E-commerce logistics distribution problem solving method and system and storage medium
Optimizing the e-commerce logistics distribution problem through adaptive neighborhood search algorithm, solving the problem of insufficient multi-objective optimization strategies in the existing technology, achieving efficient path optimization and real-time scheduling, significantly reducing costs and carbon emissions, and improving customers Satisfaction.
Patent Information
- Application Number
- CN202411569153.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-05
- Publication Date
- 2024-12-03
AI Technical Summary
The existing technology has failed to effectively solve the multi-objective optimization strategies and implementation details in the e-commerce logistics distribution problem, especially in the peak demand period, it is difficult to optimize vehicle paths and loading strategies, ignoring the real-time storage capacity limits and vehicle speeds of the front warehouse. Time dependence.
Adaptive variable neighborhood search algorithm (AVNS) is used to construct the initial solution, apply neighborhood operators to generate new solutions, and determine whether the new solution is accepted according to simulated annealing criteria, and construct it in combination with heuristic greedy algorithms and real-time orders and traffic conditions. Initial solution, optimize routes using exchange and repositioning operators, dynamically adjusting the penalty parameters to optimize the cost function.
It significantly reduces distribution costs, improves customer satisfaction, reduces operating costs and carbon emissions, and achieves efficient real-time scheduling and flexible path optimization, which is suitable for customers and transportation conditions of different sizes.
Smart Images

Figure SMS_1 
Figure SMS_2 
Figure SMS_3
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of smart logistics technology, and specifically relates to a method, system and storage medium for solving e-commerce logistics distribution problems. Background Art
[0002] The booming development of e-commerce has put forward higher requirements for logistics and distribution. This development is mainly due to the popularization of the Internet and the continuous innovation of e-commerce platforms. However, as the support of e-commerce, the logistics and distribution system is facing increasingly severe challenges. Especially during shopping festivals such as "Double Eleven" and Black Friday, the peak demand for logistics and distribution has brought tremendous pressure to the warehousing and transportation systems. During peak demand periods, the pressure on logistics and distribution doubles, and traditional logistics and distribution methods are often unable to cope with such sudden and huge demands. In order to meet these challenges, the logistics and distribution system needs to achieve efficient resource management and route optimization during peak demand periods.
[0003] For the e-commerce logistics distribution problem (ELDP), it involves two-level transportation operations. First, a group of large-capacity vehicles deliver goods from the central warehouse to various transfer stations; then, a group of small-capacity vehicles deliver goods from these transfer stations to end customers in urban areas. The goal of ELDP is to optimize the vehicle routing and loading strategies to minimize the total driving distance or time while ensuring that customer needs are met, and to meet vehicle capacity and synchronization constraints. For ELDP, most traditional models ignore the real-time storage capacity constraints of the forward warehouse, and do not fully consider the time dependence of vehicle speed and simultaneous pickup and delivery. For the forward warehouse, it can effectively relieve the pressure of the central warehouse, but its space is limited and requires efficient scheduling and path planning. The time dependence of vehicle speed means that urban traffic conditions vary significantly in different time periods, and the driving speed of logistics vehicles is time-dependent. The demand for simultaneous delivery and pickup means that during peak demand periods, the complexity of scheduling increases, and logistics vehicles must not only deliver but also pick up.
[0004] Adaptive Variable Neighborhood Search (AVNS) is a heuristic search algorithm that can be used to solve multi-objective optimization problems. It explores the solution space by switching between neighborhoods of different sizes to achieve global optimization. The core idea of the AVNS algorithm is to dynamically adjust the size of the search neighborhood to balance the breadth and depth of the search. During the execution of the algorithm, different neighborhood structures are adaptively selected based on the quality of the current solution and the search history, thereby avoiding falling into the local optimal solution and improving the search efficiency of the algorithm. The AVNS algorithm has been applied to the dynamic vehicle routing problem (DVRP) and other problems, showing good adaptability and flexibility.
[0005] However, when the AVNS algorithm is used to solve specific problems, the design of multi-objective optimization strategies and implementation details is an important issue. Only by reasonably designing multi-objective optimization strategies and implementation details can good optimization results and processing performance be achieved. However, there is currently a lack of relevant research on the application of the AVNS algorithm to the solution of the ELDP problem, and the existing technology fails to provide appropriate multi-objective optimization strategies and implementation details for the ELDP problem. Summary of the invention
[0006] In view of the problems of the prior art, the present invention provides a method, system and storage medium for solving e-commerce logistics distribution problems.
