A multi-depot open vehicle routing planning method based on ALNS algorithm

Abstract

The application discloses a multi-depot open vehicle route planning method based on an ALNS algorithm, and specifically comprises the following steps: S1, a vehicle path mathematical model is constructed with the optimization target of minimizing the total path length; S2, detailed addresses of each relay station where vehicles are located after completing end distribution are input, and the locations of large customers within the service range of the network points and relevant ticket and piece data are input; S3, a distance matrix is generated according to the address information; S4, in the iteration process of the ALNS algorithm, the model solution is initialized based on the Multi-Deport scene, and the iteration calculation is performed on each route; and S5, in the iteration process of the ALNS algorithm, the route adopts a Single-trip strategy to match the vehicle, and the matching strategy of the route and the vehicle is implemented. Through the initial solution calculation and the processing method of the destruction and repair operators of the multi-depot open vehicle route planning problem, the application links two operation tasks of collection and end delivery, and improves the efficiency of the delivery and collection of the network points.

Description

Technical Field

[0001] This invention belongs to the field of logistics and transportation technology, specifically involving a multi-vehicle yard open vehicle route planning method based on the ALNS algorithm. Background Technology

[0002] With the rapid development of the logistics industry, it faces increasing challenges in service quality, pickup efficiency, and cost control. Currently, the basic workflow for pickup and delivery at logistics outlets is as follows: in the early morning and afternoon, priority is given to handling first and second-tier parcels arriving at the outlet that day; after completing these deliveries in the afternoon, large customer pickups are completed. Cost control in the logistics industry is becoming increasingly precise and refined, and some outlets that have voluntarily changed their operating models have become textbook examples of cost control in the industry. These outlets have shifted from relying solely on couriers to a lightweight operating model of drivers plus couriers. Drivers are responsible for delivering parcels to the station, and after completing their delivery tasks, they go to the originating large customer locations to pick up parcels. Once the vehicle is full within a specified time, they head to the outlet to load the parcels onto a larger truck and transport it to the transfer center, or drive directly to the transfer center to unload. Station couriers complete door-to-door delivery within the station's designated area, while outlet couriers handle the pickup of individual parcels.

[0003] Currently, the Adaptive Large Neighborhood Search (ALNS) algorithm is widely used in vehicle route planning. However, different application scenarios have significantly different requirements for the model scheme. Therefore, in practical applications, the ALNS algorithm needs to be customized and optimized according to specific scenarios to meet different needs. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a multi-depot open vehicle route planning method based on the ALNS algorithm. The method of this invention enables the two tasks of pickup and last-mile delivery to be linked by calculating the initial solution of the multi-depot open vehicle route planning problem and the handling of destruction and repair operators, thereby improving the efficiency of network point delivery and pickup. This breaks through the limitation of the VRP problem that it is fixed at a single starting point and must return to the starting point.

[0005] To achieve the above-mentioned objectives, the technical solution provided by this invention patent is as follows:

[0006] A multi-parking lot open vehicle route planning method based on the ALNS algorithm, the method specifically includes the following steps:

[0007] S1. Construct a mathematical model for vehicle routing with the optimization objective of minimizing the total path length. This model includes several constraints, specifically: graph flow balance, limits on the number of large customer shipments along the route, vehicle carrying capacity limits, and vehicle pickup time window requirements.

[0008] S2: The route planning personnel at the network point input the detailed addresses of each station where the vehicle is located after the last-mile delivery is completed, as well as the location of major customers within the network point's service area and related ticket volume data, to provide necessary information support for subsequent route planning.

[0009] S3, based on the address information of the locations of outlets or transit centers and major customers, uses Open Street Map service to generate a distance matrix between the locations of outlets or transit centers and major customers, in order to quantify the spatial relationship between different locations;

[0010] S4. During the iteration of the ALNS algorithm, the model solution is initialized based on the Multi-Deport scenario, and specific labels and identifications are made for the parking lot and the destination in the destruction and repair operators. Iterative calculations are performed on each route to ensure that the total pickup distance is short.

[0011] S5. During the iteration of the ALNS algorithm, the route adopts the Single-trip strategy to match vehicles, that is, each vehicle only performs a single delivery task, and the route and vehicle matching strategy is implemented.

[0012] Furthermore, the mathematical model for vehicle routing is a Single-trip, Multi-capacity and Multi-Deport Open Vehicle Routing Problem with Time Window mathematical model, which is set to minimize the total travel distance of all vehicles in order to optimize overall delivery efficiency.

[0013] Furthermore, the graph flow balancing specifically involves ensuring that the last point of last-mile delivery serves as the starting point of the pickup route, and the network point or transfer center serves as the ending point. When the number of delivery vehicles is insufficient to complete the pickup task, the network point will supplement the required pickup vehicles. When the last last-mile delivery location serves as the starting point of the route, its in-degree is zero and its out-degree is 1. When the network point or transfer center serves as the ending point of the route, its out-degree is zero and its in-degree is the sum of the total number of routes. If the total capacity of vehicles completing last-mile delivery is insufficient to complete the pickup task for all major customers, the network point will dispatch additional vehicles, which will depart from the network point, pass through major customers, and cooperate with the vehicles completing last-mile delivery to complete the pickup task.

[0014] Furthermore, the route planners at the service points input the detailed address of the vehicle after the last-mile delivery is completed, as well as the location of major customers within the service area of ​​the service point and related ticket volume data, to obtain the address information of the service points or transfer centers and major customers.

