Fresh goods cross-region collaborative transportation path optimization method of simulated annealing algorithm based on genetic-large neighborhood search

By applying the simulated annealing algorithm of genetic-big neighborhood search in fresh goods transportation, the cross-regional coordinated transportation path of fresh goods is optimized, and the problem of time coupling throughout the process is solved, and a rapid convergence and stable global optimal solution is achieved. Improve transportation efficiency and market competitiveness.

CN119026769APending Publication Date: 2024-11-26GUANGXI TEACHERS EDUCATION UNIV
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411031245.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The prior art is difficult to effectively solve the problem of full-time time coupling in fresh goods transportation, resulting in path optimization easily falling into local optimality, slow convergence speed, unstable optimal solution, and difficult to obtain good solution sets within the effective time.

Method used

Using a simulated annealing algorithm based on genetic-big neighborhood search, the cross-regional distribution path optimization model of fresh goods is constructed, combined with genetic algorithm and large neighborhood search mechanism, the transportation path of fresh goods is optimized, avoiding local optimization, and improving the algorithm convergence speed and stability.

Benefits of technology

It realizes the consideration of time coupling from an overall perspective in fresh goods transportation, avoids local optimization, improves the convergence speed and optimization stability of the algorithm, obtains global optimal solutions, and improves transportation efficiency and market competitiveness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026769A_ABST
    Figure CN119026769A_ABST
Patent Text Reader

Abstract

The invention discloses a fresh goods cross-region collaborative transportation path optimization method based on a genetic-large neighborhood search simulated annealing algorithm, and relates to the technical field of fresh goods transportation, and the method comprises the following steps: S100, according to a fresh goods cross-region transportation route map, determining the position of each site on the way, and listing the distance between the sites; s200, constructing a fresh goods cross-region distribution path optimization model; s300, solving the fresh goods cross-regional distribution path optimization model through a simulated annealing algorithm based on genetic-large neighborhood search to obtain an optimized path; firstly, node and vehicle scheduling allocation is performed through a genetic algorithm, and then iterative optimization is performed by adopting a simulated annealing algorithm of large neighborhood search according to a node allocation scheme to obtain an optimized path. According to the method provided by the invention, the defects of local optimum, low convergence speed, unstable optimal solution and the like can be overcome, the global optimal solution of cross-regional collaborative transportation is improved, and the convergence speed of the algorithm is accelerated.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of fresh goods transportation, and in particular to a method for optimizing the cross-regional collaborative transportation path of fresh goods based on a simulated annealing algorithm of a genetic-large neighborhood search. Background Art

[0002] In recent years, with the rapid development of my country's logistics infrastructure, the time and space distance between products and target nodes has been expanded to a certain extent, making the problem of delivering fresh products with good freshness between cities more challenging. However, due to the low concentration of the fresh supply chain industry, irregular business operations, complex collection work, high costs, and poor timeliness of terminal delivery, fresh transportation delays and other problems are prominent. How to maximize the freshness of products when delivered, improve the efficiency of cold chain transportation, and reduce the cost of fresh cold chain is of great significance to improving the market competitiveness of logistics companies. Inter-city LTL fresh goods need to go through the process of arriving at the starting hub from the starting node, sorting and concentrating at the hub, and then being transported by the starting hub to the terminal hub by the concentrated trunk line, and the goods are distributed at the terminal hub. Decay is inevitable in the transportation of fresh goods. This attribute often makes each transportation stage have a strong time coupling. It is necessary to reasonably select the transshipment hub for the integrated transportation of goods and the routing and scheduling of distribution vehicles to reduce the transportation cost of the entire network and improve the time efficiency of the target node (Reference 1). For example, Xie et al. (reference 2) took cold chain LTL logistics as the research object and constructed a two-stage optimization model with the lowest average dispersion of front-end cargo collection and the lowest total cost of the entire chain from cold chain LTL trunk line transportation to terminal distribution. Chen et al. (reference 3) considered the time window sensitivity of the target node in the green vehicle routing problem of multi-center cold chain logistics under the central + kitchen model, aiming to minimize the sum of carbon emission costs, distribution costs and time window penalty costs.

[0003] In this context, cross-docking terminals are able to integrate and sort fast-flowing products on the supply chain network. Commercial products are collected from various sources by inbound trucks, unloaded at cross-docks, and reunited with other products (References 4 and 5). ZACHARIADIS et al. (Reference 6) combined the capacity constraints of the transshipment center in the cross-docking vehicle path planning, and adopted a first-in-first-out strategy to coordinate the vehicles entering and leaving the station and the scheduling problem. While THEOPHILUS et al. (Reference 7) constructed a mixed integer model for the cold chain cross-docking terminal truck scheduling optimization problem to capture the decay of perishable products throughout the service process of reaching the truck. SHAHABI-SHAHMIRI et al. (Reference 4) converted the delivery deadline of perishable goods into a constraint to ensure that the perishable goods arrive within the delivery deadline.

[0004] Due to the perishable nature of fresh goods and the dual constraints of the time window of the target node, the overall fresh cross-docking transportation becomes highly sensitive to the transportation time of goods. Li et al. (reference 8) considered the dual time window constraints of the closing time of the transportation tool service and the order time window in the optimization problem of the LTL multimodal transport route of order merging and cargo transfer. Lv et al. (reference 9) considered the strong coupling of the multi-level collaborative distribution network system and the timeliness of the order to establish a multi-level (three-level) collaborative path optimization model with time window constraints under complete information sharing. In view of the randomness of the arrival of goods at the terminal hub in fresh food transportation and the uncertainty of goods, existing studies often adopt a random batch distribution method, that is, according to the time window requirements of the goods, the goods are prioritized to perform distribution routes and vehicle scheduling to reduce the violation of the time window (references 10 and 11).

[0005] In summary, existing research focuses on the end-to-end delivery path problem of fresh goods, and rarely considers the time coupling of the entire fresh goods transportation process from a holistic perspective. The spatiotemporal changes in the distribution of fresh goods make the solution of this problem more complicated. Due to the non-convex characteristics of this problem and the strong constraints on transportation time, when performing path optimization, it is often easy to fall into local optimality, slow convergence speed, and unstable optimal solution. It is often impossible to obtain a relatively good solution set within an effective time.

[0006] References

[0007] 1.Ladier AL,Alpan G.Cross-docking operations:Current research versus industry practice[J].Omega,2016,62:145–162.

[0008] 2. Xie Ruhe, He Jiawen, Zou Yifeng, et al. Optimization model of the whole chain from front-end collection to terminal distribution of cold chain LTL logistics[J]. Transportation Systems Engineering and Information, 2023, 23(3): 204–213.