[0007] A method for solving e-commerce logistics distribution problems includes the following steps: Construct an initial solution to the e-commerce logistics and distribution problem; Starting from an initial solution, a neighborhood operator is applied to generate a new solution; Determine whether the new solution is accepted according to the simulated annealing criterion; Repeating the process of generating a new solution and determining whether the new solution is accepted; Wherein, the initial solution or new solution includes a list R ={ R a , R b , R c}, R a is the distribution route from the central warehouse to the transfer station. R b is the pickup and delivery route from the transfer station to the end customer. R c It is the pickup route from the transfer station to the central warehouse.
[0008] Preferably, the initial solution is constructed by a heuristic greedy algorithm; And / or, the initial solution is constructed based on real-time order conditions and real-time traffic conditions.
[0009] Preferably, the neighborhood operator is selected from a swap operator or a repositioning operator; Or, the neighborhood operator is selected from at least one of the following operators: 1) Using neighborhood operator pairs R a or R c Processing is performed to exchange two transfer stations in two different routes; 2) Use the repositioning operator R a or R c Processing is done to remove a transfer station and insert it into another route; 3) Using neighborhood operator pairs R b Processing is performed to exchange two end customers in two different routes; 4) Use the repositioning operator R b Processing is done to remove an end customer and insert it into another route.
[0010] Preferably, the probability of a neighborhood operator being selected is the weight multiplied by a preset initial probability value. The weights are updated using the following formula: in, ω j Indicates j The weight of the neighborhood operator, j represents the neighborhood operator, ω j0 represents the weight before updating, r’ is the reaction factor, ρ j is the neighborhood operator j Performance rating, n j is the neighborhood operator j The number of times to apply in generating new solutions.
[0011] Preferably, in the process of determining whether the new solution is accepted, the cost function f(sol) is defined as: f(sol)= t(sol) +α×loadViol(sol) +β×twViol(sol) +γ×capacityViol(sol) Among them, t(sol) represents the total travel time of the vehicle route, loadViol(sol) is R a , R b and R c The total number of vehicle load violations in the three stages, twViol(sol) is the total number of violations in the terminal customer time window, and capacityViol(sol) is the number of times the cargo volume at the transfer station exceeds its maximum capacity maxS , α, β and γ are penalty parameters.
[0012] Preferably, the calculation formulas of loadViol(sol), twViol(sol) and capacityViol(sol) are as follows: Among them, the symbol + The meaning is x + = max(0, x ), load r represents the vehicle load on route r, load seg represents the vehicle load of a section of route r, where r is R a , R b or R c , Q 1 is the maximum carrying capacity of the vehicle in the first stage, Q 2 is the maximum carrying capacity of the vehicle in the second stage, LT i For customers i The upper limit of the time window, at i For vehicles to reach customers i The actual time, V c For customers to gather, occupiedCapacity ( T g , s ) is the time interval T g The amount of cargo occupying the transfer station's capacity, s For transfer station, maxS is the maximum capacity of the transfer station, V s Gather for the transfer station, T For time collection.
[0013] Preferably, the penalty parameter is in the search domain [ l , k ] Dynamic adjustment within, the adjustment method is: When the new solution violates the constraint, the corresponding penalty parameter is multiplied by the penalty factor σ for update; When a new solution is accepted, the corresponding penalty parameter is divided by the penalty factor σ for update.
[0014] Preferably, the process of determining whether the new solution is accepted includes: determining whether the new solution is accepted as a global optimal solution and / or a local optimal solution.
[0015] The present invention also provides a system for implementing the above-mentioned method for solving the e-commerce logistics distribution problem, comprising: The initial solution building module is configured to build an initial solution to the e-commerce logistics distribution problem; A neighborhood search module, configured to, starting from an initial solution, apply a neighborhood operator to generate a new solution; A new solution determination module is configured to determine whether the new solution is accepted according to a simulated annealing criterion; Wherein, the initial solution or new solution includes a list R ={ R a , R b , R c}, R a is the distribution route from the central warehouse to the transfer station. R b is the pickup and delivery route from the transfer station to the end customer, which is R c The pickup route from the transfer station to the central warehouse.
