Material distribution method and system, electronic equipment and medium

Abstract

The embodiment of the invention relates to the technical field of emergency material distribution, and provides a material distribution method and system, an electronic device and a medium, and the method comprises the steps: building a mixed integer programming model with a plurality of pieces of preset constraint information as constraint conditions according to a plurality of disaster points of a target region and isolation point and non-isolation point data of each disaster point, the mixed integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model; solving the mixed integer programming model based on a multi-heuristic initialized adaptive particle swarm optimization algorithm by taking minimization of delivery time as a target; determining an optimal delivery path of the plurality of trucks and the plurality of ground unmanned vehicles based on a solving result of the mixed integer programming model; and based on the optimal distribution path, carrying out material distribution on a plurality of disaster-affected points of the target area and isolation points and non-isolation points of each disaster-affected point. Therefore, efficient path planning and resource optimization scheduling are realized.

Description

Technical Field

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

[0005] The present invention provides a material distribution method, including: 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 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; 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 the optimal distribution paths of the multi-truck and multi-ground unmanned vehicles based on the solution result of the mixed integer programming model; 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.

[0006] In a possible implementation manner, the method further includes: The plurality of constraint information includes: path constraint, load constraint, mileage constraint, collaborative operation time constraint; Define decision variables and objective variables based on multiple disaster-affected points in the target area, as well as the isolated and non-isolated point data of each disaster-affected 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 using the multiple constraint information as constraint conditions, construct a mixed-integer programming model.

[0007] In a possible implementation manner, the method further includes: 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; 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 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.

[0008] In a possible implementation manner, the method further 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 factors, convenience of node connection, and potential influencing factors for subsequent path planning; 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; Combine the first initial solution, the second initial solution, and the third initial solution according to a preset ratio to generate an initial population.

[0009] In a possible implementation manner, the method further includes: Use the linearly decreasing weight strategy to dynamically adjust the inertia weight; Use the exponential function to increase the change range 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 the candidate particle for position update; Randomly prune some velocity components, control the velocity change of the particles within a preset range and maintain the movement direction to obtain the adaptive particle swarm optimization algorithm.

[0010] In a possible implementation, the method further includes: Verify the solution result of the mixed-integer programming model based on the material distribution result.

[0011] The present invention also provides a material distribution system, including the following modules: A model construction module, configured to construct 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 isolated points and non-isolated points at each disaster-stricken point, wherein the mixed-integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model; A model solution module, configured to solve the mixed-integer programming model based on the adaptive particle swarm optimization algorithm initialized by multiple heuristics with the goal of minimizing the distribution time; 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; A material distribution module, configured to perform material distribution on a plurality of disaster-stricken points in the target area and the isolated points and non-isolated points at each disaster-stricken point based on the optimal distribution path.

[0012] 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, and when the processor executes the computer program, the material distribution method described in any one of the above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the material distribution method described in any one of the above is implemented.

[0014] The present invention also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the material distribution method described in any one of the above is implemented.

[0015] The material distribution method, system, electronic device and medium provided by the present invention construct a mixed integer programming model with a plurality of constraints in the preset 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. 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 a 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 materials are distributed to 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. 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 this model based on a multi-heuristic initialization adaptive particle swarm optimization algorithm, realizing efficient path planning and resource optimization scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] 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.

[0017] Figure 1 It is a schematic flowchart of the material distribution method provided by the present invention.

[0018] Figure 2 It is a schematic diagram of the truck-ground unmanned vehicle collaborative distribution mode provided by the present invention.

[0019] Figure 3 It is a schematic diagram of the path generation process provided by the present invention.

[0020] Figure 4 It is a schematic diagram of the influence of the learning factor on the heuristic algorithm provided by the present invention.

[0021] Figure 5 It is a flowchart of the adaptive particle swarm optimization algorithm initialized based on multiple heuristic algorithms provided by the present invention.

[0022] Figure 6 It is a collaborative distribution diagram of trucks and ground unmanned vehicles provided by the present invention.

[0023] Figure 7 It is a schematic diagram of the result analysis of different truck load capacities provided by the present invention.

[0024] Figure 8 It is a schematic diagram for the result analysis of the load capacity of different ground unmanned vehicles provided by the present invention.

[0025] Figure 9 It is a schematic diagram for the result analysis of the mileage of different ground unmanned vehicles provided by the present invention.

[0026] Figure 10 It is a schematic diagram of the structure of the material distribution system provided by the present invention.

[0027] Figure 11 It is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed implementation manners

[0028] 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.