[0009] 3. Chen Yudie, Gan Hongcheng, Cheng Liang. Joint distribution cold chain logistics model and its solution algorithm under the background of "dual carbon"[J]. Control and Decision, 2023, 38(7): 1951–1959.

[0010] 4.Shahabi-Shahmiri R,Asian S,Tavakkoli-Moghaddam R,et al.A routing and scheduling problem for cross-docking networks with perishable products,heterogeneous vehicles and split delivery[J].Computers&IndustrialEngineering,2021,157:107299.

[0011] 5. Yu VF, Anh PT, Baldacci RA robust optimization approach for the vehicle routing problem with cross-docking under demand uncertainty [J]. Transportation Research Part E: Logistics and Transportation Review, 2023,173:103106.

[0012] 6.Zachariadis EE,Nikolopoulou AI,Manousakis EG,et al.The vehiclerouting problem with capacitated cross-docking[J].Expert Systems withApplications,2022,196:116620.

[0013] 7. Theophilus O, Dulebenets MA, Pasha J, et al. Truck scheduling optimization at acold-chain cross-docking terminal with product perishability considerations[J]. Computers&Industrial Engineering, 2021,156:107240.

[0014] 8. Li Zhaojin, Liu Ya, Yang Zhen. Research on multimodal transport path optimization considering order consolidation and cargo transshipment[J]. Operations Research and Management, 2022, 31(4): 28–34.

[0015] 9. Lv Xiaohui, Wang Nengmin, Yang Zhen. Collaborative path optimization model and algorithm design for three-level distribution network under complete information sharing[J]. Operations Research and Management, 2023, 32(3): 78–84,130.

[0016] 10.Gallo A,Accorsi R,Akkerman R,et al.Scheduling cross-docking operations under uncertainty:A stochastic genetic algorithm based onscenarios tree[J].EURO Journal on Transportation and Logistics,2022,11:100095.

[0017] 11. Shi Jianli, Xie Lirong. Approximate dynamic programming for solving the vehicle routing problem of batch delivery with random demand[J]. Operations Research and Management, 2023, 32(5): 16–22. Summary of the invention

[0018] In view of the problems existing in the prior art, the present invention provides a method for optimizing the cross-regional collaborative transportation path of fresh goods based on a simulated annealing algorithm of a genetic-large neighborhood search. The method can overcome the problems of falling into local optimality, slow convergence speed, unstable optimal solution, etc., improves the global optimal solution for cross-regional collaborative transportation, and accelerates the convergence speed of the algorithm.

[0019] To achieve the above-mentioned object of the invention, the technical solution of the present invention is as follows:

[0020] The present invention provides a method for optimizing the cross-regional collaborative transportation path of fresh goods based on a simulated annealing algorithm of a genetic-large neighborhood search, comprising the following steps:

[0021] S100, based on the cross-regional transportation route map of fresh goods, determine the location of each station along the way and list the distances between the stations;

[0022] S200, building a cross-regional distribution route optimization model for fresh goods;

[0023] S300, solving the cross-regional distribution path optimization model of fresh goods by using a simulated annealing algorithm based on genetic-large neighborhood search to obtain an optimized path;

[0024] Firstly, the nodes and vehicles are dispatched and allocated by genetic algorithm. Then, the simulated annealing algorithm with large neighborhood search is used for iterative optimization according to the node allocation scheme to obtain the optimized path.

[0025] Furthermore, the construction of the cross-regional distribution route optimization model for fresh goods includes:

[0026] S201, set the objective function,

[0027]

[0028] In formula (1), M is the set of terminal hub r and starting hub s r, s∈M, Q is the set of starting node i and target node q i, q∈Q, K is the set of vehicle k k∈K, and N is the set of nodes j that the hub needs to be responsible for delivering j∈N;

[0029] represents the trunk line transportation cost of fresh goods, where: g sr It is expressed as the total volume of cargo transported between the starting hub s and the ending hub r, C sr represents the unit transportation cost between the starting hub s and the terminal hub r, d sr It is expressed as the transportation distance between the starting hub s and the end hub r, is the scale discount factor;

[0030] represents the operating cost of the starting hub s, where: Z ss It means that if the hub is at the starting hub s, then Z ss is 1, otherwise it is 0; f s represents the operating cost of the starting hub s;

[0031] represents the branch line transportation cost from the starting node i to the starting hub s, where: C1 represents the unit transportation cost on the branch line; d is It is expressed as the distance from the starting node i to the starting hub s, l is Determine whether the starting node i is served by the starting hub s. If so, it is 1, otherwise it is 0;

[0032] represents the branch line transportation cost from distribution node j to target node q, where: d jq represents the distance from the delivery node j to the target node q; It is used to determine whether there is a transportation arc from the delivery node j to the target node q for the vehicle k at the terminal hub r. If so, it is 1, otherwise it is 0; It is used to determine whether the distribution node j and the target node q are served by the same terminal hub r. If they are served by the same terminal hub r, it is 1, otherwise it is 0; Indicates whether the delivery node j and the target node q are delivered by the same vehicle k. If they are served by the same vehicle k, it is 1, otherwise it is 0;

[0033] Represents the cost of damage to fresh goods, where: iq represents the freight volume transported from the starting node i to the target node q; C2 represents the penalty cost for the decay of fresh goods. When the total transportation time is greater than the time when the fresh goods begin to decay, the decay cost will be incurred; δ is the freshness decay coefficient of fresh goods; T iq is the total time it takes to transport the goods from the starting node i to the target node q; It indicates the time when fresh products begin to decay;

[0034] represents the time cost of fresh goods, where a iq Indicates the earliest time when the target node can receive the goods, which is the left time window; b iq It represents the latest time for the target node to receive the goods, which is the right time window; α and β represent the penalty factors that exceed the left and right time window limits of the target node respectively;

[0035] S202, set constraints:

[0036]

[0037] Constraint (2) means that each starting node i is assigned to only one starting hub s; Z is Indicates that the starting node i is assigned to the starting hub s as 1; Z rq It means that the target node q is assigned to the terminal hub r, which is 1;

[0038] Constraint (3) ensures that only when the starting node i is assigned a starting hub s, there exists a route starting from the starting node i, then starting the transportation at the starting hub s, and ending at the destination node q. isrq It means that when the starting node i is assigned to the starting hub s as 1, and the target node q is assigned to the end hub r as 1, there is a route of isrq;

[0039] Constraint (4) ensures that l isrq The non-negativity of

[0040] Constraint (5) Calculate the freight volume that the originating hub s needs to handle;

[0041] Constraint (6) ensures that g s The non-negativity of g s represents the total freight volume handled by the originating hub s;

[0042]

[0043] Formula (7) is the constraint of the vehicle distribution path at the terminal hub r: All vehicles start from the terminal hub r and return to the original hub after completing all distribution tasks. It is used to determine whether there is a transportation arc from the target node q to the delivery node j for the vehicle k at the terminal hub r. If so, it is 1, otherwise it is 0;

[0044] Formula (8) is expressed as a weight constraint: the load of each vehicle cannot exceed its maximum load limit, O iq Represents the transportation volume of goods from the starting node i to the target node q; Determine whether the delivery node j and the target node q are served by the same terminal hub r. If yes, it is 1, otherwise it is 0; U represents the vehicle capacity;

[0045] Formula (9) is the customer point service constraint: each customer point is served once.

