate the initial solutions of the mixed integer programming model respectively, and combine the initial solutions according to a preset ratio to generate an initial population; optimize the particle swarm algorithm through dynamically adjusting the inertia weight, the adaptive learning factor, the tournament selection and the velocity pruning mechanism to obtain the adaptive particle swarm optimization algorithm based on multi - heuristic initialization; solve the mixed integer programming model based on the adaptive particle swarm optimization algorithm based on multi - heuristic initialization, and output the global optimal solution.

[0117] Specifically, in practice, although mathematical programming tools such as Gurobi can provide the global optimal solution, when dealing with large - scale or complex examples, Gurobi will verify whether each solution is the global optimal solution, which may face the problem of too long convergence time. Especially in the scenarios of dynamic change or real - time scheduling, it is difficult to meet the efficiency requirements. As a heuristic algorithm based on swarm intelligence, the Particle Swarm Optimization (PSO) algorithm has the characteristics of simple iterative process and low computational cost, and can obtain an approximate optimal solution within a reasonable time. Selecting the PSO algorithm as the method to solve the collaborative distribution problem of trucks and ground driverless vehicles is mainly based on its advantages in dealing with complex optimization problems.

[0118] As an optimization strategy inspired by the swarm intelligence in nature, the design concept of PSO is inspired by the collective action patterns of biological groups such as birds or fish. This method simulates the interaction and cooperation mechanism among individuals in the group to explore the optimal solution of the problem. In the algorithm framework, each potential solution is graphically defined as a "particle", which has two core attributes: position coordinates and moving speed, jointly guiding the particle to explore and update in the solution space. In each iteration of the algorithm, the particle updates its position according to its own position and speed, and is also affected by other particles in the group. After a period of search, the flock of birds can find the position with the largest amount of food, which is the global optimal solution.

[0119] For the problem of optimizing the collaborative distribution path of trucks and ground unmanned vehicles, the main idea is to unify first and then separate, and solve it through a combination of two stages. In the first stage, the whole process is regarded as a Vehicle Routing Problem (VRP); first, separate the isolation points and non-isolation points, take the non-isolation points as the truck paths, and the isolation points as the ground unmanned vehicle paths; in the second stage, merge the two paths, and merge the ground unmanned vehicle paths on the basis of the original truck paths to obtain all possible truck and ground unmanned vehicle paths. In the third stage, continuously improve the iterative search for possible results to find the optimal truck and ground unmanned vehicle paths and construct the VRP-D solution. One particle is the optimal path combination, including the truck and ground unmanned vehicle paths; the objective function is to minimize the time. A particle contains the position of a particle, the velocity of a particle, and a set of calculated paths. Calculate the set of each particle, then compare the optimality, and continuously iterate this process.

[0120] In Figure 3 , it is assumed that C1, C2, C3, C4 are non-isolation points, and C5, C6 are non-isolation points. In the first-stage optimization, 0-C1-C2-C3-C4-0 is the truck path, and 0-C5-0 is the ground unmanned vehicle path. In the second stage, merge the paths to find all possible paths. In the third stage, find the optimal result from all possible situations. The ground unmanned vehicle path changes from 0-C5-0 to C4-C5-C6 -0, launches from C4, and retrieves at C1. In the calculation of the distribution time, since these trucks are in parallel service, after optimizing each truck path, find the vehicle with the longest service time, which is the distribution time of the whole event.

[0121] Aiming at the disadvantage that the traditional PSO algorithm is prone to falling into the local optimal solution, introduce nearest neighbor initialization, dynamic inertia weight, adaptive learning factor and velocity pruning to optimize the algorithm, and propose an improved adaptive inertia weight particle swarm optimization algorithm.

[0122] In order to construct a high-quality initial population, three heuristic algorithms with unique characteristics are used for comprehensive initialization, aiming to provide diverse and potential initial solutions for the subsequent optimization process.

[0123] Greedy heuristic initialization: This algorithm is based on the greedy strategy. In each step of selection, the customer point that is closest to the current node and satisfies the time and capacity constraints is given priority as the next access node. Starting from the distribution center (warehouse), this process is repeated until no new customer points can be added, thus establishing a complete truck route. The advantage of this method is its fast convergence speed, and it can generate a relatively optimal route in a short time. Because it always selects locally optimal nodes, the total distance of the route can be optimized to a certain extent in the initial stage. For example, in areas where customers are relatively concentrated, the greedy strategy can quickly find a compact route, reduce unnecessary transfers, and thus introduce a locally optimal route pattern into the initial population, providing a good starting point for subsequent evolution.

[0124] Nearest neighbor heuristic algorithm: Similar to the greedy nearest neighbor heuristic algorithm, the nearest neighbor heuristic algorithm also takes finding neighbor nodes as the core idea. However, when selecting the next node, not only the distance factor is considered, but also other factors such as the convenience of node connection and the potential impact on subsequent route planning are comprehensively considered, and a relatively flexible selection mechanism is used to determine the next access point. This method not only ensures a certain route efficiency but also increases the diversity of the route. It can explore some route possibilities that the greedy algorithm may ignore, and avoid the initial population being overly concentrated near a single local optimal solution.

