Photoelectric hybrid solving method and system for vehicle path planning problem

Abstract

The invention relates to a photoelectric hybrid solving method and system for a vehicle path planning problem. Firstly, customer points in a vehicle path planning problem are divided into a plurality of clusters according to the number of vehicles by a clustering method with capacity consideration, then travel salesman problems in each cluster are converted into Isinus problems, and corresponding adjacency matrixes and outfield vectors are solved for each Isinus problem; and finally, aiming at each Isinus problem, according to the corresponding adjacent matrix and the external field vector, performing parallel solving by using each optical Isinus machine to obtain an optimal vehicle path in each cluster.

Description

Technical Field

[0001] The present invention relates to an optoelectronic hybrid solution method and an optoelectronic hybrid solution system for vehicle routing problems. Background Art

[0002] The vehicle routing problem (VRP) refers to the problem of how to effectively allocate a group of vehicles to visit multiple customer points and minimize the total driving distance or cost under the satisfaction of constraint conditions. Considering that each customer point has a certain demand and each vehicle also has a capacity limit, the capacitated vehicle routing problem (CVRP) has very important practical significance and is widely applied in the fields of logistics, distribution, transportation, etc. However, since problems such as CVRP are non-deterministic polynomial time (NP) solvable problems, solving the CVRP problem faces huge complexity challenges. The huge complexity of solving CVRP has prompted researchers to develop effective methods for solving CVRP. Currently, heuristic algorithms are the most effective methods for solving large-scale CVRP, and such methods include simulated annealing algorithms, tabu search algorithms, genetic algorithms, etc.

[0003] In recent years, optical computing methods for solving the maximum cut, an NP problem, using optoelectronic hybrid systems have been greatly developed, including spatial light modulation Ising machines, coherent Ising machines, and photonic neural network Ising machines based on cascaded Mach-Zehnder interferometers (MZIs). These computing systems make full use of the advantages of light, such as parallelism, low latency, and low energy consumption, and provide a way to solve combinatorial optimization problems that is completely different from classical electronic computing. They have achieved performance beyond classical computers (running specific heuristic algorithms) in solving the maximum cut, an NP problem. Summary of the Invention

[0004] However, as the electronic chip process approaches the physical limit, it is difficult to further improve the computing power and memory bandwidth of classical electronic computers, while the power consumption brought by the von Neumann architecture continues to increase. This makes classical electronic computing face problems of slow convergence speed and high energy consumption when solving large-scale CVRP problems. In addition, although great development has been achieved in using optoelectronic hybrid systems to solve the maximum cut, an NP problem, for large-scale CVRP problems, there is still a lack of corresponding research to show whether optical computing methods can be used for solving.

[0005] The present invention is proposed in view of the above situation, and its purpose is to provide an optoelectronic hybrid solution method for the vehicle routing problem. First, the customer points in the vehicle routing problem are grouped into multiple clustering clusters according to the number of vehicles by a clustering method considering capacity. Then, the traveling salesman problems within each clustering cluster are respectively converted into Ising problems, and the corresponding adjacency matrix and external field vector are obtained for each Ising problem. Finally, for each Ising problem, parallel solutions are respectively performed using each optical Ising machine according to the corresponding adjacency matrix and external field vector to obtain the optimal vehicle route within each clustering cluster.

[0006] Technical solution for solving technical problems

[0007] In order to solve the above problems, in the optoelectronic hybrid solution method for the vehicle routing problem according to the first aspect of the present invention, the following steps are included:

[0008] Clustering step, in this clustering step, the customer points are grouped into multiple clustering clusters according to the number of vehicles, so that the distances between the customer points in each clustering cluster are respectively less than or equal to a predetermined threshold, and the sum of the demands of the customer points in each clustering cluster is less than or equal to the capacity of each vehicle;

[0009] Construction step, in this construction step, the traveling salesman problem models within each clustering cluster are respectively converted into Ising problem models, and the corresponding adjacency matrix and external field vector are obtained for each Ising problem model, wherein the adjacency matrix is quantified according to the accuracy of the optical Ising machine; and

[0010] Parallel solution step, in this parallel solution step, for each Ising problem model constructed in the construction step, parallel solutions are respectively performed using the optical Ising machine according to the corresponding adjacency matrix and external field vector, and the path matching the optimal solution of the optical Ising machine is used as the optimal vehicle route within each clustering cluster.