[0029] 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.

[0030] Figure 1 It is a schematic flowchart of the material distribution method provided by the present invention. As Figure 1 shown, the method includes the following: S11. Construct a mixed integer programming model with a plurality of constraint information in the target area as constraint conditions according to the data of a plurality of disaster-affected points and the isolation points and non-isolation points of each disaster-affected point.

[0031] In the embodiments of the present invention, for example, there are multiple disaster-affected 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 manually driven trucks and unmanned ground vehicles in the distribution center for material distribution. This method mainly studies how to plan the paths of the distribution vehicles so that they can meet the material distribution needs in the disaster area and achieve the goal of the shortest total distribution time of all trucks.

[0032] First of all, it is necessary to construct a mixed integer programming model, which is a collaborative distribution model for multiple trucks and multiple ground unmanned vehicles.

[0033] Specifically, the affected points in the city are allocated by both trucks and ground-based unmanned vehicles. The trucks can also serve as mobile warehouses for the launch and recovery of ground-based unmanned vehicles. The affected points are divided into isolated points and non-isolated points. The former are distributed by ground-based unmanned vehicles, and the latter are distributed by trucks. During the distribution operation, the truck carries multiple ground-based unmanned vehicles from the distribution center and sequentially travels to the affected points or temporary parking points along the established route, delivers supplies to the affected points, and launches or recovers ground-based unmanned vehicles. As Figure 2 shown, 1 The truck sequentially travels to non-isolated points C1 - C5 - C8 along the predetermined route and launches ground-based unmanned vehicles D1 and D2 at C1. After the ground-based unmanned vehicles are launched, they sequentially travel to the affected points near C1 for material replenishment along the established routes (C1 - C2 - C3 and C1 - C4). After the distribution task is completed, they rendezvous with 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-based unmanned vehicle D4: C11 - C13 - distribution center). Each ground-based unmanned vehicle can only execute one distribution task, but each distribution task can serve multiple affected points. When the ground-based unmanned vehicle returns to the distribution center, no more material distribution is carried out.

[0034] Considering the technical characteristics and operating conditions of ground-based unmanned vehicles, the following assumptions are made: (1) Trucks and ground-based unmanned vehicles can access multiple disaster areas, and each disaster area is served by one or only one truck or ground-based unmanned vehicle. To reduce human contact and the risk of virus transmission, trucks can only serve non-isolated points, and ground-based unmanned vehicles can only serve isolated points.

[0035] (2) The locations and demands of each affected point are known.

[0036] (3) The driving speed of the truck and the number of ground-based unmanned vehicles it can carry are known. The truck has no driving distance limit but has a load limit.

[0037] (4) The driving speed, load limit, and mileage limit of ground-based unmanned vehicles are known.

[0038] (5) Ground-based unmanned vehicles can be launched from trucks or directly serve the disaster area starting from the warehouse.

[0039] (6) The time for delivering materials after the ground-based unmanned vehicle and the truck arrive at the affected location is known, and the time for the ground-based unmanned vehicle to deliver and recover operations is known. The truck can simultaneously recover and launch multiple ground-based unmanned vehicles.

[0040] (7) Each ground-based unmanned vehicle consumes a certain amount of energy during each trip. When the ground-based unmanned vehicle starts a new journey, it will be replaced with a fully charged battery.

[0041] When the truck serves non-isolated points, ground unmanned vehicles can be launched simultaneously, and the launch time is less than the service time.

[0042] (9) Understand the urban road traffic conditions.

[0043] Furthermore, decision variables and objective variables are defined based on multiple disaster-affected points in the target area and the data of isolated and non-isolated points for each disaster-affected 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, a mixed-integer programming model is constructed with multiple constraint information as constraints.

[0044] The names, meanings, and types of each variable are shown in Table 1:

[0045] Table 1 Variable names, meanings, and types

[0046] Table 1 Variable names, meanings, and types 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-affected points, where the set of disaster-affected 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, , .

[0047] 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-affected point are and respectively. Each release and recovery of the unmanned vehicle requires a certain operation time, which are respectively denoted as and W . The time point when the truck arrives at is denoted as . The time when the unmanned vehicle arrives at The time point of the point is . The truck departs from . The departure time of the truck is recorded as . The unmanned vehicle departs from . The departure time of the unmanned vehicle is recorded as .