[0016] The present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is executed by a processor, the device where the storage medium is located is controlled to execute the above-mentioned method for solving the e-commerce logistics distribution problem, or when the computer program is executed by a processor, the device where the storage medium is located is controlled to execute the above-mentioned system.
[0017] The technical solution of the present invention can achieve the following beneficial technical effects: 1. The AVNS algorithm designed by the present invention has excellent performance: The present invention applies the AVNS algorithm to solve the problem of e-commerce logistics distribution for the first time, and designs the strategy and implementation details of multi-objective optimization for this problem, so that it performs well in optimizing the distribution path. Specifically, according to the experimental results, the AVNS algorithm of the present invention can significantly reduce the cost function under different numbers of customers. For example, in the example of 25 customers, the AVNS algorithm of the present invention reduces the cost function by 12.54% to 22.87% compared with the greedy algorithm, which highlights the ability of the AVNS algorithm of the present invention to find more optimized solutions across different scenarios in a smaller scale. In the example of 50 customers, the AVNS algorithm of the present invention continues to show excellent performance, and its cost function is reduced by 10.26% to 14.75%. This shows that as the instance size increases, the AVNS algorithm of the present invention is robust in maintaining competitive solution quality, and its scalability and effectiveness in larger scenarios. In larger instances with 75 and 100 customers, the AVNS algorithm of the present invention is always better than the greedy algorithm, and the cost function is reduced by 11.37% to 14.87% and 9.57% to 13.29%, respectively. This strongly demonstrates that the present invention can effectively handle complex routing problems while consistently providing solutions of higher quality than the greedy approach. This adaptability enables the present invention to handle various complex scenarios in e-commerce logistics distribution, including different numbers of customers, time-dependent speeds, and real-time available transfer station capacity.
[0018] 2. Improve customer satisfaction: The algorithm of the present invention can optimize the delivery path of e-commerce logistics and improve the delivery efficiency, significantly shorten the delivery time, and thus improve customer satisfaction. Especially during the peak period of e-commerce, fast and accurate delivery service is a key factor in improving customer experience. The present invention can effectively cope with the large number of order demands during the peak period, ensure timely delivery, and greatly improve customer satisfaction and loyalty.
[0019] 3. Reduce operating costs: Optimizing the delivery route can not only shorten the delivery time, but also reduce the fuel consumption and labor costs of logistics vehicles. The present invention can significantly reduce the cost function, which means better route selection and lower transportation costs. This cost reduction has important economic significance for e-commerce logistics companies and can improve overall operational efficiency and profit margins.
[0020] 4. Green and low carbon: The present invention has significant environmental benefits by optimizing the delivery route, reducing vehicle mileage and fuel / power consumption, and reducing carbon dioxide emissions and other harmful gas emissions.
[0021] 5. Efficient real-time scheduling The present invention supports real-time dispatching and can dynamically adjust the delivery path according to real-time orders and traffic conditions. This feature makes the AVNS algorithm of the present invention more flexible and adaptable when dealing with emergencies and unexpected situations. For example, real-time available transfer station capacity information can help the algorithm dynamically adjust vehicle routes to ensure maximum delivery efficiency.
[0022] Obviously, according to the above contents of the present invention, in accordance with common technical knowledge and customary means in the art, without departing from the above basic technical ideas of the present invention, other various forms of modification, replacement or change may be made.
[0023] The above contents of the present invention are further described in detail below through specific implementation methods in the form of embodiments. However, this should not be understood as the scope of the above subject matter of the present invention being limited to the following examples. All technologies realized based on the above contents of the present invention belong to the scope of the present invention. DETAILED DESCRIPTION
[0024] It should be noted that the algorithms of data collection, transmission, storage and processing steps not specifically described in the embodiments, as well as the hardware structure, circuit connection, etc. not specifically described can all be implemented through the contents disclosed in the prior art.
[0025] Example 1 Method and system for solving e-commerce logistics distribution problems This embodiment applies the AVNS algorithm to solve the e-commerce logistics distribution problem, and specifically designs a multi-objective optimization strategy and implementation details for the problem, thereby providing a method and system for solving the e-commerce logistics distribution problem.