[0011] Furthermore, the predetermined threshold is obtained according to the magnitude relationship of the distances between the customer points.

[0012] Furthermore, the clustering step includes the following steps:

[0013] (S1) Set the total number of all customer points as N, set the demand of the i-th customer point as d i , set the upper limit of the load capacity of each vehicle as Q, and determine that the minimum number of vehicles required is K;

[0014] (S2) Select the coordinates of K customer points as the initial coordinate values of the centers of each clustering cluster, and set the capacity of each clustering cluster as the upper limit Q of the load capacity of the vehicle;

[0015] (S3) Allocate the customer points to the corresponding clustering clusters according to the demand quantities of each customer point;

[0016] (S4) After the allocation of all customer points is completed, recalculate and update the central coordinate values of each of the K clustering clusters; and

[0017] (S5) Determine whether the difference between the central coordinates of the new clustering clusters and those of the original clustering clusters is greater than a set threshold. If it is greater than the set threshold, transfer to step (S3); otherwise, the clustering step ends.

[0018] Further, the minimum number of vehicles required K = [Σ i N d i / Q] + 1, where [Σ i N d i / Q] is the largest integer not exceeding Σ i N d i / Q.

[0019] Further, the construction step includes the following steps:

[0020] (S6) Assign weights to the corresponding edges according to the distances between the customer points in each clustering cluster;

[0021] (S7) For each customer point in each clustering cluster, introduce a number of binary variables corresponding to (e.g., equal to) the square of the number of customer points in each clustering cluster, give the objective function of the constraint and the objective function of the total weight, and establish K quadratic unconstrained binary optimization problem models; and

[0022] (S8) Convert the K quadratic unconstrained binary optimization problem models into K Ising problem models, and respectively obtain the corresponding adjacency matrices and external field vectors for the K Ising problem models.

[0023] Further, the parallel solution step includes the following steps:

[0024] (S9) Set the K optical Ising machines according to the K adjacency matrices and external field vectors;

[0025] (S10) Use the K set optical Ising machines to respectively perform parallel recursive solutions for the K Ising problem models;

[0026] (S11) Set the current optimal solution of each optical Ising machine as Best_x k , set the current optimal energy of each optical Ising machine as Best_H k , and set the variable value of the t-th iteration of each optical Ising machine as x k(t), set the energy of the t-th round of iteration of each optical Ising machine to H k (t), for each optical Ising machine, calculate the energy H k (t) and compare it with the current optimal energy Best_H k If H k (t) < Best_H k and the current solution result of the optical Ising machine satisfies the condition of the objective function H k A = 0, then let Best_H k = H k (t), Best_x k = x k (t), otherwise Best_H k remains unchanged, and the iteration number t is incremented by 1;

[0027] (S12) For each optical Ising machine, determine whether the iteration number t is less than the maximum iteration number Nmax. If t is less than Nmax, then return to step (S10); otherwise, output the optimal solution Best_x k and the optimal energy Best_H k .

[0028] In the optoelectronic hybrid solution system for the vehicle path planning problem according to the second aspect of the present invention, it includes:

[0029] A clustering unit that groups customer points into multiple clustering clusters according to the number of vehicles, such that the distances between customer points in each clustering cluster are respectively less than or equal to a predetermined threshold, and the demands of all customer points in each clustering cluster are less than or equal to the capacity of each vehicle;

[0030] A construction unit that respectively converts the traveling salesman problem models within each clustering cluster into Ising problem models, and respectively obtains the corresponding adjacency matrix and external field vector for each Ising problem model, where the adjacency matrix is quantized according to the accuracy of the optical Ising machine; and

[0031] A parallel solution unit that respectively performs parallel solutions for each Ising problem constructed in the construction unit according to the corresponding adjacency matrix and external field vector and using the optical Ising machine, and takes the path matching the optimal solution of the optical Ising machine as the optimal vehicle path within each clustering cluster.

[0032] Furthermore, the predetermined threshold is obtained according to the magnitude relationship of the distances between customer points.