[0048] Model construction includes: The complete operation time of each truck consists of two parts: the time to complete its own distribution task and the time to cooperate with the unmanned vehicle. The specific definition of the operation time is as follows: 1) Definition of the distribution operation time: Denote as the cumulative time for the truck to complete its own distribution task, including the driving time between each disaster-affected node and the service time at the disaster-affected point.

[0049] (1) 2) Definition of the truck-unmanned vehicle cooperation operation time: Denote as the cooperation operation time generated by the truck and the unmanned vehicle due to mutual waiting, release, and recovery at point. At any point the cooperation between the truck and the unmanned vehicle is divided into two parts: the waiting time required to recover the previously dispatched unmanned vehicle and the time required to dispatch the current unmanned vehicle.

[0050] Recovery waiting time: For the unmanned vehicle paired with the truck d , if the truck carries the unmanned vehicle to reach point, the recovery waiting time is recorded as =0. If the unmanned vehicle d and the truck do not reach point together, 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 unmanned vehicle arrives and then performs the recovery operation. The time consumed in this process is recorded as ; (2) The unmanned vehicle arrives at the recovery point first, that is . Then the consumed time is the recovery operation time of the unmanned vehicle, .

[0051] Then the waiting time for the truck to recover all unmanned vehicles at point is: .

[0052] Release waiting time: The truck is at When the point requires the release of the drone, the time consumed is the total time required for the unmanned vehicle released at this release point. .

[0053] Therefore, the total collaborative operation time of the truck at point is: .

[0054] For any truck , the time from leaving the warehouse to finally returning to the warehouse is recorded as , which is the sum of the time required for the truck to complete its own distribution task and the collaborative operation time with the unmanned vehicle. Therefore, the operation time of each truck in this problem is defined as (2) The constraint conditions include the following: (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) If , then (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) The objective function (3) minimizes the maximum delay time for the trucks or drones to return to the warehouse. The constraint (4) stipulates that all truck trips must start and end at the warehouse. The constraints (5) and (6) stipulate that each disaster-stricken point can only be served by a truck or a drone alone once. The constraints (7) and (8) indicate that the trucks and drones must leave a node after entering it during the distribution process. The constraint (9) indicates that the drone must start from the release node and return to the recovery node, and each drone is released only once.

[0055] The constraint (10) indicates that if a truck serves a disaster-stricken point , then the truck must enter this point once. The constraint (11) indicates that if a truck serves a disaster-stricken point , then the truck must leave this point once. Similarly, the constraint (12) indicates that if a drone d serves a point, then there is and only the drone d arrives at the point, passing through the arc ([[]] , ). The constraint (13) indicates that if a drone d serves a point, then there is and only the drone d leaves the arc the point, passing through the arc ([[]] , ). The constraint (14) indicates when a point is selected as the release point of the drone carried by the truck d , the drone d starts from the point. The constraint (15) indicates when a point is selected as the recovery point of the drone carried by the truck d , the drone d has to return to the point. The constraints (16) and (17) indicate that the release and recovery points of the drone are on the corresponding truck path nodes.

[0056] The constraint (18) restricts that the truck d carrying the drone should first arrive at the drone dThe release node, and then arrives at the recovery node of the driverless vehicle. Constraint (19) represents the recovery waiting time of the truck at the node The time relationship between the arrival and departure of the driverless vehicle at the disaster node is represented by Constraint (20). If the driverless vehicle passes through the arc ( , ), then it must first arrive at and then arrive at , and the time difference must be greater than or equal to the service time of the driverless vehicle at and the time passing through the arc ( , ). Similarly, Constraint (21) represents the time relationship between the arrival and departure of the truck at the node. If the truck passes through the arc ( , ), then it must first arrive at and then arrive at , and the time difference must be greater than or equal to the service time of the truck at and the time passing through the arc ( , ), plus the waiting time for the truck to recover 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.

[0057] S12. With the goal of minimizing the delivery time, solve the mixed-integer programming model based on the multi-heuristic initialization adaptive particle swarm optimization algorithm.

[0058] 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 factor, 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.

[0059] Specifically, in practice, although mathematical programming tools such as Gurobi can provide global optimal solutions, when dealing with large-scale or complex examples, Gurobi will verify whether each solution is a global optimal solution, and may face the problem of too long convergence time. Especially in 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 approximate optimal solutions within a reasonable time. The choice of the PSO algorithm as a method for solving the collaborative distribution problem of trucks and ground unmanned vehicles is mainly based on its advantages in dealing with complex optimization problems.

[0060] As an optimization strategy inspired by 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.