[0046] Furthermore, the specific steps of step S300 are as follows:

[0047] S310, input parameters: transportation distance, unit transportation cost, transshipment center operation cost, node location coordinates, target node time window, penalty function;

[0048] S320, initialize the parameters of the simulated annealing algorithm of genetic-large neighborhood search: the initial population Popsize of the genetic algorithm, the crossover probability Pc, the mutation probability Pm, the first iteration number k l , the first maximum number of iterations L;

[0049] S330, based on the trunk transportation cost of fresh goods, the operating cost of the starting hub s, and the branch transportation cost from the starting node i to the starting hub s, the vehicle scheduling and node allocation scheme at the terminal hub r is initialized through the genetic algorithm chromosome;

[0050] S340, optimizing the fresh goods transportation path through a large neighborhood search simulated annealing search algorithm to obtain candidate delivery paths, candidate terminal delivery costs, and costs that violate the target node delivery time window;

[0051] S350, calculate the total time T consumed by the fresh goods from the starting node i to the target node q based on the candidate delivery paths, the candidate terminal delivery costs and the cost of violating the target node delivery time window iq , calculate the damage cost of fresh goods

[0052]

[0053] S360, solving the objective function value of the current iteration, updating the lowest objective function value, and the candidate delivery path corresponding to the lowest objective function value;

[0054] S370, determining the first iteration number k lIs it greater than the first maximum number of iterations L? If so, the candidate delivery path obtained in step S360 is the optimized path and the algorithm ends. Otherwise, k l =k l +1, go to step S380;

[0055] S380, performing genetic crossover mutation on the candidate delivery path corresponding to the lowest objective function value, the steps are as follows:

[0056] S381, generate a random number Rand1∈[0,1], if Rand1>crossover probability Pc, insert a vehicle at a random position of the candidate delivery path chromosome, otherwise, do not perform crossover operation;

[0057] S382, generate a random number Rand2∈[0,1], if Rand2> mutation probability Pm, remove a vehicle at a random position of the candidate delivery path chromosome, otherwise, do not perform mutation operation;

[0058] S383, determine whether there is an empty or unloaded condition in the vehicle in the candidate delivery path chromosome. If so, return to step S381, otherwise, return to step S340.

[0059] Furthermore, in step S330, the vehicle scheduling and node allocation scheme at the terminal hub r of the genetic algorithm chromosome initialization is specifically:

[0060] S331, generate a distance matrix from the starting node i to the starting hub s according to the coordinates of the starting node i, with the diagonal elements being 0. If the starting node i is not connected to the starting hub s, it is a maximum value, and select the starting hub s closest to the starting node i according to the distance matrix, and allocate the starting node i to the corresponding s matrix array in the P cell;

[0061] S332, the P cell array randomly generates a cross-regional transportation path: starting node i-starting hub s-end hub r-target node q, forming a chromosome structure of cross-regional transportation and storing it in the cross-regional transportation path matrix. The first column is the starting node i, the second column is the starting hub s, the third column is the end hub r, and the fourth column is the target node q. The time and related transportation costs of fresh goods from the starting node i to the end hub r are calculated.

[0062]

[0063] S333, according to the cross-regional transportation path matrix, the cross-regional transportation paths with the same terminal hub r are stored in the mp cell array, and the mp cell array forms a vehicle path set at each terminal hub r: the mp cell array is sorted according to the time when the fresh goods arrive at the terminal hub r, and then the vehicle scheduling arrangement is generated according to the location information of the terminal hub r, the location information of the target node q, the demand for fresh goods at the target node q, and the vehicle capacity. The transportation information of each vehicle includes the location of the target node q and the demand for fresh goods that need to be transported, as well as the time window constraint of the target node q; the vehicle scheduling in each terminal hub r in the mp cell array Arrange to form a specific initial group of chromosomes according to the number of vehicles. The vehicle scheduling arrangement satisfies constraints (2-6). The obtained vehicle allocation feasible solution for each area is established. The initial population of chromosomes for cargo allocation of preset vehicles is established by natural number coding. That is, the length of each chromosome is the sum of the number of target nodes q to be delivered and the number of vehicles K used. In the chromosome, the natural number 0 represents the terminal hub r, K+1 0s arrange the delivery target node q to K vehicles, and the part between every two 0s represents the delivery situation of a vehicle. Constraints (2-6) need to be satisfied, that is, the freight volume is balanced, and the total amount of freight allocated to each vehicle shall not exceed the maximum cargo capacity of the vehicle.

[0064] Furthermore, the large neighborhood simulated annealing search algorithm is used to optimize the fresh goods transportation path, including the following steps:

[0065] S341, setting the initial second iteration number v=0, the second maximum iteration number to LK, the initial annealing temperature T0 and the cooling rate p;

[0066] S342, combining the time when the seafood goods arrive at the terminal hub r, the left time window and the right time window of the target node q, and the distance between the distribution node j and the target node q, determine the distribution order of the target node q responsible for distribution in each vehicle, and generate the initial path of the vehicle by adopting the mileage saving method. First, each target node q and distribution node j form a separate path, which starts from the terminal hub r, serves the target node q or the distribution node j, and then returns to the terminal hub r; according to the saving value Δd jq =d rj +d rq -d jq Merge the paths in descending order, where d rj is the distance from the terminal hub r to the distribution node j, d rq is the distance from the terminal hub r to the target node q, d jqis the distance from the delivery node j to the target node q. The merging conditions are: the merged path does not exceed the right time window of the target node q and the delivery node j during delivery; the merged path should reduce the violation of the left time window of the target node q and the delivery node j. The merged path forms a valid path, that is, a complete path, that is, a vehicle can complete the delivery tasks of multiple starting nodes i at the same time during a delivery process; the delivery node j is the target node corresponding to other starting nodes;