[0026] The system comprises: The initial solution building module is configured to build an initial solution to the e-commerce logistics distribution problem; A neighborhood search module, configured to, starting from an initial solution, apply a neighborhood operator to generate a new solution; The new solution determination module is configured to determine whether the new solution is accepted according to the simulated annealing criterion.
[0027] The method for solving the e-commerce logistics distribution problem using the above system includes the following steps: Construct an initial solution to the e-commerce logistics and distribution problem; Starting from an initial solution, a neighborhood operator is applied to generate a new solution; Determining whether the new solution is accepted according to the simulated annealing criterion; the process of determining whether the new solution is accepted includes: determining whether the new solution is accepted as a global optimal solution and / or a local optimal solution; Repeating the process of generating a new solution and determining whether the new solution is accepted; In this embodiment, the e-commerce logistics distribution problem is designed into three stages: (1) The first stage is the main distribution stage from the central warehouse to the transfer station; (2) The second stage is the pickup and delivery stage from the transfer station to the end customer; (3) The third stage is the pickup route from the transfer station to the central warehouse.
[0028] According to the above three stages, in this embodiment, the solution to the e-commerce logistics distribution problem includes the following list: R ={ R a , R b , R c}, R a is the distribution route from the central warehouse to the transfer station. R b is the pickup and delivery route from the transfer station to the end customer. R c It is the pickup route from the transfer station to the central warehouse.
[0029] The initial solution of this embodiment is constructed by a greedy heuristic algorithm. The greedy heuristic algorithm is an existing technology, and its basic idea is to select the current optimal solution each time and gradually construct the final solution.
[0030] In the process of generating a new solution, the neighborhood operator is selected from an exchange operator or a repositioning operator. The exchange operator exchanges customers on different routes, and the repositioning operator moves customers from one route to another. Each type of operator is applied to the three stages of path planning, resulting in six different neighborhood operators.
[0031] Specifically, the six neighborhood operators are as follows: (1) Using neighborhood operator pairs R a Processing is performed to exchange two transfer stations in two different routes; (2) Using neighborhood operator pairs R c Processing is performed to exchange two transfer stations in two different routes; (3) Using the repositioning operator R a Processing is done to remove a transfer station and insert it into another route; (4) Using the repositioning operator R c Processing is done to remove a transfer station and insert it into another route; (5) Using neighborhood operator pairs R b Processing is performed to exchange two end customers in two different routes; (6) Using the repositioning operator R b Processing is done to remove an end customer and insert it into another route.
[0032] This embodiment uses a relative weighted roulette wheel selection scheme to assign weights to each neighborhood operator. The scheme adjusts the weights based on the performance of the neighborhood operator in the previous iterations. Specifically, the probability of a particular neighborhood operator being selected is the weight multiplied by a preset initial probability value. The weight ω i Update using the following formula: in, i represents a specific neighborhood operator, ω i0 represents the weight before updating, r’ is the response factor (a parameter that controls how quickly the weights adapt to changes in the score), ρ i is the neighborhood operator i Performance rating, n i is the operator usage frequency (neighborhood operator i The number of times it is applied in the process of generating a new solution). ρ i The update is based on whether the neighborhood operator has found a new global optimal solution, improved the current solution without changing the global optimal solution, or accepted an unimproved solution. Specifically, ρ i Increased by addition. When a neighborhood operator finds a new solution, the system increases the score of the operator according to the quality of the solution found: 1) When a new global optimal solution is found, increase μ 1; 2) When the current solution is improved but the global optimal solution is not updated, increase μ 2; 3) When a non-improving solution is accepted, increase μ 3.
[0033] μ 1. μ 2 or μ 3 represents the increase in the score under different circumstances. The specific value is usually set based on experiments or parameter adjustments to achieve better algorithm performance.
