Vehicle path planning method and system supporting multiple stations and customer priorities

Abstract

The invention discloses a vehicle path planning method and system supporting multiple stations and customer priorities. The method comprises the steps of 1, constructing a target function to obtain a global optimal solution of the target function; 2, if the random number is smaller than a preset experience feedback parameter, executing a development process, otherwise, entering an exploration process to obtain a local optimal solution, and adding the local optimal solution into a preset experience learning array; step 3, repeating the step 2 until a preset number of times is reached; 4, if the preset maximum number of iterations is not reached, updating preset experience feedback parameters, clearing all data in the experience array, and repeating iteration; and if the preset maximum number of iterations is reached, outputting the current global optimal solution, and completing the rescue material transport vehicle path planning. According to the method, the self optimization ability is improved, falling into a local optimal solution is prevented, the customer priority is introduced, the actual demand is better met, and casualties and pains of victims can be reduced to the maximum extent.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle path planning, and in particular to a vehicle path planning method and system supporting multiple stations and customer priorities. Background Art

[0002] The vehicle rescue process in disaster scenarios is a typical vehicle routing problem (VRP), which studies how to optimize transportation costs by rationally organizing vehicles and scientifically planning routes. It is an NP-hard problem. In recent years, the frequent occurrence of natural disasters has had an unprecedented impact and even threat on human social development and natural ecosystems. The sixth assessment report of the United Nations Intergovernmental Panel on Climate Change pointed out that the degree of harm caused by sudden natural disasters will increase in the future. Therefore, in emergency logistics, it is necessary to consider the speed and fairness of material distribution from the perspective of the disaster site to ensure that relief materials arrive at each disaster site in a timely manner. The goal of the classic vehicle routing problem is usually to minimize operating costs, but in humanitarian logistics and rescue operations, the main goal is to minimize casualties and suffering of victims.

[0003] To achieve this goal, existing methods such as the Cumulative Capacitated Vehicle Routing Problem (CCVRP) with capacity constraints aim to minimize the total arrival time of all customers, rather than the traditional path length, under vehicle capacity and quantity constraints. The solution to this problem has been shown to be more applicable to humanitarian supply chains. However, since there are usually multiple rescue centers in actual emergency logistics, the concept of cumulative waiting time is combined with the multi-yard vehicle routing problem to obtain the cumulative time multi-yard vehicle routing problem (MDCCVRP). This problem is more practical in emergency logistics.

[0004] In addition, the severity of disasters in different disaster-stricken areas is different. Although MDCCVRP considers that every customer should receive services equally, this fairness may lead to unfairness in disaster relief. Generally, the closer the area is to the disaster site, the more seriously affected it is. If the principle of equal disaster relief is followed, it may lead to casualties and suffering in severely affected areas due to failure to receive timely assistance. Therefore, the cumulative time-based multi-yard vehicle routing problem (MDCCVRP-Pr) considering the order of visits to disaster-stricken areas is derived. Specifically, according to the different disaster-stricken areas, they are divided into four different severity levels, represented by numbers from 0 to 3, where the larger the number, the more serious the disaster, the more urgent the need for materials, and these areas need to be reached first, otherwise, they will cause greater casualties and losses to these areas.

[0005] The cumulative time multi-depot vehicle routing problem with visit order in disaster-stricken areas (MDCCVRP-Pr) can be solved through local search. However, the local search algorithm in the existing methods cannot effectively utilize historical search information when solving the cumulative time multi-depot vehicle routing problem with capacity constraints and is prone to falling into local optimality. Summary of the invention

[0006] In order to at least partially solve the problem that the existing local search algorithm cannot effectively use historical search information and is prone to fall into local optimality when solving the problem of cumulative time vehicle routing of multiple stations with capacity constraints, the present invention provides a vehicle routing planning method and system that supports multiple stations and customer priorities, by using a clustering algorithm to divide customer points and considering customer priorities to generate an initial solution. An adaptive learning strategy is used to guide the algorithm process, and there are two nested large and small iteration cycles. In the small cycle, the algorithm will perform the exploration process or the development process and update the optimal solution in each iteration through the experience feedback parameter, and calculate the guidance parameters of the next small cycle in the large cycle iteration. The diversity of the algorithm is guaranteed through the exploration process, and the optimization ability of the algorithm is improved through the development process. Finally, after the large cycle iteration is completed, the planned path is output. The present invention can indicate the direction for the subsequent steps according to historical information, can avoid falling into local optimality or failure to converge, can effectively realize the cumulative time vehicle routing planning of multiple stations and considering customer priorities, thereby reducing the casualties of disaster victims.