[0067] S343, when no more paths can be merged, stop merging, and return to the terminal hub r from the last target node q in the merged path, output the final path plan, and calculate and output the total travel distance;

[0068] S344, based on the time taken by the fresh goods from the starting node i to the terminal hub r and the time taken by the terminal hub r to the demand target node q calculated in S332, calculate the transportation cost generated when the terminal hub r is delivered and the cost of violating the user's delivery time window, that is, the sum of the terminal delivery cost of the current solution and the cost of violating the user's delivery time window

[0069]

[0070] S345, based on the current solution of the path arrangement of each vehicle in each area, a new possible solution is generated in the large neighborhood, and the driving paths of two vehicles are randomly selected and the three neighborhood operators, namely, the Split operator, the Elim operator and the Icross operator, are used to transform the paths; a vehicle path is randomly selected and the distribution order in the path is transformed by the neighborhood operators swap operator and move operator in the path to generate a new neighborhood solution;

[0071] S346, calculating the terminal delivery cost of the new neighborhood solution and the cost of violating the user delivery time window,

[0072]

[0073] And compare it with the sum of the terminal delivery cost of the current solution and the cost of violating the user delivery time window to judge the quality of the new neighborhood solution. If the sum of the terminal delivery cost of the new neighborhood solution and the cost of violating the user delivery time window is smaller than the sum of the terminal delivery cost of the current solution and the cost of violating the user delivery time window, the new neighborhood solution is directly accepted; otherwise, the probability of accepting the new neighborhood solution is determined by the current temperature and the difference between the objective function value of the new neighborhood solution and the objective function value of the current solution. The probability of accepting the new neighborhood solution is calculated by the Metropolis criterion, and the probability of accepting the new neighborhood solution gradually decreases with the iteration of the algorithm.

[0074] S347, reducing the current temperature according to the set cooling rate p;

[0075] S348, when the new neighborhood solution of the preset number of consecutive iterations remains unchanged, for example, the preset number of consecutive iterations can be set to 3, when the new neighborhood solution of 3 consecutive iterations remains unchanged, or when the second number of iterations v = the second maximum number of iterations LK, terminate the large neighborhood search simulated annealing search algorithm, otherwise, v = v+1, return to S346.

[0076] Compared with the prior art, the present invention has the following beneficial effects:

[0077] 1. The cross-regional collaborative transportation path optimization method of fresh goods based on the simulated annealing algorithm of genetic-large neighborhood search provided by the present invention constructs a cross-regional distribution path optimization model for fresh goods, considers the time coupling of the entire fresh goods transportation process from an overall perspective, and avoids falling into the local optimum during path optimization solution; a large neighborhood search (LNS) mechanism is added to the GA-SA algorithm, and the LNS search mechanism improves the search ability of the (GA-SA) mechanism in solving cross-regional transportation path problems. At the same time, the LNS search mechanism accelerates the convergence speed of the algorithm by reducing the space of the search path, and also improves the stability of the algorithm optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 The present invention is a flowchart of a method for optimizing the cross-regional collaborative transportation path of fresh goods based on a simulated annealing algorithm using a genetic-large neighborhood search in a technical solution of the present invention.

[0079] Figure 2 This is a schematic diagram of logistics transportation in a technical solution of the present invention.

[0080] Figure 3 A logistics network diagram in a technical solution of the present invention

[0081] Figure 4 This is a schematic diagram of the route planning of terminal hub delivery vehicles in one technical solution of the present invention.

[0082] Figure 5 Comparison chart of iterative convergence curves of different algorithms. DETAILED DESCRIPTION

[0083] The present invention is further described in detail below in conjunction with embodiments so that those skilled in the art can implement the invention with reference to the description.

[0084] In order to explain the technical content, achieved objectives and effects of the present invention in detail, the following is a description of the embodiments in conjunction with the accompanying drawings.

[0085] like Figure 1As shown, a method for optimizing the cross-regional collaborative transportation path of fresh goods based on a simulated annealing algorithm of a genetic-large neighborhood search comprises the following steps:

[0086] S100, based on the cross-regional transportation route map of fresh goods, determine the locations of each station along the way and list the distances between the stations;

[0087] like Figure 2 As shown, the stations along the way include the starting hub s, the end hub r, the starting node i, the target node q and the delivery node j.

[0088] S200, with the goal of minimizing total cost, builds a cross-regional distribution route optimization model for fresh goods;

[0089] Specifically, building a cross-regional distribution route optimization model for fresh goods includes:

[0090] S201, set the objective function to minimize the total cost, including the trunk transportation cost of fresh goods, the operation cost of the hub, the branch transportation cost from the starting node i to the starting hub s, and the damage cost and time cost of fresh goods

[0091]

[0092] In formula (1), M is the set of terminal hub r and starting hub s r, s∈M, Q is the set of starting node i and target node q i, q∈Q, K is the set of vehicle k k∈K, and N is the set of nodes j that the hub needs to be responsible for delivering j∈N;

[0093] represents the trunk line transportation cost of fresh goods, where: g sr It is expressed as the total volume of cargo transported between the starting hub s and the ending hub r, C sr represents the unit transportation cost between the starting hub s and the terminal hub r, d sr It is expressed as the transportation distance between the starting hub s and the end hub r, is the scale discount factor;

[0094] represents the operating cost of the starting hub s, where: Z ss It means that if the hub is at the starting hub s, then Z ss is 1, otherwise it is 0; f s represents the operating cost of the starting hub s;

[0095] represents the branch line transportation cost from the starting node i to the starting hub s, where: C1 represents the unit transportation cost on the branch line; d is It is expressed as the distance from the starting node i to the starting hub s, l isDetermine whether the starting node i is served by the starting hub s. If so, it is 1, otherwise it is 0;

[0096] represents the branch line transportation cost from distribution node j to target node q, where: d jq represents the distance from the delivery node j to the target node q; Used to determine whether there is a transportation arc from vehicle k to the target node q. If so, it is 1, otherwise it is 0; It is used to determine whether the distribution node j and the target node q are served by the same terminal hub r. If they are served by the same terminal hub r, it is 1, otherwise it is 0; Indicates whether the delivery node j and the target node q are delivered by the same vehicle k. If they are served by the same vehicle k, it is 1, otherwise it is 0;

[0097] Represents the cost of damage to fresh goods, where: iq represents the freight volume transported from the starting node i to the target node q; C2 represents the penalty cost for the decay of fresh goods. When the total transportation time is greater than the time when the fresh goods begin to decay, the decay cost will be incurred; δ is the freshness decay coefficient of fresh goods; T iq is the total time it takes to transport the goods from the starting node i to the target node q; It indicates the time when fresh products begin to decay;