[0125] Random neighborhood heuristic initialization: The random neighborhood heuristic algorithm constructs the route by randomly selecting adjacent nodes, completely breaking the deterministic selection mode. Each time the next node is selected, a candidate node is randomly selected from the set of adjacent nodes that satisfy certain constraints, and then the route is continued to be extended based on this node. This highly random method greatly increases the diversity of the initial population and makes it possible to explore solution space regions that are difficult for traditional deterministic algorithms to reach. Especially in complex distribution scenarios with many uncertain factors and unpredictable situations, the random neighborhood heuristic algorithm may discover some unexpected effective route combinations.

[0126] Constructing the initial solution by combining three heuristic methods: The initialization uses the greedy nearest neighbor heuristic, the nearest neighbor heuristic, and the random nearest neighbor heuristic to generate the initial solution. Each method generates multiple routes under constraints such as time and capacity, and parallel computing is used to improve the efficiency of solution generation. The results of different methods are combined to form the initial population. A data structure is constructed to store information such as the route list, route distance, particle position, and speed. At the same time, the fitness value is calculated based on the number of trucks and the route distance, and finally the initial population and the fitness value table are returned to provide a diverse solution set for the subsequent optimization process.

[0127] Three heuristic algorithms are assigned to participate in the generation of the initial population. The total population size M is divided into three parts, and the number of particles generated by the greedy nearest neighbor heuristic algorithm can be directly specified. For the remaining number of particles, the ratio between the nearest neighbor heuristic algorithm and the random nearest neighbor heuristic algorithm is controlled by a random value phi. In the randomly generated array, the number of particles less than phi is determined as the number of initial particles generated by the nearest neighbor heuristic algorithm, and the rest are the number of initial particles generated by the random nearest neighbor heuristic algorithm. Through this dynamic proportional allocation mechanism, the greedy algorithm ensures the existence of high-quality solutions in the population, while the nearest neighbor and random algorithms enhance the diversity of the solutions. Through this method, the algorithm takes into account both the diversity of knowledge and the initial quality, laying a good foundation for the convergence of the optimization algorithm.

[0128] By organically combining these three heuristic algorithms and generating the initial population according to a preset ratio or strategy, the initial population contains both local optimal path patterns and rich diversity, laying a solid foundation for the subsequent iterative optimization process and increasing the possibility of finding the global optimal solution.

[0129] In the PSO algorithm, the inertia weight is used to control the velocity update of particles, directly affecting the balance between the exploration ability and exploitation ability of the algorithm. The adjustment of the inertia weight significantly affects the search performance of the algorithm. The linear decreasing weight strategy proposed by Shi and Y is adopted, and the inertia weight is dynamically adjusted according to Equation (28), where the k inertia weight of the generation. The initial inertia weight is usually larger (such as 0.9) to enhance the initial global exploration ability, and the final inertia weight k is usually smaller (such as 0.4) for local exploitation in the later stage of convergence. represents the current iteration number, and max_gen is the maximum number of iterations. In the initial stage of the search, a larger inertia weight promotes the particles to conduct global exploration at a faster speed, which can better cover the entire search space and avoid falling into local optima. In the later stage, a smaller inertia weight causes the particle velocity to decrease, which is beneficial to conduct detailed development and optimization near the known optimal solution, thereby improving the solution accuracy. As the number of evolutionary generations increases,

[0130] (28)

[0131] The individual learning factor c1 and the social learning factor c2 are usually between (0, 2), used to balance the influence of the particle's self-experience and group experience on its velocity, such as Figure 4As shown. A larger learning factor makes the particles pay more attention to the optimal position of the group, while a smaller learning factor makes the particles pay more attention to their own optimal positions. In the PSO algorithm, the particles update their positions based on their own historical optimal positions (pbest) and the global optimal position (gbest) of the group. In standard PSO, all particles are updated according to the same rules and parameters. This consistency sometimes causes all particles to concentrate near a certain solution prematurely, lacking diversity, and thus being prone to falling into local optima.

[0132] The adaptive setting of the learning probability adopts different learning strategies for different particles. Using the nonlinear function in Equation (29), the exponential function is used to increase the variation range of the learning probability, and the learning probability of each particle is calculated, where i is the number of particles, M is the total number of particles, P is the learning probability of each particle. This probability is used to determine whether the particle should use its pbest or update its best position through other mechanisms. For particles with a smaller number, their learning probability is lower. For particles with a larger number, the learning probability is higher. P The higher the value of, the more likely the particle is to learn and retain its own experience (i.e., use pbest); P Particles with a lower value may choose different strategies, such as updating the optimal position through tournament selection, which means they are more likely to explore new solutions. The range of the learning probability has a minimum value of 0.05, has a maximum value of 0.5, indicating that the minimum learning probability is 5% and the maximum is 50%. This ensures that different particles have different exploration capabilities. Under such a design, the change of the learning probability shows a trend from low to high, that is, particles with a larger number are more inclined to retain their historical experience, while particles with a smaller number are more likely to explore new solutions.

[0133] (29)

[0134] In the standard PSO algorithm, the position update of particles is usually determined by their own pbest and gbest. However, this update mechanism may cause all particles to tend to the same optimal point, resulting in problems such as reduced population diversity and premature convergence. To increase the diversity of solutions and reduce the probability of falling into local optima, the tournament selection mechanism is used to update the pbest of particles; this selection mechanism randomly selects two or more particles for comparison, and selects the better particle as a candidate particle for position update or optimal position update.