[0033] Furthermore, the clustering unit performs the following steps:

[0034] (S1) Set the total number of all customer points to N, set the demand of the i-th customer point to d i , set the upper limit of the load capacity of each vehicle to Q, and determine that the minimum number of vehicles required is K;

[0035] (S2) Select the coordinates of K customer points as the initial coordinate values of the centers of each clustering cluster, and set the capacity of each clustering cluster to the upper limit Q of the load capacity of the vehicle;

[0036] (S3) Allocate the customer points to the corresponding clustering clusters according to the demands of each customer point;

[0037] (S4) After all customer points are allocated, recalculate and update the center coordinate values of each of the K clustering clusters; and

[0038] (S5) Determine whether the difference between the center coordinates of the new clustering cluster and the center coordinates of the original clustering cluster is greater than the set threshold. If it is greater than the set threshold, transfer to step (S3); otherwise, the clustering step ends.

[0039] Further, the minimum number of vehicles required K = [Σ i N d i / Q] + 1, where [Σ i N d i / Q] is the largest integer not exceeding Σ i N d i / Q.

[0040] Further, the construction unit performs the following steps:

[0041] (S6) Assign weights to the corresponding edges according to the distances between the customer points in each clustering cluster;

[0042] (S7) For each customer point in each clustering cluster, introduce a number of binary variables corresponding to the square of the number of customer points in each clustering cluster, give the objective function of the constraint and the objective function of the total weight, and establish K quadratic unconstrained binary optimization problem models; and

[0043] (S8) Convert the K quadratic unconstrained binary optimization problem models into K Ising problem models, and find the corresponding adjacency matrices and external field vectors for the K Ising problem models respectively.

[0044] Further, the parallel solving unit performs the following steps:

[0045] (S9) Set the K optical Ising machines according to the K adjacency matrices and external field vectors;

[0046] (S10) Use K set optical Ising machines to recursively solve K Ising problem models in parallel;

[0047] (S11) Set the current optimal solution of each optical Ising machine as Best_x k , set the current optimal energy of each optical Ising machine as Best_H k , set the variable value of the t-th iteration of each optical Ising machine as x k (t), set the energy of the t-th iteration of each optical Ising machine as H k (t), for each optical Ising machine, calculate the energy H k (t) and compare it with the current optimal energy Best_H k . If H k (t) < Best_H k and the current solution result of the optical Ising machine satisfies the condition of the objective function H k A = 0, then let Best_H k = H k (t), Best_x k = x k (t), otherwise Best_H k remains unchanged, and the iteration number t is incremented by 1;

[0048] (S12) For each optical Ising machine, determine whether the iteration number t is less than the maximum iteration number Nmax. If t is less than Nmax, return to step (S10), otherwise output the optimal solution Best_x k and the optimal energy Best_H k .

[0049] Invention Effect

[0050] According to the optoelectronic hybrid solution method and system for the vehicle routing problem of the present invention, the advantages of high parallelism and low latency of the optical Ising machine in solving the quadratic unconstrained binary optimization problem are fully utilized. Therefore, compared with the traditional electronic computing-based solution method, it can greatly improve the solution speed and significantly reduce the energy consumption. Description of the Drawings

[0051] Figure 1 is a schematic diagram showing the overall process of the optoelectronic hybrid solution method for the vehicle routing problem involved in the embodiment of the present invention.

[0052] Figure 2 is a schematic diagram showing the detailed content of the overall process of the optoelectronic hybrid solution method for the vehicle routing problem involved in the embodiment of the present invention.

[0053] Figure 3 It is a schematic block diagram showing the overall structure of the optoelectronic hybrid solution system for the vehicle path planning problem involved in the embodiments of the present invention.

[0054] Figure 4 It is a schematic block diagram showing the computing architecture of the optical Ising machine in the optoelectronic hybrid solution system for the vehicle path planning problem involved in the embodiments of the present invention.

[0055] Label description

[0056] 801 Optical domain matrix device

[0057] 802 Threshold judgment device

[0058] 901 Clustering unit

[0059] 902 Construction unit

[0060] 903 Parallel solution unit Specific embodiments

[0061] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. These embodiments are implemented on the premise of the technical solution of the present invention, and detailed implementation methods and specific operation processes are given, but the protection scope of the present invention is not limited to the following embodiments.