[0061] For the collaborative distribution path optimization problem of trucks and ground unmanned vehicles, the main idea is to unify first and then separate, and solve it through the combination of two stages. In the first stage, the whole process is regarded as a Vehicle Routing Problem (VRP); first, separate the isolated points and non-isolated points, take the non-isolated points as the truck paths, and the isolated 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 of the 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 speed of a particle, and a set of calculated paths. Calculate the set of each particle, then compare the optimality, and continuously iterate this process.

[0062] In Figure 3Among them, it is assumed that C1, C2, C3, and C4 are non-isolated points, and C5 and C6 are non-isolated points. In the first-stage optimization, 0-C1-C2-C3-C4-0 is the truck path, and 0-C5-0 is the path of the ground-based unmanned vehicle. In the second stage, the paths are merged to find all possible paths. In the third stage, all possible cases are considered to find the optimal result. The path of the ground-based unmanned vehicle changes from 0-C5-0 to C4-C5-C6-0, launches from C4, and retrieves at C1. In calculating the delivery time, since these trucks are in parallel service, after optimizing each truck path, the vehicle with the longest service time is found, which is the delivery time of the entire event.

[0063] Aiming at the disadvantage that the traditional PSO algorithm is prone to falling into local optimal solutions, the nearest neighbor initialization, dynamic inertia weight, adaptive learning factor, and velocity pruning are introduced to optimize the algorithm, and an improved adaptive inertia weight particle swarm optimization algorithm is proposed.

[0064] 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.

[0065] 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 preferentially considered 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 path. The advantage of this method is its fast convergence speed, and it can generate a relatively optimal path in a short time. Because it always selects locally optimal nodes, the total distance of the path can be optimized to a certain extent in the initial stage. For example, in areas where customer distribution is relatively concentrated, the greedy strategy can quickly find a compact path, reduce unnecessary transfers, and thus introduce a locally optimal path pattern into the initial population, providing a good starting point for subsequent evolution.

[0066] 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 its potential impact on subsequent path planning are comprehensively considered, and the next access point is determined through a relatively flexible selection mechanism. This method not only ensures a certain path efficiency but also increases the diversity of the path. It can explore some path possibilities that the greedy algorithm may ignore and avoid the initial population being overly concentrated near a single local optimal solution.

[0067] Random Neighborhood Heuristic Initialization: The random neighborhood heuristic algorithm constructs paths by randomly selecting adjacent nodes, completely breaking the deterministic selection pattern. Each time when selecting the next node, a candidate node is randomly chosen from the set of adjacent nodes that satisfy certain constraint conditions, and then the path is continued to be expanded 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 unexpectedly effective path combinations.

[0068] Constructing the Initial Solution by Combining Three Heuristic Methods: The initialization uses the greedy nearest neighbor heuristic, the nearest neighbor heuristic, and the random neighbor heuristic to generate the initial solution. Each method generates multiple paths 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 path list, path distance, particle position, and velocity. At the same time, the fitness value is calculated based on the number of trucks and the path distance, and finally the initial population and the fitness value table are returned to provide a diverse set of solutions for the subsequent optimization process.

[0069] Allocate three heuristic algorithms to participate in the generation of the initial population. Divide the total population size M 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 neighbor heuristic algorithm is controlled by a random value phi. In the randomly generated array, determine the number of particles less than phi as the initial number of particles generated by the nearest neighbor heuristic algorithm, and the rest are the initial number of particles generated by the random neighbor heuristic algorithm. Through this dynamic ratio 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. By 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.

[0070] 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.

[0071] 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. Adopt the linearly decreasing weight strategy proposed by Shi and Y. The inertia weight is dynamically adjusted according to Equation (28), where is the k inertia weight of the It is usually large (such as 0.9) to enhance the initial global exploration ability, and the final inertia weight is usually small (such as 0.4) for local development in the later stage of convergence, k represents the current iteration number, and max_gen is the maximum iteration number. In the initial stage of the search, a larger inertia weight promotes the particles to conduct global exploration at a faster speed, can better cover the entire search space, and avoid falling into local optima. In the later stage, a smaller inertia weight reduces the particle speed, 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, the value of gradually decreases, which can make the particles move less and contribute to precise local search during the process of approaching the global optimal solution.

[0072] (28) The individual learning factor c1 and the social learning factor c2 are usually between (0, 2), which are used to balance the influence of the particle's self-experience and group experience on its speed, as Figure 4 shown. A larger learning factor will make the particle pay more attention to the optimal position of the group, while a smaller learning factor will make the particle pay more attention to its own optimal position. In the PSO algorithm, the particles update their positions according to their own historical optimal positions (pbest) and the global optimal position (gbest) of the group. In the 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.