[0098] represents the time cost of fresh goods, where: a iq Indicates the earliest time when the target node q can receive the goods, which is the left time window; b iq It represents the latest time for the target node to receive the goods, which is the right time window; α and β represent the penalty factors that exceed the left and right time window limits of the target node respectively;

[0099] S202, set constraints:

[0100]

[0101] Constraint (2) means that each starting node i is assigned to only one starting hub s; Z is Indicates that the starting node i is assigned to the starting hub s as 1; Z rq It means that the target node q is assigned to the terminal hub r, which is 1;

[0102] Constraint (3) ensures that only when the starting node i is assigned a starting hub s, there exists a route starting from the starting node ii, then starting the transportation at the starting hub s, and ending at the destination node q. isrqIt means that when the starting node i is assigned to the starting hub s as 1, and the target node q is assigned to the end hub r as 1, there is a route of isrq;

[0103] Constraint (4) ensures that l isrq The non-negativity of

[0104] Constraint (5) Calculate the freight volume that the originating hub s needs to handle;

[0105] Constraint (6) ensures that g s The non-negativity of g s represents the total freight volume handled by the originating hub s;

[0106]

[0107] Formula (7) is the constraint of the vehicle distribution path at the terminal hub r: All vehicles start from the terminal hub r and return to the original hub after completing all distribution tasks. It is used to determine whether there is a transportation arc from the target node q to the delivery node j for the vehicle k at the terminal hub r. If so, it is 1, otherwise it is 0;

[0108] Formula (8) is expressed as a weight constraint: the load of each vehicle cannot exceed its maximum load limit, O iq Represents the transportation volume of goods from the starting node i to the target node q; Determine whether the delivery node j and the target node q are served by the same terminal hub r. If yes, it is 1, otherwise it is 0; U represents the vehicle capacity;

[0109] Formula (9) is the customer point service constraint: each customer point is served once.

[0110] S300, solving the cross-regional distribution path optimization model of fresh goods by using a simulated annealing algorithm based on genetic-large neighborhood search to obtain an optimized path;

[0111] Firstly, the nodes and vehicles are dispatched and allocated by genetic algorithm. Then, the simulated annealing algorithm with large neighborhood search is used for iterative optimization according to the node allocation scheme to obtain the optimized path.

[0112] Specifically, the solution steps are as follows:

[0113] S310, input parameters: transportation distance, unit transportation cost, transshipment center operation cost, node location coordinates, target node time window, penalty function;

[0114] S320, initialize the parameters of the simulated annealing algorithm of genetic-large neighborhood search: the initial population Popsize of the genetic algorithm, the crossover probability Pc, the mutation probability Pm, the first iteration number k l, the first maximum number of iterations L;

[0115] S330, based on the trunk transportation cost of fresh goods, the operating cost of the starting hub s, and the branch transportation cost from the starting node i to the starting hub s, the vehicle scheduling and node allocation scheme at the terminal hub r is initialized through the genetic algorithm chromosome;

[0116] S340, optimizing the fresh goods transportation path through a large neighborhood search simulated annealing search algorithm to obtain candidate delivery paths, candidate terminal delivery costs, and costs that violate the target node delivery time window;

[0117] S350, calculate the total time T consumed by the fresh goods from the starting node i to the target node q based on the candidate delivery paths, the candidate terminal delivery costs and the cost of violating the target node delivery time window iq , calculate the damage cost of fresh goods

[0118]

[0119] S360, solving the objective function value of the current iteration, updating the lowest objective function value, and the candidate delivery path corresponding to the lowest objective function value;

[0120] S370, determining the first iteration number k l Is it greater than the first maximum number of iterations L? If so, the candidate delivery path obtained in step S360 is the optimized path and the algorithm ends. Otherwise, k l =k l +1, go to step S380;

[0121] S380, performing genetic crossover mutation on the candidate delivery path corresponding to the lowest objective function value, the steps are as follows:

[0122] S381, generate a random number Rand1∈[0,1], if Rand1>crossover probability Pc, insert a vehicle at a random position of the candidate delivery path chromosome, otherwise, do not perform crossover operation;

[0123] S382, generate a random number Rand2∈[0,1], if Rand2> mutation probability Pm, remove a vehicle at a random position of the candidate delivery path chromosome, otherwise, do not perform mutation operation;

[0124] S383, determine whether there is an empty or unloaded condition in the vehicle in the candidate delivery path chromosome. If so, return to step S381, otherwise, return to step S340.

[0125] Furthermore, in step S330, the vehicle scheduling and node allocation scheme at the terminal hub r of the genetic algorithm chromosome initialization is specifically:

[0126] S331, generate a distance matrix from the starting node i to the starting hub s according to the coordinates of the starting node i, with the diagonal elements being 0. If the starting node i is not connected to the starting hub s, it is a maximum value, and select the starting hub s closest to the starting node i according to the distance matrix, and allocate the starting node i to the corresponding s matrix array in the P cell;

[0127] S332, the P cell array randomly generates a cross-regional transportation path: starting node i-starting hub s-end hub r-target node q, forming a chromosome structure of cross-regional transportation and storing it in the cross-regional transportation path matrix. The first column is the starting node i, the second column is the starting hub s, the third column is the end hub r, and the fourth column is the target node q. The time and related transportation costs of fresh goods from the starting node i to the end hub r are calculated.

[0128]

[0129] S333, according to the cross-regional transportation path matrix, the cross-regional transportation paths with the same terminal hub r are stored in the mp cell array, and the mp cell array forms a vehicle path set at each terminal hub r: the mp cell array is sorted according to the time when the fresh goods arrive at the terminal hub r, and then the vehicle scheduling arrangement is generated according to the location information of the terminal hub r, the location information of the target node q, the demand for fresh goods at the target node q, and the vehicle capacity. The transportation information of each vehicle includes the location of the target node q and the demand for fresh goods that need to be transported, as well as the time window constraint of the target node q; the vehicle scheduling arrangement in each terminal hub r in the mp cell array forms a specific chromosome initial group according to the number of vehicles, and the vehicle The scheduling arrangement satisfies constraints (2-6), and the obtained vehicle allocation in each area is a feasible solution. The initial population of chromosomes for cargo allocation of preset vehicles is established by natural number coding, that is, the length of each chromosome is the sum of the number of target nodes q that need to be delivered and the number of vehicles K used. In the chromosome, the natural number 0 represents the terminal hub r, K+1 0s assign the delivery target node q to K vehicles, and the part between every two 0s represents the distribution situation of a vehicle, such as [0, 21, 23, 34, 0, 22, 12, 1, 4, 19, 0], 21, 23, 34 and other target node q numbers; it is necessary to satisfy constraint (2-6), that is, the freight volume is balanced, and the total freight allocated to each vehicle must not exceed the maximum cargo capacity of the vehicle.