[0135] A particle with a different number from the current particle is randomly selected as a competitor through the tournament function, ensuring that the number is different from the current particle. Let the total number of the particle swarm be m, for each particle , a different particle needs to be selected for tournament selection. The tournament selection process can be expressed as Equation (30). Among them is the number of selected particles, rand(1, M ) represents randomly selecting a particle from 1 to M .

[0136] (30)

[0137] In the PSO algorithm, the velocity vector determines the direction and amplitude of the particle's next movement. Usually, the PSO algorithm updates the velocity according to the pbest and gbest of each particle. However, as the iteration progresses, if the particle velocity is too large, some detailed solutions may be missed, resulting in the inability to well develop these solution regions. But if the velocity is too small, the convergence speed may be too slow. The velocity pruning mechanism is used to randomly prune some velocity components, so that the particle can maintain the movement direction with little change in velocity, thus achieving a balance between exploration and exploitation. First, a velocity matrix is randomly generated, and a pruning threshold matrix is defined; if the random value is less than the pruning threshold, the velocity is set to 0, otherwise the original velocity value is maintained. If some dimensions of the particle velocity are very large, it may cause the particle to leave the effective solution space and even oscillate back and forth, resulting in a decrease in efficiency. Through velocity pruning, the particle can more stably stay in the effective solution space and gradually approach the optimal solution.

[0138] Any particle i will synchronously update its velocity and position according to two preset formulas to update the optimization process of the entire search space:

[0139] (31)

[0140] (32)

[0141] Among them c1 and c2 are learning factors, usually c1 = c2 = 2, rand () is a random number between (0, 1). According to Equation (31) and Equation (32), the standard form of PSO is:

[0142] (33)

[0143] Among them is the inertia factor, and its value is non - negative. In practical applications, the selection of the learning factor and the inertia coefficient has an important impact on the performance of the PSO algorithm. Usually, it is necessary to determine the optimal values of the learning factor and the inertia coefficient through experiments and parameter adjustments to obtain better optimization effects. The pseudo - code for initializing the adaptive particle swarm optimization algorithm based on multiple heuristic algorithms is shown in Table 2, and the flowchart for initializing the adaptive particle swarm optimization algorithm based on multiple heuristic algorithms is as Figure 5 shown.

[0144]

[0145] Table 2 Pseudo - code for initializing the adaptive particle swarm optimization algorithm based on multiple heuristic algorithms

[0146]

[0147] Table 2 Pseudo - code for initializing the adaptive particle swarm optimization algorithm based on multiple heuristic algorithms

[0148] S13. Determine the optimal distribution path of the multiple trucks and multiple ground unmanned vehicles based on the solution results of the mixed - integer programming model.

[0149] S14. Conduct material distribution for multiple disaster - affected points in the target area and the isolation points and non - isolation points of each disaster - affected point based on the optimal distribution path.

[0150] Through the solution of the above algorithm, the optimal distribution paths of multiple trucks and multiple ground unmanned vehicles are obtained. These paths include:

[0151] The distribution path of the truck, starting from the distribution center, visiting non - isolation points in sequence, and simultaneously launching and retrieving ground unmanned vehicles.

[0152] The distribution path of the ground unmanned vehicle, starting from the truck or the distribution center, visiting isolation points, and returning after completing the distribution task.

[0153] Distribution execution:

[0154] Truck distribution: The truck goes to non - isolation points in sequence according to the planned path for material distribution. At non - isolation points, the truck can simultaneously launch ground unmanned vehicles to go to nearby isolation points for distribution. After the truck completes the distribution task, it returns to the distribution center.

[0155] Ground unmanned vehicle distribution: The ground unmanned vehicle starts from the truck or the distribution center and goes to isolation points for material distribution. After the ground unmanned vehicle completes the distribution task, it returns to the truck or the distribution center for battery replacement and material loading. During the distribution process, the ground unmanned vehicle visits multiple isolation points in sequence according to the planned path.

[0156] Cooperative operation: The truck and the ground-based autonomous vehicle maintain cooperative operation during the distribution process. When the truck stops at a non-isolated point, it can launch or retrieve the ground-based autonomous vehicle. The distribution routes of the truck and the ground-based autonomous vehicle are optimized to ensure the delivery of supplies to all disaster-stricken points in the shortest possible time.

[0157] Example verification:

[0158] Adaptability and advantages of APSO-MHI in dealing with complex scenarios: The APSO-MHI algorithm effectively improves the exploration and exploitation capabilities of the particle swarm optimization algorithm by introducing three heuristic algorithms to generate initial solutions, dynamically adjusting the inertia weight, probability learning factor, tournament selection method, and velocity pruning mechanism. Constructing the initial solution by combining three heuristic algorithms provides a high-quality initial solution for PSO and reduces the randomness of early convergence. By dynamically adjusting the inertia weight, the probabilities of individual learning and swarm learning are balanced, improving the adaptability of the algorithm and enabling more efficient exploration and utilization in the search space. The tournament selection method strengthens the competition mechanism among particles, resulting in a higher retention probability of excellent solutions and avoiding the loss of high-quality solutions. The velocity pruning mechanism limits the excessive transition of particles and improves the accuracy and stability of the solution. The APSO-MHI algorithm is applicable to scenarios with complex search spaces and multi-objective optimization. Compared with the traditional PSO algorithm, the improved algorithm not only improves the convergence speed and optimization accuracy but also enhances the robustness, especially in dynamic or high-dimensional problems.