[0073] The adaptive setting of the learning probability is to adopt different learning strategies for different particles. Using the non-linear function of Equation (29), the exponential function is used to increase the variation range of the learning probability, and calculate the learning probability of each particle, 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, the more likely the particle is to learn and maintain 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, The maximum value is 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 in the learning probability shows a trend from low to high, that is, particles with a larger number are more inclined to maintain their historical experience, while particles with a smaller number are more likely to explore new solutions.

[0074] (29) 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, a tournament selection mechanism is adopted 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.

[0075] 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). Where is the number of selected particles, rand(1, M ) represents randomly selecting a particle from 1 to M .

[0076] (30) 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, enabling the particle to 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 reduced efficiency. Through velocity pruning, the particle can more stably stay in the effective solution space and gradually approach the optimal solution.

[0077] 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: (31) (32) Among them c1 and c2 are learning factors, usually c1 = c2 = 2, rand () is a random number between (0, 1). According to formula (31) and formula (32), the standard form of PSO is: (33) Among them is the inertia factor, and its value is non - negative. In practical applications, the selection of learning factors and inertia coefficients has an important impact on the performance of the PSO algorithm. Usually, it is necessary to determine the optimal values of learning factors and inertia coefficients 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 flow chart for initializing the adaptive particle swarm optimization algorithm based on multiple heuristic algorithms is as Figure 5 shown.

[0078]

[0079] Table 2 Pseudo - code for initializing the adaptive particle swarm optimization algorithm based on multiple heuristic algorithms

[0080] Table 2 Pseudo - code for initializing the adaptive particle swarm optimization algorithm based on multiple heuristic algorithms S13. Determine the optimal distribution path of the multiple trucks and multiple ground unmanned vehicles based on the solution result of the mixed - integer programming model.

[0081] 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.

[0082] Through the solution of the above algorithm, the optimal distribution paths of multiple trucks and multiple ground unmanned vehicles are obtained. These paths include: The distribution path of the truck, starting from the distribution center, sequentially visiting non - isolation points, and simultaneously launching and recovering ground unmanned vehicles.

[0083] 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.

[0084] Distribution execution: Truck delivery: The truck travels to non-isolated points in sequence according to the planned route for material delivery. At non-isolated points, the truck can simultaneously launch ground unmanned vehicles to travel to nearby isolated points for delivery. After the truck completes the delivery task, it returns to the distribution center.

[0085] Ground unmanned vehicle delivery: The ground unmanned vehicle departs from the truck or the distribution center to travel to isolated points for material delivery. After the ground unmanned vehicle completes the delivery task, it returns to the truck or the distribution center for battery replacement and material loading. During the delivery process, the ground unmanned vehicle visits multiple isolated points in sequence according to the planned route.

[0086] Collaborative operation: The truck and the ground unmanned vehicle maintain collaborative operation during the delivery process. When the truck stays at non-isolated points, it can launch or retrieve ground unmanned vehicles. The delivery routes of the truck and the ground unmanned vehicle are optimized to ensure that all disaster-stricken points are supplied with materials in the shortest time.

[0087] Example verification: 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 initial solutions by combining three heuristic algorithms provides high-quality initial solutions for PSO and reduces the randomness of early convergence. By dynamically adjusting the inertia weight, the probability of individual learning and group learning is 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, making the retention probability of excellent solutions higher and avoiding the loss of high-quality solutions. The velocity pruning mechanism restricts the excessive transition of particles and improves the accuracy and stability of the solution. The APSO-MHI algorithm is suitable for 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.

[0088] Taking the Solomon dataset as an example, this dataset is a relatively classic dataset for studying VRRP-related issues. The instance definitions and pointers of the most famous solutions for 25 and 50 customer instances of the Solomon VRPTW benchmark problem in 1987 can be found on Solomon's official website. The Solomon standard test data has a 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 the convenience of calculation, the distance between nodes is used as the transportation cost between nodes. The demand is the demand of this 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 points, 8 points are randomly selected as isolation points. Among the 50 disaster points, 16 points are randomly selected as isolation points, and only ground unmanned vehicles can reach them.