[0034] In the process of determining whether the new solution is accepted, let globalSolution , localSolution and newSolution Represent the global optimal solution, local optimal solution and new solution respectively. The simulated annealing criterion determines the newSolution Whether it is accepted as the global best solution or the local best solution. Specifically, if newSolution The cost function is localSolution Small, then localSolution Updated to newSolution At the same time, if newSolution The cost is lower than globalSolution ,but globalSolution Also updated to newSolution Otherwise, the acceptance is calculated according to the simulated annealing criterion. newSolution As localSolution The probability is calculated by the following formula: e -(f(newSolution)−f(localSolution)) / Td Where T d is the temperature in the current iteration (T>0 d ), f ( newSolution ) is the cost value of the new solution, f ( localSolution ) is the cost value of the local optimal solution. d It is used to control the algorithm's acceptance probability of non-optimal solutions. As the temperature decreases, the difference f ( newSolution ) −f ( localSolution ) is larger, the probability of accepting a new inferior solution is smaller. The initial temperature is set to T start , in each iteration, T new =T prev ×c,T new is the temperature after the current iteration, T prev is the temperature of the previous iteration, and the temperature decreases by a factor c∈ [0,1], where c is called the cooling rate. f ( newSolution ) 、f ( localSolution ) represent the values of the new solution and the local solution in the current iteration after the cost function is calculated. When the cost of the new solution is less than the local solution, the local solution will be updated to the new solution. The temperature decreases as the number of iterations increases, which allows the algorithm to explore more solution spaces in the early stage (that is, there is a greater probability of accepting inferior solutions). As the iterations proceed, the temperature decreases, and the algorithm gradually converges to the local optimal solution. The probability of accepting inferior solutions becomes smaller and smaller until it converges completely.
[0035] In this embodiment, in the process of generating new solutions, a policy penalty function is used to guide the search towards the feasibility boundary, where high-quality solutions may exist. Therefore, all restrictions on vehicle capacity, time window and satellite capacity are relaxed during the search process, and a linear penalty function with three penalty coefficients is introduced to penalize violations of these constraints. These penalty coefficients are reflected in the calculation of the cost function. Specifically, the cost function f(sol) of solution s is defined as: f(sol)= t(sol) +α×loadViol(sol) +β×twViol(sol) +γ×capacityViol(sol) Among them, t(sol) represents the total travel time of the vehicle route, loadViol(sol) is R a , R b and R c The total number of vehicle load violations in the three stages, twViol(sol) is the total number of violations in the terminal customer time window, and capacityViol(sol) is the amount of cargo that exceeds its maximum capacity maxS , α, β and γ are penalty parameters. For capacityViol(sol), it refers to the g In a transfer station s The amount of cargo exceeds its maximum capacity maxS. When the cargo exceeds the maximum capacity, the violation value is occupiedCapacity(T g ,s)-maxS, and if it is not exceeded, the violation value is 0. In the formula, all violation items are calculated using x + = max(0, x ) to ensure that calculations are only done when the limit is exceeded.
[0036] The penalty parameter is in the search domain [ l , k ] The search domain is dynamically adjusted within the range that can be set according to the experience of the technicians. The search domain ensures that the weights start from reasonable values when the constraints are violated and prevents them from diverging to infinity. The penalty parameter dynamic adjustment method is: when the new solution violates the constraints (vehicle load constraints, time window constraints or transfer station capacity constraints), the corresponding penalty parameter is multiplied by the penalty factor σ (this parameter is an experimental parameter that can be defined and adjusted according to actual needs) for update; when the new solution is accepted, the corresponding penalty parameter is divided by the penalty factor σ for update.
[0037] The calculation formulas of loadViol(sol), twViol(sol) and capacityViol(sol) are as follows: Among them, the symbol + The meaning is x + = max(0, x ), load r represents the vehicle load on route r, load seg represents the vehicle load of a section of route r, where r is R a , R b or R c , Q 1 is the maximum carrying capacity of the vehicle in the first stage, Q 2 is the maximum carrying capacity of vehicles in the second stage (such as electric tricycles), LT i For customers i The upper limit of the time window, at i For vehicles to reach customers i The actual time, V c For customers to gather, occupiedCapacity ( T g , s ) is the time interval T g The amount of cargo occupying the transfer station's capacity, s For transfer station, maxS is the maximum capacity of the transfer station, V s Gather for the transfer station, T For time collection.