[0007] In order to achieve the above object, the technical solution of the present invention is:

[0008] A first aspect of the present invention provides a vehicle path planning method supporting multiple stations and customer priorities, comprising:

[0009] Step 1: Construct an objective function based on customer information and station information, cluster and group customers using the K-means algorithm, and construct a global optimal solution for the objective function based on customer priority and clustering results; wherein the global optimal solution includes multiple paths, each path represents the delivery sequence of a vehicle, so as to facilitate subsequent optimization of the global optimal solution;

[0010] Step 2: Construct a random number and compare the random number with the size of the preset experience feedback parameter;

[0011] If the random number is smaller than the preset experience feedback parameter, the development process is executed to obtain the local optimal solution of the objective function; if the local optimal solution is smaller than the global optimal solution, the global optimal solution is updated and the local optimal solution is added to the preset experience learning array; the development process is conducive to improving the optimization ability of the objective function;

[0012] If the random number is greater than the preset experience feedback parameter, the exploration process is executed to obtain the local optimal solution of the objective function, and the global optimal solution is directly updated, and the local optimal solution is added to the preset experience learning array; the exploration process makes it easy to jump out of the local optimal solution;

[0013] Step 3: Repeat step 2 until the data in the experience learning array reaches the preset experience upper limit;

[0014] Step 4: Determine whether the preset maximum number of iterations has been reached. If the preset maximum number of iterations has not been reached, use the experience array to update the preset experience feedback parameters and clear all data in the experience array, and repeat the above steps 2, 3 and 4 to iterate;

[0015] If the preset maximum number of iterations is reached, the current global optimal solution is output and the path planning for the relief material transport vehicle is completed.

[0016] Furthermore, the objective function is expressed by the following formula:

[0017]

[0018] Where F is the objective function, is the time from the corresponding station to customer point i, I ij The penalty time for customer i to be served before customer j, p ij is a binary variable, The time from the corresponding station to customer point j, R is the total number of paths, V′ is the total number of customers, and k is the kth path.

[0019] Furthermore, the development process specifically includes:

[0020] S301, using a destruction and reconstruction local search operator to process the global optimal solution and update the global optimal solution;

[0021] S302, using a station allocation operator to process the updated global optimal solution to obtain a solution after station allocation;

[0022] S303, defining multiple neighborhood structures, neighborhood global optimal solutions S0 and neighborhood local optimal solutions S1, and saving the neighborhood global optimal solution S0 as the neighborhood local optimal solution S1, wherein the initial values ​​of the neighborhood global optimal solution S0 and the neighborhood local optimal solution S1 are both solutions after the station is allocated;

[0023] A plurality of said neighborhood structures include one-point movement, two-point exchange, point-edge exchange, edge exchange and OrOpt;

[0024] One-point move includes moving the selected customer to other locations, two-point exchange includes exchanging the positions of two selected customers, point-edge exchange includes exchanging the positions of one customer on the path with two other customers, edge exchange includes exchanging the positions of two customers on the path with two other customers, and OrOpt includes moving the selected consecutive 2-4 customers to other locations;

[0025] S304, judging the current neighborhood structure, if the current neighborhood structure is the last neighborhood structure, executing S310, if the current neighborhood structure is not the last neighborhood structure, performing a local search for the neighborhood local optimal solution S1 on the current neighborhood structure to obtain a solution S2;

[0026] S305, if the solution S2 is smaller than the neighborhood local optimal solution S1, execute S306; if the solution S2 is larger than the neighborhood local optimal solution S1, execute S307;

[0027] S306, using 2-opt and the station exchange operator to process the solution S2 to obtain a solution S3, and using the solution S3 as the new neighborhood local optimal solution S1, and then returning to S304; wherein 2-opt includes flipping the order of customers on the same path, and the station exchange operator includes exchanging different stations on every two paths in the solution S2;

[0028] S307, compare the neighborhood global optimal solution S0 with the neighborhood local optimal solution S1, if the neighborhood local optimal solution S1 is smaller than the neighborhood global optimal solution S0, then execute S308, if the neighborhood local optimal solution S1 is larger than the neighborhood global optimal solution S0, then enter S309;

[0029] S308, updating the neighborhood global optimal solution S0 according to the neighborhood local optimal solution S1, and returning to execute S304;

[0030] S309, executing S304 on the next neighborhood structure;

[0031] S310, outputting a neighborhood global optimal solution S0; wherein the neighborhood global optimal solution S0 is a local optimal solution.