[0089] APSO-MHI implements algorithm programming and operations using the Python language on the Pycharm platform. All numerical experiments were carried out on a computer with an AMD Ryzen 7-4800H, which has a Radeon graphics card with 2.90 GHz and 16GB of memory. The optimization solver used is the GUROBI Optimizer 10.0, which runs under the Windows 11 operating system. The results of running the algorithm 10 times are taken 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, the learning factor c = 2. Call Gurobi 10.0.1 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 unmanned vehicle is 0.42 km / min, the full-load limited distance of the ground unmanned vehicle is 60 km, the load limit of the truck is 500 kg, and the load limit of the ground unmanned vehicle is 200 kg.

[0090] APSO-MH Numerical Example and Analysis: The running results are shown in Table 7, which presents the performance comparison of the Gurobi solver, the proposed APSO-MHI algorithm 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 percentage error 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 percentage error between the optimal solution obtained by the PSO algorithm and the optimal solution obtained by the Gurobi solver.

[0091] 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 small, indicating that its solution is close to the optimal. For large-scale examples, Gurobi cannot obtain a feasible solution or an optimal solution within 3600s. Since the VRP problem is essentially an NP-hard problem, as the number of disaster points 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.

[0092] On the dataset containing 25 disaster sites, the GAP2 (%) value ensures a balance between solution quality and computing power within the 5% threshold, providing a reliable performance benchmark comparison, which is 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 solving quality and running efficiency, and is a robust and efficient algorithm for dealing with complex task scenarios.

[0093]

[0094] Table 3 Comparison of Test Results between GUROBI Solver and APSO - MHI Case Verification and Analysis Taking the distribution stations in Qingbaijiang District, Xindu District, Chenghua District and other areas 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 setting. 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.

[0095] The longitude and latitude coordinates of each distribution station are obtained through Baidu Map. When simulating the route, due to the curvature of the ground road, it is assumed that the driving distance of the truck and the ground unmanned vehicle is 20% more than the straight - line distance. The longitude and latitude coordinates, demand volume, and accessibility of the truck and the ground unmanned vehicle 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.

[0096]

[0097] Table 4 Dataset of Disaster - affected Points

[0098] Table 4 Dataset of Disaster - affected Points Results of collaborative distribution of trucks and unmanned vehicles: In the scenario of collaborative 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 with using only the same number of trucks for distribution. Especially in the case of public health events, when trucks cannot directly access isolation points, it can effectively reduce the overall route length and distribution time, and reduce the risk of virus transmission.

[0099] The specific distribution route of the collaborative 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, where 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 whole 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: Route 1: 0 → 19 → 15 → 13 → 10 → 9 → 4 → 1 → 2 → 0 13 → 18 → 0 Route 2: 0 → 16 → 11 → 7 → 6 → 14 → 17 → 0 0 → 12 → 7, 6 → 5 → 3 → 0, 11 → 8 → 6, 0 → 20 → 21 → 17 Route 3: 0 → 24 → 36 → 28 → 22 → 0 36 → 38 → 33 → 26 → 0 Route 4: 0 → 23 → 29 → 34 → 32 → 35 → 29 → 37 → 31 → 27 → 0 32 → 30 → 25 → 0 Sensitivity analysis Truck load: The APSO - MHI algorithm was programmed using the Python language and run on the Pycharm platform. The algorithm parameters were set as follows: inertia coefficient 1 was 0.9, inertia coefficient 2 was 0.4, and the learning factor was 2. Gurobi 10.0.1 was called to experimentally verify the APSO - MHI algorithm, and the results are shown in Table 5.

[0100]

[0101] Table 5 Influence of truck load on the results

[0102] Table 5 Influence of truck load on the results With other parameters unchanged, the influence of the truck loading capacity on the total delivery time was analyzed. Table 5 records the results for different loads (150 kg, 200 kg, 250 kg, 300 kg, 350 kg). From Figure 6 it can be seen that the truck loading capacity has a significant impact on the delivery time. As the load increases, the total distribution time shows a downward trend, but the rate of decline gradually slows down, indicating that increasing the load when the load is small (150 kg - 250 kg) can effectively reduce the distribution time. When the load 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 and stabilizes after 300 kg, indicating that as the load 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 task volume for a single vehicle, 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.

[0103] 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 8The intersection point of the two curves corresponds to a specific load capacity, indicating the optimal balance between the total lead time and the average lead time for each vehicle. According to Figure 7 the intersection point 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, which plays a crucial role in improving the distribution efficiency and shortening the total distribution time within a certain range.