[0062] For the sake of convenience of description, spatial relative terms such as "under", "below", "lower", "above", "upper", etc. may be used herein to describe the relationship of one element or feature relative to another element or feature as shown in the figures. It should be understood that the spatial relative terms are intended to include different orientations of the device used or operated in addition to the orientation shown in the figures. For example, if the device in the figure is flipped, the element described as "under" or "below" another element or feature will be oriented "above" another element or feature.

[0063] Unless otherwise defined, the terms used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the present invention belongs. The terms should be understood to have a meaning consistent with the context of the related art, and should not be understood in an idealized or overly formalized manner unless clearly defined herein.

[0064] <Optoelectronic Hybrid Solution Method for Vehicle Path Planning Problem>

[0065] Figure 1 It is a schematic overall flow diagram showing the optoelectronic hybrid solution method for the vehicle path planning problem involved in the embodiments of the present invention.

[0066] In the optoelectronic hybrid solution method for the vehicle path planning problem involved in the embodiments of the present invention, the following steps are included:

[0067] Clustering step ST11. In this clustering step, customer points are grouped into multiple clusters according to the number of vehicles, such that the distances between customer points in each cluster are respectively less than or equal to a predetermined threshold, and the sum of the demands of the customer points in each cluster is less than or equal to the capacity of each vehicle. In one embodiment, the predetermined threshold is dynamically variable and can be obtained based on the magnitude relationship of the distances between customer points. For example, it is determined by comparing the distances between customer points, so that the distances between customer points in each cluster are as small as possible.

[0068] Construction step ST12. In this construction step, the traveling salesman problem models within each of the clusters are respectively converted into Ising problem models, and corresponding adjacency matrices and external field vectors are respectively obtained for each of the Ising problem models. Among them, the adjacency matrix is quantized according to the accuracy of the optical Ising machine; and; and

[0069] Parallel solution step ST13. In this parallel solution step, for each of the Ising problem models constructed in the construction step, parallel solutions are respectively performed using the optical Ising machine according to the corresponding adjacency matrix and external field vector, and the path matching the optimal solution of the optical Ising machine is used as the optimal vehicle path within each cluster.

[0070] Figure 2 It is a schematic diagram showing the detailed content of the overall process of the optoelectronic hybrid solution method for the vehicle path planning problem involved in the embodiment of the present invention. Next, the detailed content of the overall process of the optoelectronic hybrid solution method for the vehicle path planning problem involved in the embodiment of the present invention will be specifically described.

[0071] First, in step S101, the total number of all customer points is set to N, all customer points are labeled as 1, 2, ···, N, the demand of the i-th customer point is set to d i , the upper limit of the load capacity of each vehicle is set to Q, and the number of vehicles required at least is calculated as K = [Σ i N d i / Q]+1, where [Σ i N d i / Q] is the largest integer not exceeding Σ i N d i / Q. Then, the coordinates of K customer points are randomly selected as the initial coordinate values of the centers of each cluster, and the capacity of each cluster is respectively set to the upper limit Q of the load capacity of the vehicle.

[0072] Then, in step S102, sort each customer point in descending order of demand, and sequentially assign the customer points to the corresponding clustering clusters. For a certain customer point i, the specific assignment process is as follows: Calculate the distance between customer point i and the centers of each clustering cluster, and determine whether the remaining capacity of the clustering cluster m meets the demand of customer point i when customer point i is assigned to the clustering cluster m with the smallest distance. If it meets the requirement, assign customer point i to clustering cluster m; otherwise, assign customer point i to the clustering cluster m' with the second-best distance and determine whether the remaining capacity of clustering cluster m' meets the demand of customer point i. Repeat the above steps until the assignment of all customer points is completed.

[0073] Then, in step S103, after the assignment of all customer points is completed, recalculate and update the center coordinate values of each of the K clustering clusters.

[0074] Then, in step S104, calculate the difference between the center coordinates of the new clustering cluster and the center coordinates of the original clustering cluster, and in step S105, determine whether this difference is greater than the set threshold. If it is greater than the set threshold, transfer to step (S3); otherwise, the clustering step ends. In addition, the above steps S101 - S105 correspond to clustering step ST11.