[0130] Furthermore, the large neighborhood simulated annealing search algorithm is used to optimize the fresh goods transportation path, including the following steps:

[0131] S341, setting the initial second iteration number v=0, the second maximum iteration number to LK, the initial annealing temperature T0 and the cooling rate p;

[0132] S342, combining the time when the seafood goods arrive at the terminal hub r, the left time window and the right time window of the target node q, and the distance between the distribution node j and the target node q, determine the distribution order of the target node q responsible for distribution in each vehicle, and generate the initial path of the vehicle by adopting the mileage saving method. First, each target node q and distribution node j form a separate path, which starts from the terminal hub r, serves the target node q or the distribution node j, and then returns to the terminal hub r; according to the saving value Δd jq =d rj +d rq -d jq Merge the paths in descending order, where d rj is the distance from the terminal hub r to the distribution node j, d rq is the distance from the terminal hub r to the target node q, d jq is the distance from the delivery node J to the target node q. The merging conditions are: the merged path does not exceed the right time window of the target node q and the delivery node j during delivery; the merged path should reduce the violation of the left time window of the target node q and the delivery node j. The merged path forms a valid path, that is, a complete path, that is, a vehicle can complete the delivery tasks of multiple starting nodes i at the same time during a delivery process; the delivery node j is the target node corresponding to other starting nodes;

[0133] S343, when no more paths can be merged, stop merging, and return to the terminal hub r from the last target node q in the merged path, output the final path plan, and calculate and output the total travel distance;

[0134] S344, based on the time taken for the fresh goods to be delivered from the starting node i to the terminal hub r and the time taken for the fresh goods to be delivered from the terminal hub r to the target node q calculated in S332, calculate the transportation cost incurred when delivering from the terminal hub r and the cost of violating the user's delivery time window, that is, the sum of the terminal delivery cost of the current solution and the cost of violating the user's delivery time window

[0135]

[0136] S345, based on the current solution of the path arrangement of each vehicle in each area, a new possible solution is generated in the large neighborhood, and the driving paths of two vehicles are randomly selected to use the roulette driving selection neighborhood operator Split operator, Elim operator and Icross operator to perform path transformation; a vehicle path is randomly selected to transform the distribution order in the path through the neighborhood operator swap operator and move operator in the path to generate a new neighborhood solution, and the diversity of the population is improved by changing the distribution order of the current solution;

[0137] S346, calculating the terminal delivery cost of the new neighborhood solution and the cost of violating the user delivery time window,

[0138]

[0139] And compare it with the sum of the terminal delivery cost of the current solution and the cost of violating the user delivery time window to judge the quality of the new neighborhood solution. If the sum of the terminal delivery cost of the new neighborhood solution and the cost of violating the user delivery time window is smaller than the sum of the terminal delivery cost of the current solution and the cost of violating the user delivery time window, the new neighborhood solution is directly accepted; otherwise, the probability of accepting the new neighborhood solution is determined by the current temperature T and the difference between the objective function value of the new neighborhood solution and the objective function value of the current solution. The probability of accepting the new neighborhood solution is calculated by the Metropolis criterion, and the probability of accepting the new neighborhood solution gradually decreases with the iteration of the algorithm.

[0140] S347, reducing the current temperature T according to the set cooling rate p;

[0141] S348, when the new neighborhood solution of the preset number of consecutive iterations remains unchanged, for example, the preset number of consecutive iterations can be set to 3, when the new neighborhood solution of 3 consecutive iterations remains unchanged, or when the second number of iterations v = the second maximum number of iterations LK, terminate the large neighborhood search simulated annealing search algorithm, otherwise, v = v+1, return to S346.

[0142] Example

[0143] In order to verify the correctness and rationality of this method, Matlab language programming is used to Figure 3 The algorithm is simulated under the logistics network model, and the results are compared with genetic-simulated annealing (GA-SA), particle swarm simulated annealing algorithm (PSO-SA), and genetic-tabu search algorithm (GA-TS). The algorithm's initial population Popsize is set to 100, the maximum number of iterations is L=200, the crossover probability Pc=0.8, the mutation probability Pm=0.1, the initial annealing temperature T0=1000, the termination temperature t=0.001, the cooling rate q=0.95, and the second maximum number of iterations LK=300. The simulation results are shown in Figure 2. Figure 4 and Figure 5 The selected case is based on 177 target nodes in Guangxi public data as an example to verify the algorithm and model.

[0144] The route planning of the terminal hub distribution vehicle is obtained through the algorithm iteration. Figure 4 As shown in Table 1 and Table 2, a total of 55 vehicles were used, the total driving distance was 84281.7 kilometers, and the total violation time was 746 minutes. The total cost was 35367731.26 yuan.

[0145] Table 1 Number of vehicles at different terminal hubs and related driving path results

[0146]

[0147] Due to the conflict between the characteristics of fresh produce and the user's time window restrictions, more vehicles need to be arranged for direct delivery in the random batch delivery scenario to reduce user time violations. At the same time, due to the dynamic randomness of different batches of goods arriving at the terminal hub, the frequency of vehicle use and the total mileage of the vehicle will increase.

[0148] Figure 5 and Table 2 show the results and convergence curves of different algorithm iterations, respectively.

[0149] Table 2 Comparison of results of different algorithm iterations

[0150]

[0151] Depend on Figure 5 It can be seen that the optimal solution obtained by the improved algorithm (GA-LNS-SA) in the early stage of algorithm iteration is better than that of other algorithms, and the optimization ability is relatively strong and the convergence speed is faster. It converges to the optimal solution in about 75 generations. At the same time, the large neighborhood search (LNS) mechanism added to the algorithm improves the stability of the algorithm optimization. The present invention constructs a cross-regional distribution path optimization model for fresh goods, considers the time coupling of the entire fresh goods transportation process from an overall perspective, and avoids falling into local optimality during path optimization. The large neighborhood search (LNS) mechanism is added to the GA-SA algorithm. The LNS search mechanism improves the search ability of the (GA-SA) mechanism in solving cross-regional transportation path problems. At the same time, the LNS search mechanism accelerates the convergence speed of the algorithm by reducing the space of the search path. Therefore, the optimization ability in the early stage of the algorithm iteration is stronger than that of the (PSO-SA) and (GA-TS) mechanisms, and it shows good advantages in path search.