[0038] In addition, in this embodiment, the initial solution is constructed based on the real-time order situation and the real-time traffic situation, so that the neighborhood structure can be dynamically adjusted, the solution efficiency can be improved, and a better solution can be found in a shorter time. Considering the urban traffic conditions, using the real road network data, the distribution path is dynamically adjusted to deal with time-dependent factors and optimize the distribution efficiency. By introducing the time window and traffic flow data, the actual driving conditions are simulated to more accurately reflect the actual traffic conditions. The real demand data of my country's leading e-commerce companies are integrated to ensure that the algorithm can reflect the actual demand changes and improve the practicality of the optimization results. By analyzing the historical order data and real-time order information, accurate demand forecasts are made and the logistics resource allocation is adjusted in advance.
[0039] The beneficial effects of the present invention are described below through experimental examples.
[0040] Experimental Example 1 Performance comparison between the AVNS algorithm of the present invention and the greedy algorithm 1. Experimental Methods 1. Sample set The sample data used in this experiment comes from an e-commerce company in Chengdu. The company obtained the coordinates of all businesses and buildings in its area, as well as the statistics of packages collected and distributed by each building every day. In addition, this experiment also obtained vehicle speed data from the actual road network in the city, covering 1902 nodes and taking into account time changes.
[0041] 2. AVNS algorithm of the present invention Execute according to the method of Example 1.
[0042] 3. Greedy Algorithm The method is implemented according to the following literature: H. Zhou, H. Qin, Z. Zhang, and J. Li, “Two-echelon vehicle routing problem withtime windows and simultaneous pickupanddelivery,” Soft Comput, vol. 26, no. 7, pp. 3345–3360,Apr. 2022, doi:10.1007 / s00500-021-06712-2.
[0043] 2. Experimental Results In the 25-client instance, the AVNS algorithm of the present invention achieved a 12.54% to 22.87% reduction in the cost function over the greedy algorithm, highlighting the ability of the AVNS algorithm of the present invention to find more optimal solutions across different scenarios at a smaller scale. In the 50-client instance, the AVNS algorithm of the present invention continued to show excellent performance, with a 10.26% to 14.75% reduction in the cost function. This demonstrates the robustness of the AVNS algorithm of the present invention in maintaining competitive solution quality as the instance size increases, and its scalability and effectiveness in larger scenarios. In larger instances with 75 and 100 clients, the AVNS algorithm of the present invention consistently outperformed the greedy algorithm, with a cost function reduction of 11.37% to 14.87% and 9.57% to 13.29%, respectively.
[0044] It can be seen that the solution obtained by the AVNS algorithm provided by the present invention has higher quality, and the algorithm has good robustness and is suitable for processing data of various scales.
[0045] According to the above embodiments and experimental examples, it can be seen that the present invention realizes the application of the AVNS algorithm to solve the problem of e-commerce logistics and distribution for the first time, and designs a multi-objective optimization strategy and implementation details for the problem, thereby providing a method and system for solving the problem of e-commerce logistics and distribution. The present invention has the advantages of high problem solving quality, dynamic adjustment according to real-time information of orders and traffic, good robustness, good applicability to data of different scales, etc., and has a good application prospect.
Claims
1. A method for solving e-commerce logistics distribution problems, characterized in that: The steps include: Construct an initial solution to the e-commerce logistics and distribution problem; Starting from an initial solution, a neighborhood operator is applied to generate a new solution; Determine whether the new solution is accepted according to the simulated annealing criterion; Repeating the process of generating a new solution and determining whether the new solution is accepted; Wherein, the initial solution or new solution includes a list R ={ R a , R b , R c }, R a is the distribution route from the central warehouse to the transfer station. R b is the pickup and delivery route from the transfer station to the end customer. R c It is the pickup route from the transfer station to the central warehouse.
2. The method for solving the e-commerce logistics distribution problem according to claim 1, characterized in that: The initial solution is constructed by a heuristic greedy algorithm; And / or, the initial solution is constructed based on real-time order conditions and real-time traffic conditions.
3. The method for solving the e-commerce logistics distribution problem according to claim 1, characterized in that: The neighborhood operator is selected from a swap operator or a relocation operator; Or, the neighborhood operator is selected from at least one of the following operators: 1) Using neighborhood operator pairs R a or R c Processing is performed to exchange two transfer stations in two different routes; 2) Use the repositioning operator R a or R c Processing is done to remove a transfer station and insert it into another route; 3) Using neighborhood operator pairs R b Processing is performed to exchange two end customers in two different routes; 4) Use the repositioning operator R b Processing is done to remove an end customer and insert it into another route.