[0075] Then, in step S106, introduce a weighted graph G k =(V k , E k , W k ) to describe the path planning problem of each of the K clustering clusters. Among them, V k represents the set of all customer points in the k-th clustering cluster, E k represents the set of all drivable paths in the k-th clustering cluster, and each path e k i,j has a corresponding weight W k i,j , and W k i,j represents the distance between point i and point j or the weight of the edge.

[0076] Then, in step S107, use the modeling method of the quadratic unconstrained binary optimization (QUBO) problem for the path planning problem of the k-th clustering cluster to obtain the total objective function. Specifically, introduce a binary variable x k v,i , where v represents the vertex and i represents the order of the vertex in the path, that is, x k v,i = 1 means that the order of the v-th vertex in the path is i. Construct the following mathematical formula of the objective function, which represents the constraint on the binary variable:

[0077]

[0078] where A k is a constant greater than 0, and n k represents the number of all customer points in cluster k. The first term represents that each vertex can only be assigned to one sequence j, the second term represents that each sequence j contains only one customer point v, and the third term represents the penalty for the edges not existing in E k . When j = n k , j + 1 represents the first point. When all constraints are satisfied, H k A = 0. In addition, we also need to add an objective function representing the total weight:

[0079]

[0080] Then the total objective function is To ensure that all constraints are satisfied, the penalty coefficient satisfies the following mathematical relationship:

[0081] 0 < B k max(W k i,j ) < A k (Equation 3)

[0082] Since the highest term in the total objective function is a quadratic term about the binary variable x k v,i , the total objective function can be written in the form of the objective function of a QUBO problem through rearrangement:

[0083] H k = min[(x k ) T Q k x k (Equation 4)

[0084] where the binary variables x k v,i form a binary vector x k = (x k 1,1 , x k 1,2 ,..., x k nk,nk ).

[0085] Then, in step S108, Ising spin variables are introduced to convert the K quadratic unconstrained binary optimization problem models into K Ising problem models, and the corresponding adjacency matrices and external field vectors are obtained for the K Ising problem models respectively. Specifically, according to the objective function of the QUBO problem, the adjacency matrix M and the external field vector L to be set in the optical Ising machine are calculated. By introducing spin variables σ in each cluster kk i = 2x k i -1, the total objective function of clustering cluster k can be rewritten in the form of an Ising problem:

[0086] H k = min [-1 / 2Σ i,j M k i,j σ k i σ k j -Σ i L k i σ k i +1 / 4Tr(Q k )+1 / 4Σ i,j Q k i,j (Equation 5)

[0087] Among them, M k is the adjacency matrix of the Ising problem corresponding to clustering cluster k, with a dimension of n k ×n k , and i, j are the subscripts of the matrix elements of M k . In addition, let M k = -Q k / 2, and M k i,i = 0, L k is the external field vector of the Ising problem corresponding to clustering cluster k, with a dimension of n k ×1, and let L k i = -1 / 2Σ j Q k i,j . Among them, let i = 1, 2, 3 ···, n k , j = 1, 2, 3 ··· n k . Tr(Q k ) is to find the trace of the Q k matrix, that is, the sum of the diagonal elements. The last two terms are independent of the values of the spin variables and together are a constant. In addition, the above steps S106 - S108 correspond to the construction step ST12.

[0088] Then, in step S109, according to the K adjacency matrices and external field vectors, the K optical Ising machines are set. Figure 4 is a schematic block diagram showing the calculation architecture of the optical Ising machine in the optoelectronic hybrid solution system for the vehicle path planning problem involved in the embodiment of the present invention. Before using the optical Ising machine, it is necessary to according to the derived adjacency matrix M kand the external field vector L k Set the optical domain matrix device 801 and the threshold judgment device 802, where the threshold T k i can be calculated from the adjacency matrix M k and the external field vector L k as follows:

[0089] T k i =(Σ j M k i,j -L k i ) / 2 (Equation 6)

[0090] Then, in step S110, use the K set optical Ising machines to recursively solve the K Ising problem models in parallel. For each set optical Ising machine, its recursive solution algorithm can be expressed by the following mathematical formula:

[0091]

[0092] where x k (t) is the variable value of the t-th iteration of each optical Ising machine, and x k (t + 1) is the variable value of the (t + 1)-th iteration, is a Gaussian noise vector with a mean of 0 and a variance of s, used to represent the physical noise existing in the optical Ising machine, and sign(x) is the sign function (assuming that when x > 0, sign(x) = 1; when x ≤ 0, sign(x) = 0). Where x k (0) is a randomly generated initial value vector.