[0104] Ground unmanned vehicle load: 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. It can be seen from the experimental results in the table that the load-bearing capacity 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 8 shown, when the weight of the ground unmanned vehicle increases from 60 kg to 100 kg, the total delivery time decreases from 615.34 to 659.54, a decrease of approximately 7.18%. It can be seen that increasing the loading capacity of the ground unmanned 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 distribution time flattens out, indicating that when the load capacity reaches a certain threshold, its role in 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 unmanned vehicle to determine the optimal load setting.

[0105]

[0106] Table 6 Influence of ground unmanned vehicle load on the results

[0107] Table 6 Influence of ground unmanned vehicle load on the results Unmanned vehicle mileage: Table 7 records the results under different ground unmanned vehicle load mileages (20 km, 30 km, 40 km, 50 km, 60 km). The analysis shows that increasing the driving mileage limit of autonomous vehicles 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 unmanned vehicle can reduce the transportation burden of the truck, enabling the ground unmanned vehicle to complete more isolated point distribution tasks, thereby optimizing the overall scheduling. In Figure 9 it, when the mileage increases from 20 km to 50 km, the distribution time decreases significantly, but when the mileage increases from 50 km to 60 km, the decline amplitude becomes smaller, indicating that the marginal effect of the driving range of autonomous vehicles on improving the distribution efficiency is decreasing. In practical applications, it is necessary to comprehensively consider the endurance of the ground unmanned vehicle, the driving mileage of the ground unmanned vehicle, and the coordinated scheduling of the truck and the ground unmanned vehicle to optimize the distribution efficiency and maintain the balance and stability of the scheduling.

[0108]

[0109] Table 7 Influence of the mileage of the ground unmanned vehicle on the results

[0110] Table 7 Influence of the mileage of the ground unmanned vehicle on the results 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 random 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.

[0111] 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 increase the 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 the 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 deployment 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 managers 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.

[0112] 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 constraints such as time windows and diverse distribution modes of ground unmanned vehicles is not sufficient. Future research can study the scenarios of ground unmanned vehicles starting and recovering from different trucks and incorporate time window constraints to enhance the diversity and flexibility of trucks during multi-ground unmanned vehicle deliveries.

[0113] The material distribution method provided by the present invention constructs a mixed-integer programming model with a plurality of preset constraint information as constraints based on the data of multiple disaster-stricken points in the target area and the isolated and non-isolated points of 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 materials are distributed to the multiple disaster-stricken points in the target area and the isolated and non-isolated points of each disaster-stricken point 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 the 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.

[0114] The material distribution system provided by the present invention is described below, and the material distribution system described below can be mutually referred to the material distribution method described above.

[0115] Figure 10 It is a schematic structural diagram of the material distribution system provided by the present invention, specifically including: A model construction module 1001, configured to construct a mixed-integer programming model with a plurality of preset constraint information as constraints based on the data of multiple disaster-stricken points in the target area and the isolated and non-isolated points of 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, and details are not described herein again.

[0116] 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, and details are not described herein again.

[0117] A path determination module 1003 is configured to determine an optimal distribution path for 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.

[0118] A material distribution module 1004 is configured to perform material distribution for 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.

[0119] 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 communicate with each other through the communication bus 840. The processor 810 may call logic 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 the data of the isolation points and non-isolation points of each disaster-stricken point, where the mixed-integer programming model is a collaborative distribution model for 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 for the multiple trucks and multiple ground-based unmanned vehicles based on the solution result of the mixed-integer programming model; and performing material distribution for 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.

[0120] In addition, when the logical instructions in the above-mentioned memory 830 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an 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. This 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 memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0121] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that 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 the above-mentioned various methods. The method includes: constructing a mixed-integer programming model with a plurality of constraints as constraint conditions according to the data of a plurality of 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; taking minimizing the distribution time as the goal, solving the mixed-integer programming model based on a multi-heuristic initialization 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 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 paths.

[0122] On 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 execution of the material distribution method provided by the above-mentioned various methods. The method includes: constructing a mixed-integer programming model with a plurality of constraints as constraint conditions according to the data of a plurality of 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; taking minimizing the distribution time as the goal, solving the mixed-integer programming model based on a multi-heuristic initialization 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 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 paths.

[0123] 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 work.

[0124] 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. The 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 for causing 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.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended 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 cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A material distribution method, characterized in that: include: According to the data of multiple disaster-affected points in the target area and the isolation points and non-isolation points of each disaster-affected point, a mixed integer programming model is constructed with multiple preset constraint information as constraint conditions, wherein the mixed integer programming model is a collaborative distribution model of multiple trucks and multiple ground unmanned vehicles; With the goal of minimizing the delivery time, an adaptive particle swarm optimization algorithm based on multi-heuristic initialization is used to solve the mixed integer programming model; Determining the optimal delivery path of the multiple trucks and multiple ground unmanned vehicles based on the solution results of the mixed integer programming model; Based on the optimal distribution path, materials are distributed to multiple disaster-stricken points in the target area and to isolation points and non-isolation points of each disaster-stricken point.