[0159] Taking the Solomon dataset as an example, this dataset is a relatively classic dataset for studying VRRP-related problems. The instance definitions and pointers of the 25- and 50-customer instances of the 1987 Solomon VRPTW benchmark problem's most famous solutions can be found on Solomon's official website. The Solomon standard test data has one starting point (CUST NO. == 0) and 100 customer points. In the Solomon dataset, different subsets represent different types of problems, where R1, R2, C1, C2, RC1, RC2 are the identities of these subsets. R represents a randomly generated dataset, C represents a cluster-generated dataset, and RC represents both a random and a cluster dataset, including datasets of different sizes, customer demands, and density levels. K represents the maximum number of schedulable vehicles, Q represents the maximum load per vehicle, XCOORD and YCOORF are the horizontal and vertical coordinates of the starting point and customer points. For computational convenience, the distance between nodes is used as the transportation cost between nodes. The demand is the demand of that node, and the demand of the starting point depot is 0. The service time is the duration of service for each node. Among the 25 disaster-stricken points, 8 points are randomly selected as isolated points. Among the 50 disaster-stricken points, 16 points are randomly selected as isolated points, which can only be reached by the ground-based autonomous vehicle.

[0160] APSO-MHI implements algorithm programming and operations using the Python language on the Pycharm platform. All numerical experiments were conducted on a computer with an AMD Ryzen 7-4800H, which has a Radeon graphics card with 2.90 GHz and 16 GB of memory. The optimization solver used was the GUROBI Optimizer 10.0, running under the Windows 11 operating system. The results of running the algorithm 10 times were used as the optimal value. The parameter settings of this algorithm are as follows: the population size M = 100, the inertia weight = 0.9 = 0.4, and the learning factor c = 2. Gurobi 10.0.1 was called to conduct experimental verification on the APSO-MHI algorithm, and the results are shown in Table 1. In all case trials, the full-load speed of the truck is 0.67 km / min, the full-load speed of the ground-based unmanned vehicle is 0.42 km / min, the full-load distance limit of the ground-based unmanned vehicle is 60 km, the load limit of the truck is 500 kg, and the load limit of the ground-based unmanned vehicle is 200 kg.

[0161] APSO-MH numerical examples and analysis: The running results are shown in Table 7, which presents the performance comparison of the Gurobi solver, the APSO-MHI algorithm proposed in this method, and PSO on different datasets. The "Dataset" column lists the tested datasets; DT (min) represents the optimal distribution time of the dataset; "RT (s)" represents the total running time of the program. Different symbols are used to distinguish various usages of "GAP (%)": among them, "GAP1 (%)" in the Gurobi column represents the relative gap between the currently found solution and the optimal solution of the problem, "GAP2 (%)" in the APSO-MHI column represents the error percentage between the optimal solution obtained by the APSO-MHI algorithm and the optimal solution obtained by the Gurobi solver, and "GAP3 (%)" in the PSO column represents the error percentage between the optimal solution obtained by the PSO algorithm and the optimal solution obtained by the Gurobi solver.

[0162] As can be seen from Table 3, the performance differences of the three methods on datasets of different sizes and types are significant. On the dataset containing 25 disaster sites, GAP1 (%) is relatively small, indicating that its solution is close to the optimal. For large-scale cases, Gurobi cannot obtain a feasible solution or an optimal solution within 3600 s. Since the VRP problem is essentially an NP-hard problem, as the number of disaster sites increases, the problem scale and the number of variables increase sharply, making it difficult for the solver to complete the calculations of branching, bounding, and cutting planes within a reasonable time, and thus unable to return a solution within a limited time. This shows that although the exact method can provide a theoretical guarantee of the global optimal solution, it consumes too much computing resources and has limited solving efficiency when solving large-scale problems, and is not suitable for emergency logistics scenarios that require quick decision-making.

[0163] On a dataset containing 25 disaster sites, the GAP2 (%) value ensures a balance between solution quality and computational power within the 5% threshold, providing a reliable performance benchmark comparison sufficient to prove the effectiveness of the APSO MHI results. APSO-MHI has significant adaptability and comprehensive advantages in complex scenarios, mainly because its strategy combines multiple heuristic initialization methods, effectively enhancing the diversity of the population and the quality of the initial solution, enabling the algorithm to quickly converge to a better solution. Compared with ordinary PSO, APSO-MHI significantly improves the optimization effect of the delivery time, has better adaptability to complex distributed datasets, and significantly reduces GAP3 (%). At the same time, compared with the exact algorithm of Gurobi, the running time of APSO-MHI is only 10% - 15% of that of Gurobi, and it can provide a solution close to the global optimum. This efficiency is very suitable for scenarios with high requirements for real-time and response speed, such as emergency logistics. Therefore, APSO-MHI achieves a good balance between solution quality and running efficiency and is a robust and efficient algorithm for dealing with complex task scenarios.

[0164]

[0165] Table 3 Comparison of test results between GUROBI solver and APSO-MHI

[0166] Case verification and analysis

[0167] Taking the distribution stations in Qingbaijiang District, Xindu District, Chenghua District, etc. of Chengdu City, Sichuan Province as examples, the collaborative path problem of trucks and ground unmanned vehicles is deeply optimized. To ensure the feasibility and practicality of the research, a method combining real data and simulation is adopted in parameter settings. Specifically, the case study in this paper includes 1 urban distribution center and 39 disaster areas. The urban distribution center is numbered 0, and the 39 disaster areas are numbered 1, 2, 3, … 39.