[0152] Although the present invention has been described in detail above with specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A method for optimizing the cross-regional collaborative transportation path of fresh goods based on a simulated annealing algorithm of genetic-large neighborhood search, characterized in that: The following steps are involved: S100, based on the cross-regional transportation route map of fresh goods, determine the location of each station along the way and list the distances between the stations; S200, building a cross-regional distribution route optimization model for fresh goods; S300, solving the cross-regional distribution path optimization model of fresh goods by using a simulated annealing algorithm based on genetic-large neighborhood search to obtain an optimized path; Firstly, the nodes and vehicles are dispatched and allocated by genetic algorithm. Then, the simulated annealing algorithm with large neighborhood search is used for iterative optimization according to the node allocation scheme to obtain the optimized path.

2. The method for optimizing the cross-regional collaborative transportation path of fresh goods based on the simulated annealing algorithm of genetic-large neighborhood search according to claim 1 is characterized in that: In step S200, constructing a cross-regional distribution route optimization model for fresh goods includes: S201, set the objective function, In formula (1), M is the set of terminal hub r and starting hub s r, s∈M, Q is the set of starting node i and target node q i, q∈Q, K is the set of vehicle k k∈K, and N is the set of nodes j that the hub needs to be responsible for delivering j∈N; represents the trunk line transportation cost of fresh goods, where: g sr It is expressed as the total volume of cargo transported between the starting hub s and the ending hub r, C sr represents the unit transportation cost between the starting hub s and the terminal hub r, d sr It is expressed as the transportation distance between the starting hub s and the end hub r, is the scale discount factor; represents the operating cost of the starting hub s, where: Z ss It means that if the hub is at the starting hub s, then Z ss is 1, otherwise it is 0; f s represents the operating cost of the starting hub s; represents the branch line transportation cost from the starting node i to the starting hub s, where: C1 represents the unit transportation cost on the branch line; d is It is expressed as the distance from the starting node i to the starting hub s, l is Determine whether the starting node i is served by the starting hub s. If so, it is 1, otherwise it is 0; represents the branch line transportation cost from distribution node j to target node q, where: d jq represents the distance from the delivery node j to the target node q; It is used to determine whether there is a transportation arc from the delivery node j to the target node q for the vehicle k at the terminal hub r. If so, it is 1, otherwise it is 0; It is used to determine whether the distribution node j and the target node q are served by the same terminal hub r. If they are served by the same terminal hub r, it is 1, otherwise it is 0; Indicates whether the delivery node j and the target node q are delivered by the same vehicle k. If they are served by the same vehicle k, it is 1, otherwise it is 0; Represents the cost of damage to fresh goods, where: iq represents the freight volume transported from the starting node i to the target node q; C2 represents the penalty cost for the decay of fresh goods. When the total transportation time is greater than the time when the fresh goods begin to decay, the decay cost will be incurred; δ is the freshness decay coefficient of fresh goods; T iq is the total time it takes to transport the goods from the starting node i to the target node q; It indicates the time when fresh products begin to decay; represents the time cost of fresh goods, where a iq Indicates the earliest time when the target node can receive the goods, which is the left time window; b iq It represents the latest time for the target node to receive the goods, which is the right time window; α and β represent the penalty factors that exceed the left and right time window limits of the target node respectively; S202, set constraints: Constraint (2) means that each starting node i is assigned to only one starting hub s; Z is Indicates that the starting node i is assigned to the starting hub s as 1; Z rq It means that the target node q is assigned to the terminal hub r, which is 1; Constraint (3) ensures that only when the starting node i is assigned a starting hub s, there exists a route starting from the starting node i, then starting the transportation at the starting hub s, and ending at the destination node q. isrq It means that when the starting node i is assigned to the starting hub s as 1, and the target node q is assigned to the end hub r as 1, there is a route of isrq; Constraint (4) ensures that l isrq The non-negativity of Constraint (5) Calculate the freight volume that the originating hub s needs to handle; Constraint (6) ensures that g s The non-negativity of g s represents the total freight volume handled by the originating hub s; Formula (7) is the constraint of the vehicle distribution path at the terminal hub r: All vehicles start from the terminal hub r and return to the original hub after completing all distribution tasks. It is used to determine whether there is a transportation arc from the target node q to the delivery node j for the vehicle k at the terminal hub r. If so, it is 1, otherwise it is 0; Formula (8) is expressed as a weight constraint: the load of each vehicle cannot exceed its maximum load limit, O iq Represents the transportation volume of goods from the starting node i to the target node q; Determine whether the delivery node j and the target node q are served by the same terminal hub r. If yes, it is 1, otherwise it is 0; U represents the vehicle capacity; Formula (9) is the customer point service constraint: each customer point is served once.

3. The method for optimizing the cross-regional collaborative transportation path of fresh goods based on the simulated annealing algorithm of genetic-large neighborhood search according to claim 2 is characterized in that: The specific steps of step S300 are as follows: S310, input parameters: transportation distance, unit transportation cost, transshipment center operation cost, node location coordinates, target node time window, penalty function; S320, initialize the parameters related to the simulated annealing algorithm of genetic-large neighborhood search: the initial population Popsize of the genetic algorithm, the crossover probability Pc, the mutation probability Pm, the first iteration number k l , the first maximum number of iterations L; S330, based on the trunk transportation cost of fresh goods, the operating cost of the starting hub s, and the branch transportation cost from the starting node i to the starting hub s, the vehicle scheduling and node allocation scheme at the terminal hub r is initialized through the genetic algorithm chromosome; S340, optimizing the fresh goods transportation path through a large neighborhood search simulated annealing search algorithm to obtain candidate delivery paths, candidate terminal delivery costs, and costs that violate the target node delivery time window; S350, calculate the total time T consumed by the fresh goods from the starting node i to the target node q based on the candidate delivery paths, the candidate terminal delivery costs and the cost of violating the target node delivery time window iq , calculate the damage cost of fresh goods S360, solving the objective function value of the current iteration, updating the lowest objective function value, and the candidate delivery path corresponding to the lowest objective function value; S370, determining the first iteration number k l Is it greater than the first maximum number of iterations L? If so, the candidate delivery path obtained in step S360 is the optimized path and the algorithm ends. Otherwise, k l =k l +1, go to step S380; S380, performing genetic crossover mutation on the candidate delivery path corresponding to the lowest objective function value, the steps are as follows: S381, generate a random number Rand1∈[0,1], if Rand1>crossover probability Pc, insert a vehicle at a random position of the candidate delivery path chromosome, otherwise, do not perform crossover operation; S382, generate a random number Rand2∈[0,1], if Rand2> mutation probability Pm, remove a vehicle at a random position of the candidate delivery path chromosome, otherwise, do not perform mutation operation; S383, determine whether there is an empty or unloaded condition in the vehicle in the candidate delivery path chromosome. If so, return to step S381, otherwise, return to step S340.