[0032] Furthermore, the use of the destruction and reconstruction local search operator to process the global optimal solution and update the global optimal solution specifically includes:

[0033] Randomly select a customer from the global optimal solution, remove a preset number of customers closest to the customer from the solution, and obtain a destroyed solution;

[0034] Insert the removed customers according to their priorities in random order to the path that minimizes the increase in the objective function value, and obtain the repaired solution;

[0035] If the repaired solution is smaller than the global optimal solution, the repaired solution is used as the global optimal solution and the global optimal solution is updated. If the repaired solution is larger than the global optimal solution, the above process is repeated. After reaching the preset maximum number of updates, the last repaired solution is used as the global optimal solution and the global optimal solution is updated.

[0036] Furthermore, the updated global optimal solution is processed by the station allocation operator to obtain a solution after station allocation, specifically including:

[0037] Remove all stations on the updated optimal solution path of the entire set, assign the station closest to the first customer on each path to the path, and obtain the solution after assigning the stations.

[0038] Furthermore, the local search includes transforming the neighborhood local optimal solution S1 according to the current neighborhood structure to obtain a solution S2.

[0039] Furthermore, the local search also includes a capacity-based infeasible solution acceptance strategy, and the capacity-based infeasible solution acceptance strategy includes accepting a solution with a lower function value and always giving priority to a feasible solution. The specific process is expressed by the following formula:

[0040]

[0041] Among them, F(s) is the feasible objective function, Q(s) is the acceptance function when the capacity constraint is violated, is the sum of the squares of the demands of all customers on path k, and R is the total number of paths.

[0042] Furthermore, the exploration process specifically includes:

[0043] Randomly label all paths in the global optimal solution with k, where k = 1, 2, ..., R, and R is the total number of paths;

[0044] A customer on the path labeled k is randomly selected to exchange positions with a customer on the path labeled k+1, and the exchange operation is repeated until all paths have completed the customer position exchange; the path labeled R also needs to exchange customer positions with the path labeled 1 to obtain the perturbed solution;

[0045] Among them, the solution after perturbation is the local optimal solution.

[0046] Furthermore, the updating preset experience feedback parameter is expressed by the following formula:

[0047]

[0048] Among them, α is the experience feedback parameter, σ(C′) is the standard deviation of the normalized experience, and C′i is the normalized value of the ith experience, is the average value of the data in C′, n is the number of data in C′, the preset upper limit of experience, C′ is the normalized experience, C min and C max are the minimum and maximum values ​​in experience C respectively.

[0049] A second aspect of the present invention provides a vehicle path planning system supporting multiple stations and customer priorities, comprising:

[0050] The objective function module is used to construct an objective function based on customer information and station information, cluster and group customers using the K-means algorithm, and construct a global optimal solution for the objective function based on customer priority and clustering results; wherein the global optimal solution includes multiple paths, each path represents the delivery sequence of a vehicle, so as to facilitate subsequent optimization of the global optimal solution;

[0051] An optimization module is used to construct a random number and compare the random number with the size of a preset experience feedback parameter;

[0052] If the random number is smaller than the preset experience feedback parameter, the development process is executed to obtain the local optimal solution of the objective function; if the local optimal solution is smaller than the global optimal solution, the global optimal solution is updated and the local optimal solution is added to the preset experience learning array; the development process is conducive to improving the optimization ability of the objective function;

[0053] If the random number is greater than the preset experience feedback parameter, the exploration process is executed to obtain the local optimal solution of the objective function, and the global optimal solution is directly updated, and the local optimal solution is added to the preset experience learning array; the exploration process makes it easy to jump out of the local optimal solution;

[0054] The first iteration module is used to repeatedly perform the optimization module until the data in the experience learning array reaches a preset experience upper limit;

[0055] The second iteration module is used to determine whether the preset maximum number of iterations has been reached. If the preset maximum number of iterations has not been reached, the experience array is used to update the preset experience feedback parameters and clear all data in the experience array, and the above optimization module, the first iteration module and the second iteration module are repeated for iteration;

[0056] If the preset maximum number of iterations is reached, the current global optimal solution is output and the path planning for the relief material transport vehicle is completed.

[0057] Beneficial effects of the present invention:

[0058] The present invention combines the characteristics of customer priority and uses a learning strategy to indicate the direction of subsequent steps based on historical information, thereby avoiding falling into a local optimum or failure to converge. The repair strategy of infeasible solutions based on capacity is integrated into the operator's acceptance of new solutions, enabling it to explore a larger solution space, expanding its acceptance of new solutions, simplifying the acceptance of solutions, and strengthening the search for local optimal solutions during the development process, thereby improving the optimization ability of the development process. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 One of the flowcharts of a vehicle path planning method supporting multiple stations and customer priorities provided in an embodiment of the present invention.