[0093] Then, in step S111, update the current optimal energy Best_H k and the current optimal solution Best_x k of each optical Ising machine, and check whether the condition that the objective function H k A = 0 is satisfied. Specifically, set the current optimal solution of each optical Ising machine as Best_x k , set the current optimal energy of each optical Ising machine as Best_H k , set the variable value of the t-th iteration of each optical Ising machine as x k (t), set the energy of the t-th iteration of each optical Ising machine as H k (t), for each optical Ising machine, calculate the energy H k (t) and compare it with the current optimal energy Best_H k . If H k(t) < Best_H k and the current solution result of the optical Ising machine satisfies the objective function H of the constraint k A = 0, then let Best_H k = H k (t), Best_x k = x k (t), otherwise Best_H k remains unchanged, and the iteration number t is incremented by 1.

[0094] Finally, in step S112, for each optical Ising machine, it is determined whether the iteration number t is less than the maximum iteration number Nmax. If t is less than Nmax, then return to step (S10). Otherwise, in step S113, the optimal solution Best_x k and the optimal energy Best_H k are output. Here, the optimal solution Best_x k is the optimal vehicle path within each cluster. In addition, the above steps S109 - S113 correspond to the parallel solution step ST13.

[0095] Thus, according to the optoelectronic hybrid solution method for the vehicle routing problem of the present invention, the advantages of high parallelism and low latency of the optical Ising machine in solving quadratic unconstrained binary optimization problems are fully utilized. Therefore, compared with the traditional electronic - computing - based solution method, it can greatly improve the solution speed and significantly reduce the energy consumption.

[0096] <Optoelectronic Hybrid Solution System for Vehicle Routing Problem>

[0097] Figure 3 is a schematic block diagram showing the overall structure of the optoelectronic hybrid solution system for the vehicle routing problem according to the embodiment of the present invention.

[0098] Corresponding to the optoelectronic hybrid solution method for the vehicle routing problem according to the embodiment of the present invention, as Figure 3 shown, in the optoelectronic hybrid solution system for the vehicle routing problem according to the embodiment of the present invention, it includes:

[0099] A clustering unit 901 that groups customer points into multiple clusters according to the number of vehicles, such that the distances between customer points in each cluster are respectively less than or equal to a predetermined threshold, and the demands of all customer points in each cluster are less than or equal to the capacity of each vehicle;

[0100] A building block 902 that respectively converts the traveling salesman problem models within each of the clustering clusters into Ising problem models, and respectively obtains corresponding adjacency matrices and external field vectors for each of the Ising problem models, where the adjacency matrices are quantized according to the precision of the optical Ising machine; and

[0101] A parallel solving unit 903 that respectively performs parallel solving on each of the Ising problems constructed in the building block according to the corresponding adjacency matrices and external field vectors and using the optical Ising machine, and takes the path that matches the optimal solution of the optical Ising machine as the optimal vehicle path within each of the clustering clusters.

[0102] Thus, the optoelectronic hybrid solving system for the vehicle routing problem according to the present invention makes full use of the advantages of the optical Ising machine in solving quadratic unconstrained binary optimization problems, such as high parallelism and low latency. Therefore, compared with the traditional electronic computing-based solving method, it can greatly improve the solving speed and significantly reduce the energy consumption.

[0103] It should be understood that the above description is illustrative and not restrictive. For example, the above embodiments (and / or aspects thereof) can be used in combination with each other. In addition, many modifications can be made without departing from the scope of the present invention to adapt a specific situation or material to the teachings of various embodiments of the present invention. Although the size and type of the materials described herein are used to define the parameters of various embodiments of the present invention, each embodiment is not meant to be restrictive, but rather an exemplary embodiment. Many other embodiments will be apparent to those skilled in the art upon reading the above description. Therefore, the scope of each embodiment of the present invention should be determined with reference to the appended claims and the full scope of the equivalent forms claimed thereby.