4. The method for optimizing the cross-regional collaborative transportation path of fresh goods based on the simulated annealing algorithm of genetic-large neighborhood search according to claim 3 is characterized in that: In step S330, the vehicle scheduling and node allocation scheme at the terminal hub r initialized by the genetic algorithm chromosome is specifically: S331, generate a distance matrix from the starting node i to the starting hub s according to the coordinates of the starting node i, with the diagonal elements being 0. If the starting node i is not connected to the starting hub s, it is a maximum value, and select the starting hub s closest to the starting node i according to the distance matrix, and allocate the starting node i to the corresponding s matrix array in the P cell; S332, the P cell array randomly generates a cross-regional transportation path: starting node i-starting hub s-end hub r-target node q, forming a chromosome structure of cross-regional transportation and storing it in the cross-regional transportation path matrix. The first column is the starting node i, the second column is the starting hub s, the third column is the end hub r, and the fourth column is the target node q. The time and related transportation costs of fresh goods from the starting node i to the end hub r are calculated. S333, according to the cross-regional transportation path matrix, the cross-regional transportation paths with the same terminal hub r are stored in the mp cell array, and the mp cell array forms a vehicle path set at each terminal hub r: the mp cell array is sorted according to the time when the fresh goods arrive at the terminal hub r, and then the vehicle scheduling arrangement is generated according to the location information of the terminal hub r, the location information of the target node q, the demand for fresh goods at the target node q, and the vehicle capacity. The transportation information of each vehicle includes the location of the target node q and the demand for fresh goods that need to be transported, as well as the time window constraint of the target node q; the vehicle scheduling in each terminal hub r in the mp cell array Arrange to form a specific initial group of chromosomes according to the number of vehicles. The vehicle scheduling arrangement satisfies constraints (2-6). The obtained vehicle allocation feasible solution for each area is established. The initial population of chromosomes for cargo allocation of preset vehicles is established by natural number coding. That is, the length of each chromosome is the sum of the number of target nodes q to be delivered and the number of vehicles K used. In the chromosome, the natural number 0 represents the terminal hub r, K+1 0s arrange the delivery target node q to K vehicles, and the part between every two 0s represents the delivery situation of a vehicle. Constraints (2-6) need to be satisfied, that is, the freight volume is balanced, and the total amount of freight allocated to each vehicle shall not exceed the maximum cargo capacity of the vehicle.

5. The method for optimizing the cross-regional collaborative transportation path of fresh goods based on the simulated annealing algorithm of genetic-large neighborhood search according to claim 4 is characterized in that: The large neighborhood simulated annealing search algorithm is used to optimize the transportation path of fresh goods, including the following steps: S341, setting the initial second iteration number v=0, the second maximum iteration number to LK, the initial annealing temperature T0 and the cooling rate p; S342, combining the time when the seafood goods arrive at the terminal hub r, the left time window and the right time window of the target node q, and the distance between the distribution node j and the target node q, determine the distribution order of the target node q responsible for distribution in each vehicle, and generate the initial path of the vehicle by adopting the mileage saving method. First, each target node q and distribution node j form a separate path, which starts from the terminal hub r, serves the target node q or the distribution node j, and then returns to the terminal hub r; according to the saving value Δd jq =d rj +d rq -d jq Merge the paths in descending order, where d rj is the distance from the terminal hub r to the distribution node j, d rq is the distance from the terminal hub r to the target node q, d jq is the distance from the delivery node j to the target node q. The merging condition is: the merged path does not exceed the right time window of the target node q and the delivery node j during delivery; the merged path should reduce the violation of the left time window of the target node q and the delivery node j. The merged path forms a valid path, that is, a complete path, that is, a vehicle can complete the delivery tasks of multiple starting nodes i at the same time during a delivery process, and the delivery node j is the target node corresponding to other starting nodes; S343, when no more paths can be merged, stop merging, and return to the terminal hub r from the last target node q in the merged path, output the final path plan, and calculate and output the total travel distance; S344, based on the time taken for the fresh goods to be delivered from the starting node i to the terminal hub r and the time taken for the fresh goods to be delivered from the terminal hub r to the target node q calculated in S332, calculate the transportation cost incurred when delivering from the terminal hub r and the cost of violating the user's delivery time window, that is, the sum of the terminal delivery cost of the current solution and the cost of violating the user's delivery time window S345, based on the current solution of the path arrangement of each vehicle in each area, a new possible solution is generated in the large neighborhood, and the driving paths of two vehicles are randomly selected and the three neighborhood operators, namely, the Split operator, the Elim operator and the Icross operator, are used to transform the paths; a vehicle path is randomly selected and the distribution order in the path is transformed by the neighborhood operators swap operator and move operator in the path to generate a new neighborhood solution; S346, calculating the terminal delivery cost of the new neighborhood solution and the cost of violating the user delivery time window, And compare it with the sum of the terminal delivery cost of the current solution and the cost of violating the user delivery time window to judge the quality of the new neighborhood solution. If the sum of the terminal delivery cost of the new neighborhood solution and the cost of violating the user delivery time window is smaller than the sum of the terminal delivery cost of the current solution and the cost of violating the user delivery time window, the new neighborhood solution is directly accepted; otherwise, the probability of accepting the new neighborhood solution is determined by the current temperature and the difference between the objective function value of the new neighborhood solution and the objective function value of the current solution. The probability of accepting the new neighborhood solution is calculated by the Metropolis criterion, and the probability of accepting the new neighborhood solution gradually decreases with the iteration of the algorithm. S347, reducing the current temperature according to the set cooling rate p; S348, s When the new neighborhood solution of the preset number of consecutive iterations remains unchanged, for example, the preset number of consecutive iterations can be set to 3. When the new neighborhood solution of 3 consecutive iterations remains unchanged, or when the second number of iterations v = the second maximum number of iterations LK, terminate the large neighborhood search simulated annealing search algorithm, otherwise, v = v+1, and return to S346.

Citation Information

Cited By

  • Supply chain scheduling optimization method based on transport outsourcing and time window constraint

    CN119740824A

  • Supply chain scheduling optimization method based on transportation outsourcing and time window constraint

    CN119740824B