[0060] Figure 2 The present invention provides a second flowchart of a vehicle path planning method supporting multiple stations and customer priorities according to an embodiment of the present invention.

[0061] Figure 3 A flowchart of a development process provided for an embodiment of the present invention.

[0062] Figure 4 A flowchart of the exploration process provided by an embodiment of the present invention.

[0063] Figure 5 An architecture diagram of a vehicle path planning system supporting multiple stations and customer priorities provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] Example 1

[0066] like Figure 1 and Figure 2 As shown, a vehicle path planning method supporting multiple stations and customer priorities includes:

[0067] S101: construct an objective function based on customer information and station information, cluster and group customers using the K-means algorithm, and construct a global optimal solution for the objective function based on customer priorities and clustering results. The global optimal solution includes multiple paths, each of which represents the delivery order of a vehicle.

[0068] Specifically, the order in which each customer on a solution is visited is recorded. This order is not affected by different stations and vehicles. Each customer is traversed once to determine whether there are customers with lower priority than this customer among all customers with lower order. If there are customers, penalty time is calculated for these customers, even if these customers are on different paths. The sum of the penalty time obtained at the end is the total penalty time.

[0069] The objective function is expressed as follows:

[0070]

[0071] Where F is the objective function, is the time from the corresponding station to customer point i, I ij The penalty time for customer i to be served before customer j, p ij is a binary variable, The time from the corresponding station to customer point j, R is the total number of paths, V′ is the total number of customers, and k is the kth path.

[0072] The K-means clustering algorithm is used to cluster the coordinates of all customer nodes (users who need relief supplies) in a given area, and the number of groups is the number of stations. The center coordinates of each cluster represent the overall coordinates of the group of customers, and they are assigned to the nearest station service. Each group of customers is sorted in descending order, and the rule for comparing the size is customer priority first, followed by customer demand.

[0073] Starting from the station assigned to each group of customers, visit them in order and add them to the current path. If the path capacity limit is exceeded, add a new path and add it to it. Each path represents a vehicle. If the total path is greater than the maximum number of vehicles, split and reduce the path, split the path with the smallest load among all the paths, and add each of the above customers to the remaining path with the smallest load in turn. Repeat the above operation until the total path is equal to the maximum number of vehicles or paths. The generated initial solution allows for paths that may exceed the load, but the total number of paths will not exceed the maximum number of vehicles.

[0074] S102: construct a random number, and compare the random number with a preset experience feedback parameter.

[0075] If the random number is smaller than the preset experience feedback parameter, the development process is executed to obtain the local optimal solution of the objective function; if the local optimal solution is smaller than the global optimal solution, the global optimal solution is updated and the local optimal solution is added to the preset experience learning array.

[0076] If the random number is greater than the preset experience feedback parameter, the exploration process is executed to obtain the local optimal solution of the objective function, and the global optimal solution is directly updated, and the local optimal solution is added to the preset experience learning array.

[0077] Specifically, local search is a simple and effective random method with wide applicability and has been proven to provide good solutions in various VRP fields. As a mature heuristic algorithm, local search has derived many variants after decades of development. Variable Neighborhood Descent (VND) is an algorithm based on variable neighborhood local search. It continuously changes the neighborhood structure and tries to find the local optimal solution in each neighborhood, so as to continuously jump out of the local optimal solution during the search process and move towards a better solution. In the field of vehicle path planning, algorithms that use local search can find the best solution more effectively than those that do not.

[0078] Therefore, the present invention uses local search to find the global optimal solution. Specifically, the preset experience feedback parameter α is 0.6. A random number r ranging from 0 to 1 is generated. If r<α, the development process is executed, otherwise the exploration process is executed. Only one of the processes can be executed in each iteration.

[0079] The development process has the ability to quickly repair the solution and extremely high optimization ability, using the destruction and reconstruction optimization operator and the variable neighborhood descent algorithm. In this process, the destruction and reconstruction optimization local search operator is first used for the current global optimal solution, and then the station allocation operator is used to reallocate the stations, and finally the variable neighborhood descent algorithm is executed. In the variable neighborhood process, 2-Opt is used once for path regularization and station exchange operators after each update of the neighborhood solution.

[0080] like Figure 3 The development process shown specifically includes:

[0081] S102.1.1: Apply the destruction and reconstruction local search operator to the current global optimal solution, where the destruction and reconstruction local search operator includes:

[0082] Destruction phase: randomly select a customer from the current global optimal solution (S0) and remove the N customers closest to it from the solution to obtain a destroyed solution, where N is no more than half of all customers.