[0104] Industrial applicability

[0105] The optoelectronic hybrid solving method and system for the vehicle routing problem of the present invention can be applied to solve large-scale capacitated vehicle routing problems (CVRP).

Claims

1. An optoelectronic hybrid solution method for vehicle path planning problems, characterized in that, It includes the following steps: Clustering step: In this clustering step, customer points are grouped into multiple clusters according to the number of vehicles, such that the distances between customer points within each cluster are respectively less than or equal to a predetermined threshold, and the sum of demands of customer points within each of the clusters is less than or equal to the capacity of each vehicle; Construction step: In this construction step, the traveling salesman problem models within each of the clusters are respectively converted into Ising problem models, and corresponding adjacency matrices and external field vectors are respectively obtained for each of the Ising problem models, wherein the adjacency matrices are quantized according to the precision of the optical Ising machine; And Parallel solving step: In this parallel solving step, for each of the Ising problem models constructed in the construction step, parallel solving is respectively performed by using the optical Ising machine according to the corresponding adjacency matrices and external field vectors, and the path matching the optimal solution of the optical Ising machine is used as the optimal vehicle path within each of the clusters.

2. The optoelectronic hybrid solving method for the vehicle path planning problem according to claim 1, wherein the predetermined threshold is obtained according to the magnitude relationship of distances between customer points.

3. The optoelectronic hybrid solving method for the vehicle path planning problem according to claim 1, wherein the clustering step includes the following steps: (S1) Set the total number of all customer points as N, set the demand of the i-th customer point as d i , set the upper limit of the load capacity of each vehicle as Q, and determine that the minimum number of vehicles required is K; (S2) Select the coordinates of K customer points as the initial coordinate values of the centers of each cluster, and respectively set the capacity of each cluster to the upper limit Q of the load capacity of the vehicle; (S3) According to the demands of each customer point, allocate the customer points to the corresponding clusters; (S4) After the allocation of all customer points is completed, recalculate and update the center coordinate values of each of the K clusters; and (S5) Determine whether the difference between the center coordinates of the new cluster and the center coordinates of the original cluster is greater than a set threshold. If it is greater than the set threshold, transfer to step (S3); otherwise, the clustering step ends.

4. The optoelectronic hybrid solving method for the vehicle path planning problem according to claim 3, wherein The minimum required number of vehicles K = [Σ i N d i / Q] + 1, where [Σ i N d i / Q] is the largest integer not exceeding Σ i N d i / Q.

5. The optoelectronic hybrid solving method for the vehicle path planning problem according to claim 3, wherein the construction step includes the following steps: (S6) According to the distances between each customer point in each cluster, assign weights to the corresponding edges; (S7) For each customer point in each cluster, introduce a number of binary variables corresponding to the square of the number of customer points in each cluster, give the objective function of the constraint and the objective function of the total weight, and establish K quadratic unconstrained binary optimization problem models; and (S8) Convert the K quadratic unconstrained binary optimization problem models into K Ising problem models, and respectively obtain the corresponding adjacency matrices and external field vectors for the K Ising problem models.

6. The optoelectronic hybrid solving method for the vehicle path planning problem according to claim 5, wherein the parallel solving step includes the following steps: (S9) Set the K optical Ising machines according to the K adjacency matrices and external field vectors; (S10) Use the K set optical Ising machines to respectively perform parallel recursive solving on the K Ising problem models; (S11) Set the current optimal solution of each optical Ising machine as Best_x k , set the current optimal energy of each optical Ising machine as Best_H k , set the variable value of the t-th iteration of each optical Ising machine as x k (t), set the energy of the t-th iteration of each optical Ising machine as H k (t). For each optical Ising machine, calculate the energy H k (t) and compare it with the current optimal energy Best_H k . If H k (t) < Best_H k and the current solution result of the optical Ising machine satisfies the condition of the objective function H k A = 0, then let Best_H k = H k (t), Best_x k = x k (t). Otherwise, Best_H k remains unchanged, and the iteration number t is incremented by 1; (S12) For each optical Ising machine, determine whether the number of iterations t is less than the maximum number of iterations Nmax. If t is less than Nmax, return to step (S10); otherwise, output the optimal solution Best_x k and the optimal energy Best_H k .