4. The method for solving e-commerce logistics distribution problems according to claim 1 or 3, characterized in that: The probability of a neighborhood operator being selected is the weight multiplied by the preset initial probability value. The weights are updated using the following formula: in, ω j Indicates j The weight of the neighborhood operator, j represents the neighborhood operator, ω j0 represents the weight before updating, r’ is the reaction factor, ρ j is the neighborhood operator j Performance rating, n j is the neighborhood operator j The number of times to apply in generating new solutions.
5. The method for solving the e-commerce logistics distribution problem according to claim 1, characterized in that: In the process of determining whether a new solution is accepted, the cost function f(sol) is defined as: f(sol)= t(sol) +α×loadViol(sol) +β×twViol(sol) +γ×capacityViol(sol) Among them, t(sol) represents the total travel time of the vehicle route, loadViol(sol) is R a , R b and R c The total number of vehicle load violations in the three stages, twViol(sol) is the total number of violations in the terminal customer time window, and capacityViol(sol) is the number of times the cargo volume at the transfer station exceeds its maximum capacity maxS , α, β and γ are penalty parameters.
6. The method for solving the e-commerce logistics distribution problem according to claim 5, characterized in that: The calculation formulas of loadViol(sol), twViol(sol) and capacityViol(sol) are as follows: Among them, the symbol + The meaning is x + = max(0, x ), load r represents the vehicle load on route r, load seg represents the vehicle load of a section of route r, where r is R a , R b or R c , Q 1 is the maximum carrying capacity of the vehicle in the first stage, Q 2 is the maximum carrying capacity of the vehicle in the second stage, LT i For customers i The upper limit of the time window, at i For vehicles to reach customers i The actual time, V c For customers to gather, occupiedCapacity ( T g , s ) is the time interval T g The amount of cargo occupying the transfer station's capacity, s For transfer station, maxS is the maximum capacity of the transfer station, V s Gather for the transfer station, T For time collection.
7. The method for solving the e-commerce logistics distribution problem according to claim 5, characterized in that: The penalty parameter is in the search domain [ l , k ] Dynamic adjustment within, the adjustment method is: When the new solution violates the constraint, the corresponding penalty parameter is multiplied by the penalty factor σ for update; When a new solution is accepted, the corresponding penalty parameter is divided by the penalty factor σ for update.
8. The method for solving the e-commerce logistics distribution problem according to claim 1, characterized in that: The process of determining whether the new solution is accepted includes: determining whether the new solution is accepted as a global optimal solution and / or a local optimal solution.
9. A system for implementing the method for solving the e-commerce logistics distribution problem according to any one of claims 1 to 8, characterized in that: include: The initial solution building module is configured to build an initial solution to the e-commerce logistics distribution problem; A neighborhood search module, configured to, starting from an initial solution, apply a neighborhood operator to generate a new solution; A new solution determination module is configured to determine whether the new solution is accepted according to a simulated annealing criterion; Wherein, the initial solution or new solution includes a list R ={ R a , R b , R c }, R a is the distribution route from the central warehouse to the transfer station. R b is the pickup and delivery route from the transfer station to the end customer, which is R c The pickup route from the transfer station to the central warehouse.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed by the processor, the device where the storage medium is located is controlled to execute the method for solving the e-commerce logistics distribution problem as described in any one of claims 1 to 8, or, when the computer program is executed by the processor, the device where the storage medium is located is controlled to execute the system as described in claim 9.
Citation Information
Patent Citations
A joint scheduling method for site selection and logistics distribution of an electric vehicle battery swap station
CN109583650A
A heterogeneous hazardous chemical substance transportation path planning method with a time window
CN109948855A
Urban two-stage distribution and scheduling method with mobile distribution station
CN110909952A
Logistics distribution path planning method based on variable neighborhood parallel annealing algorithm
CN111882099A
Dynamic vehicle path optimization method based on two-stage heuristic algorithm
CN112766614A
Cited By
Medical waste transportation scheme solving method and system based on double-layer optimization model
CN119692586A