[0083] Repair phase: Select the removed customers in a random order and insert them into the damaged solution. Specifically, insert them into the path that minimizes the increase in the objective function value according to the principle of satisfying the customer priority on the path. Until all customers are inserted into the appropriate position, the repaired solution is obtained.

[0084] Compare whether the repaired solution is better than the current global optimal solution (the value of the repaired solution is less than the value of the current global optimal solution). If so, the repaired solution is output as the current global optimal solution. If not, repeat the repair phase until the condition is met. If the number of repetitions reaches the maximum number of attempts, the last repaired solution is forced to be output as the current global optimal solution.

[0085] S102.1.2: Apply the station allocation operator to the current global optimal solution, where the station allocation operator includes:

[0086] Remove all stations on the current global optimal solution path, and assign the station closest to the first customer on each path of the current global optimal solution to the path. Output the solution after assigning stations.

[0087] S102.1.3: Define multiple neighborhood structures, neighborhood global optimal solutions S0 and neighborhood local optimal solutions S1, and save the neighborhood global optimal solution S0 as the neighborhood local optimal solution S1, where the initial values ​​of the neighborhood global optimal solution S0 and the neighborhood local optimal solution S1 are both the solutions after the stations are allocated.

[0088] The neighborhood structure includes one-point movement, two-point exchange, point-edge exchange, edge exchange and OrOpt. Among them, one-point movement includes moving the selected customer to another location, two-point exchange includes exchanging the positions of two selected customers, point-edge exchange includes exchanging the positions of one customer on the path with the other two customers, edge exchange includes exchanging the positions of two customers on the path with the other two customers, and OrOpt includes moving the selected continuous 2-4 customers to other locations.

[0089] S102.1.4: Determine whether the current neighborhood structure is the last one. If so, execute step S102.1.12; if not, execute step S102.1.5.

[0090] S102.1.5: Perform a local search for the current neighborhood solution within the current neighborhood structure to obtain neighborhood solution S2.

[0091] Specifically, the neighborhood local optimal solution S1 is transformed according to the current neighborhood structure (such as one-point movement, two-point exchange, point-edge exchange, edge exchange, and OrOpt) to obtain solution S2.

[0092] Preferably, the local search also includes a capacity-based infeasible solution acceptance strategy, which includes accepting feasible solutions with lower function values ​​and always giving priority to them. The capacity-based acceptance strategy has a repairing effect on infeasible solutions that violate capacity constraints. The specific process is expressed by the following formula:

[0093]

[0094] Among them, F(s) is the feasible objective function, Q(s) is the acceptance function when the capacity constraint is violated, is the sum of the squares of the demands of all customers on path k, and R is the total number of paths.

[0095] S102.1.6: Determine whether solution S2 is better than the neighborhood local optimal solution S1. If so, execute S102.7; if not, execute S102.1.9.

[0096] S102.1.7: Execute the 2-opt path regularization and station exchange operator on solution S2. Specifically, execute 2-Opt once on each path on S2, and then exchange the different stations on every two paths of solution S2. If the objective function value of the new solution after the exchange is reduced, accept solution S3.

[0097] S102.1.8: Update the neighborhood local optimal solution S1 with solution S3.

[0098] S102.1.9: Determine whether the neighborhood local optimal solution S1 is better than the neighborhood global optimal solution S0. If so, execute S102.1.10; if not, execute S102.1.11.

[0099] S102.1.10: Update S0 with S1 and jump back to the first neighborhood and execute S102.1.4.

[0100] S102.1.11: Jump to the next neighborhood structure and execute S102.1.4.

[0101] S102.1.12: Output the current neighborhood global optimal solution S0, where the neighborhood global optimal solution S0 is a local optimal solution.

[0102] Among them, S102.1.3 to S102.1.12 are the variable neighborhood descent process.

[0103] The exploration process uses the projectile chain perturbation operator to perturb the current global optimal solution. The exploration process is as follows: Figure 4 As shown, the specific steps are as follows:

[0104] S102.2.1: All paths on the current global optimal solution (S0) are numbered in sequence, starting from 1 and in random order.

[0105] S102.2.2: Set the path iteration k of the projectile chain to 1, and R to kmax as the number of paths.

[0106] S102.2.3: Randomly select a customer on the path numbered k to exchange positions with a random customer on the successor path numbered k+1. The exchanged path may violate the capacity constraint.

[0107] S102.2.4: Serial number k is incremented by 1.