[0168] The longitude and latitude coordinates of each distribution station are obtained through Baidu Maps. When simulating the route, due to the curvature of the ground roads, it is assumed that the driving distance of trucks and ground unmanned vehicles is 20% more than the straight-line distance. The longitude and latitude coordinates, demand, and accessibility of trucks and ground unmanned vehicles at each affected point are shown in Table 4. All numerical experiments are carried out on a computer with an AMD Ryzen 7-4800H, which has a Radeon graphics card with 2.90 GHz and 16 GB of memory. The optimization solver used is GUROBI Optimizer 10.0, running under the Windows 11 operating system.

[0169]

[0170] Table 4 Dataset of disaster-stricken points

[0171]

[0172] Table 4 Dataset of disaster-stricken points

[0173] Cooperative distribution result of trucks and ground unmanned vehicles: In the scenario of cooperative distribution of trucks and ground unmanned vehicles, the total distribution time is 615.34 min. Using trucks and ground unmanned vehicles for distribution reduces the total time by 27.29% compared to using only the same number of trucks for distribution. Especially in the case of public health events, when trucks cannot directly access quarantine points, it can effectively reduce the overall route length and distribution time and lower the risk of virus transmission.

[0174] The specific distribution route of the cooperative distribution of trucks and ground unmanned vehicles is as Figure 6 shown. The green dots in the figure represent non-isolated points, and the red dots represent isolated points. Among them, point 0 is the urban distribution center. The blue arrows represent the truck distribution route, and the red arrows represent the ground unmanned vehicle distribution route. During the entire distribution process, 4 trucks and 6 ground unmanned vehicles work together to transport emergency supplies from the urban distribution center to each disaster area. The specific routes of the trucks and ground unmanned vehicles are as follows:

[0175] Route 1: 0→19→15→13→10→9→4→1→2→0

[0176] 13→18→0

[0177] Route 2: 0→16→11→7→6→14→17→0

[0178] 0→12→7, 6→5→3→0, 11→8→6, 0→20→21→17

[0179] Route 3: 0→24→36→28→22→0

[0180] 36→38→33→26→0

[0181] Route 4: 0→23→29→34→32→35→29→37→31→27→0

[0182] 32→30→25→0

[0183] Sensitivity analysis

[0184] Truck load: The APSO-MHI algorithm is programmed using the Python language and run on the Pycharm platform. The algorithm parameters are set as follows: inertia coefficient 1 is 0.9, inertia coefficient 2 is 0.4, and learning factor is 2. Gurobi 10.0.1 is called to experimentally verify the APSO-MHI algorithm, and the results are shown in Table 5.

[0185]

[0186] Table 5 Influence of Truck Load on Results

[0187]

[0188] Table 5 Influence of Truck Load on Results

[0189] With other parameters unchanged, the influence of the truck's loading capacity on the total delivery time was analyzed. Table 5 records the results for different load weights (150 kg, 200 kg, 250 kg, 300 kg, 350 kg). From Figure 6 it can be seen that the truck's loading capacity has a significant impact on the delivery time. As the load weight increases, the total distribution time shows a downward trend, but the rate of decline gradually slows down, indicating that increasing the load weight can effectively reduce the distribution time when the load is small (150 kg - 250 kg). When the load weight reaches more than 300 kg, the optimization benefit decreases significantly. At the same time, the average driving time per vehicle increases with the increase in load weight and stabilizes after 300 kg, indicating that as the load weight increases, the tasks are more concentrated on fewer vehicles, increasing the burden on the bicycles. This shows that reasonably increasing the load capacity can reduce the overall distribution time, but excessive increase may lead to an excessive workload for individual vehicles, affecting the balance. When optimizing the scheduling, it is necessary to balance the total distribution efficiency and the task distribution of the bicycles to ensure the maximum overall operating efficiency.

[0190] To achieve the dual optimization of the total lead time and the average lead time per vehicle, a balance must be found between the two. This point is Figure 8 the intersection of the two curves in, corresponding to a specific load capacity, indicating that the total lead time and the average lead time per vehicle reach the optimal balance. According to Figure 7 the intersection in, the load of this vehicle is set to 250 kg. In the emergency logistics distribution system, the loading capacity of the truck is a key parameter and plays a crucial role in improving the distribution efficiency and shortening the total distribution time within a certain range.

[0191] Load of the ground unmanned vehicle: The results of the ground unmanned vehicle under loads of 60 kg, 70 kg, 80 kg, 90 kg, and 100 kg are shown in Table 6. From the experimental results in the table, it can be seen that the load-bearing weight of the ground unmanned vehicle has a certain impact on the total delivery time. As the weight of the ground unmanned vehicle increases, the total delivery time gradually decreases. As Figure 8As shown, when the weight of the ground-based autonomous vehicle increases from 60 kg to 100 kg, the total delivery time decreases from 615.34 to 659.54, a reduction of approximately 7.18%. It can be seen that increasing the loading capacity of the ground-based autonomous vehicle can improve the distribution efficiency to a certain extent and reduce the overall transportation time. From 90 kg to 100 kg, the downward trend of the delivery time flattens out, indicating that when the load capacity reaches a certain threshold, its effect on improving the distribution efficiency begins to weaken. In practical applications, it is necessary to comprehensively consider the load, cost, and scheduling requirements of the ground-based autonomous vehicle to determine the optimal load setting.