2. The method according to claim 1, characterized in that The plurality of constraint information includes: path constraint, load constraint, mileage constraint, and collaborative operation time constraint; The mixed integer programming model is constructed based on multiple disaster-affected points in the target area and the isolation point and non-isolation point data of each disaster-affected point with multiple preset constraint information as constraint conditions, including: Decision variables and target variables are defined according to multiple disaster-affected points in the target area and the isolated point and non-isolated point data of each disaster-affected point, wherein the decision variables include path selection variables, release and recovery node selection variables, and the target variables include the total time for each truck to complete the delivery task, and the total time includes its own delivery operation time and collaborative operation time; A mixed integer programming model is constructed based on the decision variables and the target variables and using the plurality of constraint information as constraint conditions.

3. The method according to claim 2, characterized in that The method of solving the mixed integer programming model based on an adaptive particle swarm optimization algorithm with multiple heuristic initializations with the goal of minimizing the delivery time includes: Using a greedy heuristic method, a nearest neighbor heuristic method, and a random nearest neighbor heuristic method to respectively generate initial solutions of the mixed integer programming model, and combining the initial solutions according to a preset ratio to generate an initial population; The particle swarm optimization algorithm is optimized by dynamically adjusting the inertia weight, the adaptive learning factor, the tournament selection and the speed pruning mechanism to obtain the adaptive particle swarm optimization algorithm with multi-heuristic initialization; The mixed integer programming model is solved based on the adaptive particle swarm optimization algorithm initialized by the multiple heuristics, and a global optimal solution is output.

4. The method according to claim 3, characterized in that The method of using a greedy heuristic method, a nearest neighbor heuristic method and a random nearest neighbor heuristic method to respectively generate initial solutions of the mixed integer programming model, and combining the initial solutions according to a preset ratio to generate an initial population includes: The greedy heuristic method is based on the greedy strategy. In each step of path selection, the customer point closest to the current node and satisfying the constraints is selected as the next access node. Starting from the distribution center, the process is repeated until no new customer points can be added, and the first initial solution is obtained; A nearest neighbor heuristic method is used to select a next access point according to multiple factors to obtain a second initial solution, wherein the multiple factors include distance factors, convenience of node connection, and potential influencing factors on subsequent path planning; A random nearest neighbor heuristic method is used to randomly select a candidate node from a set of adjacent nodes that satisfy the constraint condition, and the path is further expanded based on the candidate node to obtain a third initial solution; The first initial solution, the second initial solution and the third initial solution are combined according to a preset ratio to generate an initial population.

5. The method according to claim 3, characterized in that: The method optimizes the particle swarm algorithm by dynamically adjusting the inertia weight, the adaptive learning factor, the tournament selection and the speed pruning mechanism to obtain the adaptive particle swarm optimization algorithm with multi-heuristic initialization, including: Adopt linear decreasing weight strategy to dynamically adjust inertia weight; Using an exponential function to increase the range of variation of the learning probability, determining the learning probability of each particle, and using 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 the candidate particle for position update; Part of the velocity components are randomly trimmed to control the velocity change of the particles within a preset range and maintain the moving direction, thereby obtaining the adaptive particle swarm optimization algorithm.

6. The method according to claim 1, characterized in that After the materials are distributed to the 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, the following steps are included: The solution results of the mixed integer programming model are verified based on the material distribution results.

7. A material distribution system, characterized in that: include: A model building module is used to build a mixed integer programming model based on multiple disaster-stricken points in the target area and the isolation point and non-isolation point data of each disaster-stricken point, with multiple preset constraint information as constraint conditions, wherein the mixed integer programming model is a multi-truck and multi-ground unmanned vehicle collaborative distribution model; A model solving module, for solving the mixed integer programming model based on an adaptive particle swarm optimization algorithm with multi-heuristic initialization with the goal of minimizing the delivery time; A path determination module, used for determining the optimal delivery path of the multiple trucks and multiple ground unmanned vehicles based on the solution result of the mixed integer programming model; The material distribution module is used to distribute 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.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the material distribution method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the material distribution method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the material distribution method according to any one of claims 1 to 6 is implemented.

Images