[0108] S102.2.5: Determine whether the sequence number k is the last sequence number. If so, execute S102.2.6. If not, return to S102.2.3 to continue the next path.

[0109] S102.2.6: Perform the same operation as S102.2.3 on the last path and the first path, randomly swapping the positions of the two points.

[0110] S102.2.7: Update the current solution S0.

[0111] S102.2.8: Determine whether the termination condition is met. If so, execute S102.2.9. If not, return to S102.2.2 and repeat once to enhance the disturbance effect.

[0112] S102.2.9: Output the perturbed solution S0, where the perturbed solution S0 is the local optimal solution.

[0113] The exploration process tries to escape from the local optimal solution as much as possible through perturbations.

[0114] S103: Repeat S102 until the data in the experience learning array reaches a preset experience upper limit.

[0115] Specifically, the preset experience cap is 20.

[0116] S104: Determine whether the preset maximum number of iterations has been reached. If the preset maximum number of iterations has not been reached, use the experience array to update the preset experience feedback parameters and clear all data in the experience array, and repeat the above S102, S103 and S104 to iterate.

[0117] Specifically, updating the preset experience feedback parameters specifically includes the following process:

[0118] Normalize the experience C, the formula is:

[0119]

[0120] Among them, C′ is the normalized experience, C min and C max are the minimum and maximum values ​​in experience C respectively.

[0121] For the normalized experience C′, the standard deviation σ(C′) is calculated by the following formula. The specific process is expressed by the following formula:

[0122]

[0123] Among them, σ(C′) is the normalized empirical standard deviation, n is the number of data in C′, is the average value of the data in C′, C′ i is the normalized value of the ith experience.

[0124] Finally, update the experience feedback parameter according to the following formula:

[0125]

[0126] Among them, α is the experience feedback parameter.

[0127] If the preset maximum number of iterations is reached, the current global optimal solution is output and the path planning for the relief material transport vehicle is completed.

[0128] Specifically, the preset maximum number of iterations is 20.

[0129] The present invention constructs an objective function and obtains an initial global optimal solution based on the K-means algorithm and customer priority, and then optimizes the global optimal solution through an adaptive learning strategy (two cycles include an experience learning array cycle and an iteration cycle) to obtain the final global optimal solution and complete the path planning of the rescue material transportation vehicle. Among them, the adaptive learning strategy is divided into two iteration cycles of different nested sizes. The small cycle is nested in the large cycle. The number of iterations of the large cycle is 20, which is the iterative process of the present invention, and the number of iterations of the small cycle is 20, which is the iterative process of the experience learning strategy. Among them, each iteration in the small cycle adds the result to the experience array C, and at the end of the iteration, that is, the experience array C reaches the maximum C max When the learning feedback parameter α and the experience array C are updated, the number of iterations of the large cycle is increased by one, and then the small cycle is entered again to continue iteration. Until the termination condition is met. The adaptive learning strategy provided by the present invention can effectively accelerate the convergence of the algorithm, improve the optimization ability of the algorithm, and prevent the algorithm from falling into the local optimal solution. Among them, the introduction of customer priority is more in line with actual needs and can minimize the casualties and pain of the victims.

[0130] Example 2

[0131] Based on the above embodiments, Figure 5 As shown, an embodiment of the present invention provides a vehicle path planning system supporting multiple stations and customer priorities, including:

[0132] The objective function module is used to construct an objective function based on customer information and station information, cluster and group customers using the K-means algorithm, and construct a global optimal solution for the objective function based on customer priority and clustering results; wherein the global optimal solution includes multiple paths, each path representing the delivery order of a vehicle.

[0133] The optimization module is used to construct a random number and compare the random number with the size of a preset experience feedback parameter.

[0134] If the random number is smaller than the preset experience feedback parameter, the development process is executed to obtain the local optimal solution of the objective function; if the local optimal solution is smaller than the global optimal solution, the global optimal solution is updated and the local optimal solution is added to the preset experience learning array.

[0135] If the random number is greater than the preset experience feedback parameter, the exploration process is executed to obtain the local optimal solution of the objective function, and the global optimal solution is directly updated, and the local optimal solution is added to the preset experience learning array.

[0136] The first iteration module is used to repeatedly perform the optimization module until the data in the experience learning array reaches a preset experience upper limit.

[0137] The second iteration module is used to determine whether the preset maximum number of iterations has been reached. If the preset maximum number of iterations has not been reached, the experience array is used to update the preset experience feedback parameters and clear all data in the experience array, and the above optimization module, the first iteration module and the second iteration module are repeated to iterate.