[0192]

[0193] Table 6 Influence of the Load of the Ground-Based Autonomous Vehicle on the Results

[0194]

[0195] Table 6 Influence of the Load of the Ground-Based Autonomous Vehicle on the Results

[0196] Mileage of the autonomous vehicle: Table 7 records the results under different mileage of the ground-based autonomous vehicle (20 km, 30 km, 40 km, 50 km, 60 km). The analysis shows that increasing the mileage limit of the autonomous vehicle can reduce the total delivery time. When the mileage limit increases from 20 km to 60 km, the total distribution time decreases from 638.62 min to 615.34 min, indicating that appropriately increasing the mileage of the ground-based autonomous vehicle can reduce the transportation burden of the truck, enabling the ground-based autonomous vehicle to complete more isolated point distribution tasks, thereby optimizing the overall scheduling. Figure 9 Among them, when the mileage increases from 20 km to 50 km, the delivery time decreases significantly, but when the mileage increases from 50 km to 60 km, the decrease amplitude becomes smaller, indicating that the marginal effect of the cruising range of the autonomous vehicle on improving the distribution efficiency is decreasing. In practical applications, it is necessary to comprehensively consider the cruising range of the ground-based autonomous vehicle, the driving mileage of the ground-based autonomous vehicle, and the collaborative scheduling of the truck and the ground-based autonomous vehicle to optimize the distribution efficiency and maintain the balance and stability of the scheduling.

[0197]

[0198] Table 7 Influence of the Mileage of the Ground-Based Autonomous Vehicle on the Results

[0199]

[0200] Table 7 Influence of the Mileage of the Ground-Based Autonomous Vehicle on the Results

[0201] For the problem of emergency material distribution, the embodiments of the present invention mainly include: First, a mixed-integer programming model is constructed with the goal of minimizing the total distribution time; Then, an adaptive particle swarm optimization algorithm based on multi-heuristic initialization is proposed to solve the model. The algorithm combines greedy heuristic, nearest neighbor heuristic, and stochastic nearest neighbor heuristic to improve the quality of the initial solution. At the same time, an adaptive inertia weight, probabilistic learning, tournament selection, and velocity mechanism are introduced into the PSO framework to enhance the global search ability and convergence stability of the algorithm. Finally, through actual case analysis and comparative experiments with the commercial solver Gurobi, the significant advantages of the proposed method in terms of solution quality and computational efficiency are verified, demonstrating its application potential in complex distribution problems. In most scenarios, Gurobi cannot find the optimal solution within 1800 s, while APSO-MHI can find the optimal solution within 200 s within a range of 4% from the optimal solution, which shows its superiority in terms of efficiency and solution quality, especially in large-scale distribution scenarios.

[0202] Compared with only using the same number of trucks, the total distribution time of the collaborative distribution of trucks and ground unmanned vehicles is reduced by 27.29%. However, too many ground unmanned vehicles may lead to resource waste or increased scheduling complexity. It is recommended that managers reasonably allocate the proportion of trucks and ground unmanned vehicles according to the geographical environment, the distribution of isolated island areas, and the distribution of disaster area tasks to ensure distribution efficiency and avoid excessive scheduling costs. According to the results of sensitivity analysis, emergency managers need to pay attention to the following aspects during the actual emergency logistics scheduling process to improve the overall allocation efficiency. (i) According to the relationship between the truck load and the distribution time, reasonably adjust the truck load capacity. It is recommended that emergency management personnel reasonably adjust the truck load during the scheduling process, rather than thinking that the larger the load, the better. Avoid over-concentration of tasks caused by excessive load, which affects the balanced distribution of tasks and the life cycle of a single vehicle. (ii) When planning the scheduling of unmanned vehicles, emergency managers should not blindly pursue a higher load, but should pay attention to appropriate load settings to avoid the negative impact of excessive load on the performance of unmanned vehicles and ensure the efficient operation and flexible task allocation of unmanned vehicles. (iii) Managers should reasonably set the upper limit of the driving mileage of unmanned vehicles according to the distribution of task locations. For tasks with a large distribution range requirement, the task range of unmanned vehicles should be appropriately expanded to improve the distribution efficiency. However, it should be noted that when exceeding a certain mileage limit, the improvement of distribution efficiency gradually becomes smaller. Therefore, it is recommended to combine the distribution range with the task requirements to avoid unnecessary resource waste.

[0203] Regarding the problem of collaborative path optimization of multiple ground unmanned vehicles and trucks for emergency material distribution, the actual characteristics of truck accessibility are considered. However, several other factors have not been taken into account, highlighting the limitations in the depth and breadth of the research. For example, the exploration of constraint conditions such as time windows and diverse distribution modes of ground unmanned vehicles is insufficient. Future research can study the scenarios of ground unmanned vehicles starting and recovering from different trucks and combine time window constraints to enhance the diversity and flexibility of trucks during multi-ground unmanned vehicle deliveries.