7. An optoelectronic hybrid solution system for vehicle path planning problems, characterized in that, It includes: A clustering unit that groups customer points into multiple clusters according to the number of vehicles, such that the distances between customer points in each cluster are respectively less than or equal to a predetermined threshold, and the demands of all customer points in each cluster are less than or equal to the capacity of each vehicle; A construction unit that respectively converts the traveling salesman problem models within each cluster into Ising problem models, and respectively obtains corresponding adjacency matrices and external field vectors for each Ising problem model, where the adjacency matrices are quantized according to the precision of the optical Ising machine; and A parallel solving unit that respectively performs parallel solving on each Ising problem constructed in the construction unit according to the corresponding adjacency matrices and external field vectors and using the optical Ising machine, and takes the path matching the optimal solution of the optical Ising machine as the optimal vehicle path within each cluster.

8. The optoelectronic hybrid solving system for the vehicle routing problem according to claim 7, wherein the predetermined threshold is obtained according to the magnitude relationship of the distances between customer points.

9. The optoelectronic hybrid solving system for the vehicle routing problem according to claim 7, wherein the clustering unit performs the following steps: (S1) Set the total number of all customer points to N, set the demand of the i-th customer point to d i , set the upper limit of the load capacity of each vehicle to Q, and determine that the minimum number of vehicles required is K; (S2) Select the coordinates of K customer points as the initial coordinate values of the centers of each cluster, and respectively set the capacity of each cluster to the upper limit Q of the load capacity of the vehicle; (S3) According to the demands of each customer point, allocate the customer points to the corresponding clusters; (S4) After all customer points are allocated, recalculate and update the center coordinate values of each of the K clusters; and (S5) Determine whether the difference between the center coordinates of the new cluster and the center coordinates of the original cluster is greater than a set threshold. If it is greater than the set threshold, transfer to step (S3), otherwise the clustering step ends.

10. The optoelectronic hybrid solving system for the vehicle routing problem according to claim 9, wherein The minimum required number of vehicles K = [Σ i N d i / Q] + 1, where [Σ i N d i / Q] is the largest integer not exceeding Σ i N d i / Q.

11. The optoelectronic hybrid solving system for the vehicle routing problem according to claim 9, wherein the construction unit performs the following steps: (S6) According to the distances between customer points in each cluster, assign weights to the corresponding edges; (S7) For each customer point in each cluster, introduce a number of binary variables corresponding to the square of the number of customer points in each cluster, give the objective function of the constraint and the objective function of the total weight, and establish K quadratic unconstrained binary optimization problem models; and (S8) Convert the K quadratic unconstrained binary optimization problem models into K Ising problem models, and respectively obtain corresponding adjacency matrices and external field vectors for the K Ising problem models.

12. The optoelectronic hybrid solving system for the vehicle routing problem according to claim 11, wherein the parallel solving unit performs the following steps: (S9) Set the K optical Ising machines according to the K adjacency matrices and external field vectors; (S10) Use the K set optical Ising machines to respectively perform parallel recursive solving on the K Ising problem models. (S11) Set the current optimal solution of each optical Ising machine to Best_x k , set the current optimal energy of each optical Ising machine to Best_H k , set the variable value of the t-th iteration of each optical Ising machine to x k (t), set the energy of the t-th iteration of each optical Ising machine to H k (t). For each optical Ising machine, calculate the energy H k (t) and compare it with the current optimal energy Best_H k . If H k (t) < Best_H k and the current solution result of the optical Ising machine satisfies the condition of the objective function H k A = 0, then let Best_H k = H k (t), Best_x k = x k (t), otherwise Best_H k remains unchanged, and the iteration number t is incremented by 1; (S12) For each optical Ising machine, determine whether the iteration number t is less than the maximum iteration number Nmax. If t is less than Nmax, return to step (S10); otherwise, output the optimal solution Best_x k and the optimal energy Best_H k .

Images