[0138] If the preset maximum number of iterations is reached, the current global optimal solution is output and the path planning for the relief material transport vehicle is completed.

[0139] It should be noted that the vehicle path planning system supporting multiple stations and customer priorities provided in an embodiment of the present invention is for realizing the above-mentioned vehicle path planning method supporting multiple stations and customer priorities. Its specific functions can be referred to the above-mentioned method embodiments and will not be repeated here.

[0140] In summary, the present invention combines the characteristics of customer priority and uses a learning strategy to indicate the direction of subsequent steps based on historical information, thereby avoiding falling into a local optimum or failure to converge. The repair strategy of infeasible solutions based on capacity is integrated into the operator's acceptance of new solutions, enabling it to explore a larger solution space, expanding its acceptance of new solutions, simplifying the acceptance of solutions, and strengthening the search for local optimal solutions during the development process, thereby improving the optimization ability of the development process.

[0141] 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 it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A vehicle path planning method supporting multiple stations and customer priorities, characterized in that: include: Step 1: construct an objective function based on customer information and station information, cluster and group customers using the K-means algorithm, and construct a global optimal solution for the objective function based on customer priority and clustering results; wherein the global optimal solution includes multiple paths, each path representing the delivery order of a vehicle; Step 2: Construct a random number and compare the random number with the size of the preset experience feedback parameter; If the random number is smaller than the preset experience feedback parameter, the development process is executed to obtain the local optimal solution of the objective function; if the local optimal solution is smaller than the global optimal solution, the global optimal solution is updated and the local optimal solution is added to the preset experience learning array; If the random number is greater than the preset experience feedback parameter, the exploration process is executed to obtain the local optimal solution of the objective function, and the global optimal solution is directly updated, and the local optimal solution is added to the preset experience learning array; Step 3: Repeat step 2 until the data in the experience learning array reaches the preset experience upper limit; Step 4: Determine whether the preset maximum number of iterations has been reached. If the preset maximum number of iterations has not been reached, use the experience array to update the preset experience feedback parameters and clear all data in the experience array, and repeat the above steps 2, 3 and 4 to iterate; If the preset maximum number of iterations is reached, the current global optimal solution is output and the path planning for the relief material transport vehicle is completed.

2. A vehicle path planning method supporting multiple stations and customer priorities according to claim 1, characterized in that: The objective function is expressed as follows: Where F is the objective function, is the time from the corresponding station to customer point i, I ij The penalty time for customer i to be served before customer j, p ij is a binary variable, The time from the corresponding station to customer point j, R is the total number of paths, V′ is the total number of customers, and k is the kth path.

3. A vehicle path planning method supporting multiple stations and customer priorities according to claim 1, characterized in that: The development process specifically includes: S301, using a destruction and reconstruction local search operator to process the global optimal solution and update the global optimal solution; S302, using a station allocation operator to process the updated global optimal solution to obtain a solution after station allocation; S303, defining multiple neighborhood structures, neighborhood global optimal solutions S0 and neighborhood local optimal solutions S1, and saving the neighborhood global optimal solution S0 as the neighborhood local optimal solution S1, wherein the initial values ​​of the neighborhood global optimal solution S0 and the neighborhood local optimal solution S1 are both solutions after the station is allocated; A plurality of said neighborhood structures include one-point movement, two-point exchange, point-edge exchange, edge exchange and OrOpt; One-point move includes moving the selected customer to other locations, two-point exchange includes exchanging the positions of two selected customers, point-edge exchange includes exchanging the positions of one customer on the path with two other customers, edge exchange includes exchanging the positions of two customers on the path with two other customers, and OrOpt includes moving the selected consecutive 2-4 customers to other locations; S304, judging the current neighborhood structure, if the current neighborhood structure is the last neighborhood structure, executing S310, if the current neighborhood structure is not the last neighborhood structure, performing a local search for the neighborhood local optimal solution S1 on the current neighborhood structure to obtain a solution S2; S305, if the solution S2 is smaller than the neighborhood local optimal solution S1, execute S306; if the solution S2 is larger than the neighborhood local optimal solution S1, execute S307; S306, using 2-opt and the station exchange operator to process the solution S2 to obtain a solution S3, and using the solution S3 as the new neighborhood local optimal solution S1, and then returning to S304; wherein 2-opt includes flipping the order of customers on the same path, and the station exchange operator includes exchanging different stations on every two paths in the solution S2; S307, compare the neighborhood global optimal solution S0 with the neighborhood local optimal solution S1, if the neighborhood local optimal solution S1 is smaller than the neighborhood global optimal solution S0, then execute S308, if the neighborhood local optimal solution S1 is larger than the neighborhood global optimal solution S0, then enter S309; S308, updating the neighborhood global optimal solution S0 according to the neighborhood local optimal solution S1, and returning to execute S304; S309, executing S304 on the next neighborhood structure; S310, outputting a neighborhood global optimal solution S0; wherein the neighborhood global optimal solution S0 is a local optimal solution.