[0204] The material distribution method provided by the present invention constructs a mixed-integer programming model with a plurality of preset constraint information as constraint conditions according to multiple disaster-stricken points in the target area and the data of isolation points and non-isolation points at each disaster-stricken point. Among them, the mixed-integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model; with the goal of minimizing the distribution time, the mixed-integer programming model is solved based on an adaptive particle swarm optimization algorithm with multi-heuristic initialization; the optimal distribution paths of the multi-truck and multi-ground unmanned vehicle are determined based on the solution result of the mixed-integer programming model; and the multiple disaster-stricken points in the target area and the isolation points and non-isolation points at each disaster-stricken point are subjected to material distribution based on the optimal distribution paths. Compared with the defect in the prior art that trucks and ground unmanned vehicles cannot be efficiently scheduled, by this method, the accessibility of trucks and the flexibility of ground unmanned vehicles in the emergency scenario are comprehensively considered, a multi-truck and multi-ground unmanned vehicle collaborative distribution model is constructed with the goal of reducing the total distribution time of all trucks, and an optimal distribution path is obtained by solving this model based on an adaptive particle swarm optimization algorithm with multi-heuristic initialization, realizing efficient path planning and resource optimization scheduling.

[0205] The material distribution system provided by the present invention is described below. The material distribution system described below can be mutually referred to the corresponding description of the above-mentioned material distribution method.

[0206] Figure 10 It is a schematic structural diagram of the material distribution system provided by the present invention, specifically including:

[0207] A model construction module 1001, configured to construct a mixed-integer programming model with a plurality of preset constraint information as constraint conditions according to multiple disaster-stricken points in the target area and the data of isolation points and non-isolation points at each disaster-stricken point. Among them, the mixed-integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model. For detailed description, refer to the relevant description corresponding to the above method embodiment, which will not be elaborated here.

[0208] A model solving module 1002, configured to solve the mixed-integer programming model based on an adaptive particle swarm optimization algorithm with multi-heuristic initialization with the goal of minimizing the distribution time. For detailed description, refer to the relevant description corresponding to the above method embodiment, which will not be elaborated here.

[0209] A path determination module 1003 is configured to determine an optimal distribution path of the multiple trucks and multiple ground-based unmanned vehicles based on the solution result of the mixed integer programming model. For detailed descriptions, please refer to the relevant descriptions corresponding to the above method embodiments, which will not be elaborated here.

[0210] A material distribution module 1004 is configured to perform material distribution to multiple disaster-stricken points in the target area and the isolation points and non-isolation points of each disaster-stricken point based on the optimal distribution path. For detailed descriptions, please refer to the relevant descriptions corresponding to the above method embodiments, which will not be elaborated here.

[0211] Figure 11 An entity structure diagram of an electronic device is exemplified, as Figure 11 shown. The electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840. Among them, the processor 810, the communication interface 820, and the memory 830 complete communication with each other through the communication bus 840. The processor 810 can call the logical instructions in the memory 830 to execute a material distribution method, which includes: constructing a mixed integer programming model with a plurality of preset constraint information as constraint conditions according to multiple disaster-stricken points in the target area and data of isolation points and non-isolation points of each disaster-stricken point, where the mixed integer programming model is a collaborative distribution model of multiple trucks and multiple ground-based unmanned vehicles; solving the mixed integer programming model based on a multi-heuristic initialization-based adaptive particle swarm optimization algorithm with the goal of minimizing the distribution time; determining an optimal distribution path of the multiple trucks and multiple ground-based unmanned vehicles based on the solution result of the mixed integer programming model; and performing material distribution to multiple disaster-stricken points in the target area and the isolation points and non-isolation points of each disaster-stricken point based on the optimal distribution path.

[0212] In addition, when the logical instructions in the above-mentioned memory 830 are implemented in the form of software function units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0213] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the material distribution method provided by each of the above methods. The method includes: constructing a mixed-integer programming model with a plurality of constraints as constraint conditions according to a plurality of disaster-affected points in a target area and the data of isolation points and non-isolation points of each disaster-affected point, wherein the mixed-integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model; taking minimizing the distribution time as the goal, solving the mixed-integer programming model based on an adaptive particle swarm optimization algorithm with multi-heuristic initialization; determining the optimal distribution paths of the multi-truck and the multi-ground unmanned vehicle based on the solution result of the mixed-integer programming model; and performing material distribution on the plurality of disaster-affected points in the target area and the isolation points and non-isolation points of each disaster-affected point based on the optimal distribution paths.

[0214] In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it realizes the material distribution method provided by each of the above methods. The method includes: constructing a mixed-integer programming model with a plurality of constraints as constraint conditions according to a plurality of disaster-affected points in a target area and the data of isolation points and non-isolation points of each disaster-affected point, wherein the mixed-integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model; taking minimizing the distribution time as the goal, solving the mixed-integer programming model based on an adaptive particle swarm optimization algorithm with multi-heuristic initialization; determining the optimal distribution paths of the multi-truck and the multi-ground unmanned vehicle based on the solution result of the mixed-integer programming model; and performing material distribution on the plurality of disaster-affected points in the target area and the isolation points and non-isolation points of each disaster-affected point based on the optimal distribution paths.