4. A vehicle path planning method supporting multiple stations and customer priorities according to claim 3, characterized in that: The method of processing the global optimal solution by using the destruction and reconstruction local search operator to update the global optimal solution specifically includes: Randomly select a customer from the global optimal solution, remove a preset number of customers closest to the customer from the solution, and obtain a destroyed solution; Insert the removed customers according to their priorities in random order to the path that minimizes the increase in the objective function value, and obtain the repaired solution; If the repaired solution is smaller than the global optimal solution, the repaired solution is used as the global optimal solution and the global optimal solution is updated. If the repaired solution is larger than the global optimal solution, the above process is repeated. After reaching the preset maximum number of updates, the last repaired solution is used as the global optimal solution and the global optimal solution is updated.

5. A vehicle path planning method supporting multiple stations and customer priorities according to claim 3, characterized in that: The updated global optimal solution is processed by the station allocation operator to obtain a solution after station allocation, specifically including: Remove all stations on the updated optimal solution path of the entire set, assign the station closest to the first customer on each path to the path, and obtain the solution after assigning the stations.

6. A vehicle path planning method supporting multiple stations and customer priorities according to claim 3, characterized in that: The local search includes transforming the neighborhood local optimal solution S1 according to the current neighborhood structure to obtain a solution S2.

7. A vehicle path planning method supporting multiple stations and customer priorities according to claim 6, characterized in that: The local search also includes a capacity-based infeasible solution acceptance strategy, which includes accepting a solution with a lower function value and always giving priority to a feasible solution. The specific process is expressed by the following formula: Among them, F(s) is the feasible objective function, Q(s) is the acceptance function when the capacity constraint is violated, is the sum of the squares of the demands of all customers on path k, and R is the total number of paths.

8. The vehicle path planning method supporting multiple stations and customer priorities according to claim 1 is characterized in that: The exploration process specifically includes: Randomly label all paths in the global optimal solution with k, where k = 1, 2, ..., R, and R is the total number of paths; A customer on the path labeled k is randomly selected to exchange positions with a customer on the path labeled k+1, and the exchange operation is repeated until all paths have completed the customer position exchange; the path labeled R also needs to exchange customer positions with the path labeled 1 to obtain the perturbed solution; Among them, the solution after perturbation is the local optimal solution.

9. A vehicle path planning method supporting multiple stations and customer priorities according to claim 1, characterized in that: The updating preset experience feedback parameter is expressed by the following formula: Among them, α is the experience feedback parameter, σ(C′) is the standard deviation of the normalized experience, and C′ i is the normalized value of the ith experience, is the average value of the data in C′, n is the number of data in C′, the preset upper limit of experience, C′ is the normalized experience, C min and C max are the minimum and maximum values ​​in experience C respectively.

10. A vehicle path planning system supporting multiple stations and customer priorities, characterized in that: include: The objective function module is used to construct an objective function based on customer information and station information, cluster and group customers using the K-means algorithm, and construct a global optimal solution for the objective function based on customer priorities and clustering results; wherein the global optimal solution includes multiple paths, each path representing the delivery order of a vehicle; An optimization module is used to construct a random number and compare the random number with the size of a preset experience feedback parameter; If the random number is smaller than the preset experience feedback parameter, the development process is executed to obtain the local optimal solution of the objective function; if the local optimal solution is smaller than the global optimal solution, the global optimal solution is updated and the local optimal solution is added to the preset experience learning array; If the random number is greater than the preset experience feedback parameter, the exploration process is executed to obtain the local optimal solution of the objective function, and the global optimal solution is directly updated, and the local optimal solution is added to the preset experience learning array; The first iteration module is used to repeatedly perform the optimization module until the data in the experience learning array reaches a preset experience upper limit; The second iteration module is used to determine whether the preset maximum number of iterations has been reached. If the preset maximum number of iterations has not been reached, the experience array is used to update the preset experience feedback parameters and clear all data in the experience array, and the above optimization module, the first iteration module and the second iteration module are repeated for iteration; If the preset maximum number of iterations is reached, the current global optimal solution is output and the path planning for the relief material transport vehicle is completed.

Images