[0215] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0216] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0217] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. A material distribution method, characterized in that, Including: Construct a mixed-integer programming model with the data of multiple disaster-stricken points in the target area, as well as the isolation points and non-isolation points of each disaster-stricken point, taking a plurality of preset constraint information as constraint conditions, wherein the mixed-integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model; Taking minimizing the distribution time as the goal, solve the mixed-integer programming model based on the multi-heuristic initialization adaptive particle swarm optimization algorithm; The step of taking minimizing the distribution time as the goal and solving the mixed-integer programming model based on the multi-heuristic initialization adaptive particle swarm optimization algorithm includes: Use the greedy heuristic method, the nearest neighbor heuristic method, and the stochastic nearest neighbor heuristic method to generate the initial solutions of the mixed-integer programming model respectively, and combine the initial solutions according to a preset ratio to generate an initial population; Optimize the particle swarm algorithm by dynamically adjusting the inertia weight, adaptive learning factors, tournament selection, and velocity pruning mechanism to obtain the multi-heuristic initialization adaptive particle swarm optimization algorithm; Solve the mixed-integer programming model based on the multi-heuristic initialization adaptive particle swarm optimization algorithm and output the global optimal solution; Determine the optimal distribution paths of the multi-truck and multi-ground unmanned vehicles based on the solution results of the mixed-integer programming model; Carry out material distribution for the multiple disaster-stricken points in the target area and the isolation points and non-isolation points of each disaster-stricken point based on the optimal distribution paths.

2. The method according to claim 1, characterized in that, The plurality of constraint information includes: path constraint, load constraint, mileage constraint, collaborative operation time constraint; The step of constructing a mixed-integer programming model with the data of multiple disaster-stricken points in the target area, as well as the isolation points and non-isolation points of each disaster-stricken point, taking a plurality of preset constraint information as constraint conditions includes: Define decision variables and objective variables according to the data of multiple disaster-stricken points in the target area and the isolation points and non-isolation points of each disaster-stricken point. Among them, the decision variables include path selection variables, release and recovery node selection variables, and the objective variable includes the total time for each truck to complete the distribution task, and the total time includes its own distribution operation time and collaborative operation time; Based on the decision variables and objective variables, and taking the plurality of constraint information as constraint conditions, construct a mixed-integer programming model.

3. The method according to claim 1, characterized in that, The step of using the greedy heuristic method, the nearest neighbor heuristic method, and the stochastic nearest neighbor heuristic method to generate the initial solutions of the mixed-integer programming model respectively, and combining the initial solutions according to a preset ratio to generate an initial population includes: Use the greedy heuristic method based on the greedy strategy. In each step of path selection, select the customer point that is the nearest to the current node and satisfies the constraint conditions as the next access node. Starting from the distribution center, repeat this process until no new customer points can be added to obtain the first initial solution; Use the nearest neighbor heuristic method to select the next access point according to multiple factors to obtain the second initial solution, where the multiple factors include distance factor, convenience of node connection, and potential influencing factors for subsequent path planning; Randomly select a candidate node from the set of adjacent nodes that satisfy the constraint conditions using a random nearest neighbor heuristic method, and continue to expand the path based on the candidate node to obtain a third initial solution; Combine the first initial solution, the second initial solution, and the third initial solution according to a preset ratio to generate an initial population.

4. The method according to claim 1, characterized in that The optimization of the particle swarm algorithm by dynamically adjusting the inertia weight, adaptive learning factor, tournament selection, and velocity pruning mechanism to obtain the multi-heuristic initialization adaptive particle swarm optimization algorithm includes: Dynamically adjust the inertia weight using a linearly decreasing weight strategy; Use an exponential function to increase the range of variation of the learning probability, determine the learning probability of each particle, and use the learning probability of each particle as the adaptive learning factor; Randomly select two or more particles for comparison, and select the better particle as a candidate particle for position update; Randomly prune some velocity components to control the velocity change of the particle within a preset range and maintain the movement direction to obtain the adaptive particle swarm optimization algorithm.

5. The method according to claim 1, characterized in that, After the material distribution to multiple disaster-stricken points in the target area and the isolation points and non-isolation points of each disaster-stricken point based on the optimal distribution path, it includes: Verify the solution result of the mixed integer programming model based on the material distribution result.

6. A material distribution system, characterized in that, It includes: A model construction module for constructing a mixed integer programming model with a preset number of constraint information as constraint conditions according to the data of multiple disaster-stricken points in the target area and the isolation points and non-isolation points of each disaster-stricken point, where the mixed integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model; A model solution module for solving the mixed integer programming model based on the multi-heuristic initialization adaptive particle swarm optimization algorithm with the goal of minimizing the distribution time; the solution of the mixed integer programming model based on the multi-heuristic initialization adaptive particle swarm optimization algorithm with the goal of minimizing the distribution time includes: using the greedy heuristic method, the nearest neighbor heuristic method, and the random nearest neighbor heuristic method to generate the initial solutions of the mixed integer programming model respectively, and combining the initial solutions according to a preset ratio to generate an initial population; optimizing the particle swarm algorithm by dynamically adjusting the inertia weight, adaptive learning factor, tournament selection, and velocity pruning mechanism to obtain the multi-heuristic initialization adaptive particle swarm optimization algorithm; solving the mixed integer programming model based on the multi-heuristic initialization adaptive particle swarm optimization algorithm and outputting the global optimal solution; A path determination module for determining the optimal distribution path of the multi-truck and multi-ground unmanned vehicle based on the solution result of the mixed integer programming model; A material distribution module for distributing materials to multiple disaster-stricken points in the target area and the isolation points and non-isolation points of each disaster-stricken point based on the optimal distribution path.

7. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the material distribution method according to any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the material distribution method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the material distribution method according to any one of claims 1 to 5.

Images