[0015] Furthermore, the distance matrix acquisition specifically involves: obtaining the shortest driving distance between the network point or transfer center and each major customer in the actual road network through the Open Street Map open-source map service, and organizing the calculated shortest driving distance between the network point or transfer center and each major customer into the form of a distance matrix, which intuitively shows the spatial relationship between different locations.

[0016] Furthermore, during the iterative calculation process, the ALNS employs a specific initial solution algorithm and destruction / repair operators; the initial solution algorithm specifically includes:

[0017] Place each starting point in the starting point list and each major customer point in the unvisited list. Iterate through each starting point to initialize the route. Each time, select the unvisited point closest to the last point of the route. If the point meets constraints such as vehicle capacity and pickup time, insert the unvisited point into the route and delete it from the unvisited list. Otherwise, select the next major customer point that is relatively close to the last transit point of the route. If no point meets the constraints, insert a network point or transfer center into the route as the endpoint and proceed with the next route planning until all starting points are routes or the unvisited point list is not empty. If all starting points have been visited and there are still unvisited points, set the starting point as a network point and repeat the above steps until the unvisited point list is empty. If there are still unvisited starting points but the unvisited point list is empty, set the starting point of the route as the unvisited starting point and the endpoint as a network point or transfer center.

[0018] The destruction and repair operators are as follows: In the destruction operator, the start and end points of a route cannot be deleted; in the repair operator, if a route has a start point that is a network point, an end point that is a network point or a transfer center, and no intermediate points, this route is a network point supplementary route and has no actual physical meaning, so the route is deleted; if a route has a start point that is one of the set multiple start points, no intermediate points, and an end point that is a network point or a transfer center, then the vehicle must return empty, this route has actual physical meaning and cannot be deleted.

[0019] Furthermore, the route and vehicle matching strategy is as follows: when the number of routes is greater than the number of vehicles, it is confirmed whether the route capacity meets the requirements of the vehicle carrying capacity; when the number of routes is reduced to less than or equal to the number of vehicles, it is determined whether each route can find a vehicle with a matching capacity; if a route can be matched with a vehicle with a suitable carrying capacity, the iterative calculation continues; if a match cannot be found, the large customer point is set up as a separate route, with the starting point as the network point and the ending point as the network point or transfer center, and the iterative calculation is performed again.

[0020] Based on the above technical solutions, the multi-parking lot open vehicle route planning method based on the ALNS algorithm of this invention has achieved the following technical advantages through practical application:

[0021] 1. The present invention provides a multi-depot open vehicle route planning method based on the ALNS algorithm. By calculating the initial solution of the multi-depot open vehicle route planning problem and handling the destruction and repair operators, the two tasks of pickup and last-mile delivery are linked, thereby improving the efficiency of delivery and pickup at the network point. This breaks through the limitation of the VRP problem that it is fixed at a single starting point and must return to the starting point.

[0022] 2. The present invention provides a multi-depot open vehicle route planning method based on the ALNS algorithm, which offers a selection method for vehicles with different carrying capacities. This allows for the selection of multiple vehicle types for pickup based on demand in practical applications. This breaks through the limitation of using only single-capacity vehicles in the traditional ALNS method, resulting in a significant improvement in vehicle utilization. In practical applications, existing network operation vehicles can be effectively utilized and assigned to suitable pickup routes, thereby improving the overall utilization efficiency of vehicles. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention is described below with reference to specific examples shown in the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0024] Example 1

[0025] This invention relates to a multi-parking lot open vehicle route planning method based on the ALNS algorithm. The method specifically includes the following steps:

[0026] S1. Construct a mathematical model for vehicle routing with the optimization objective of minimizing the total path length. This model includes several constraints, specifically: graph flow balance, limits on the number of large customer shipments along the route, vehicle carrying capacity limits, and vehicle pickup time window requirements.

[0027] S2: The route planning personnel at the network point input the detailed addresses of each station where the vehicle is located after the last-mile delivery is completed, as well as the location of major customers within the network point's service area and related ticket volume data, to provide necessary information support for subsequent route planning.

[0028] S3 uses Open Street Map service to generate a distance matrix between the locations of outlets or transit centers and major customers, based on the address information of the outlets or transit centers and major customers, in order to quantify the spatial relationship between different locations; Open Street Map is an open source map.

[0029] S4. During the iteration of the ALNS algorithm, the model solution is initialized based on the Multi-Deport scenario, and specific labels and identifications are made for the parking lot and the destination in the destruction and repair operators. Iterative calculations are performed on each route to ensure that the total pickup distance is short.

[0030] S5. During the iteration of the ALNS algorithm, the route adopts the Single-trip strategy to match vehicles, that is, each vehicle only performs a single delivery task, and the route and vehicle matching strategy is implemented.

[0031] The mathematical model for vehicle routing is a Single-trip, Multi-capacity and Multi-Deport Open Vehicle Routing Problem with Time Window. The mathematical model is set to minimize the total travel distance of all vehicles in order to optimize overall delivery efficiency.

[0032] The Single-trip, Multi-capacity and Multi-Deport Open Vehicle Routing Problem with Time Window is a model for an open vehicle routing problem with multiple capacities and multiple depots, using a time window.

[0033] The graph flow balancing is specifically as follows: ensure that the last point of last-mile delivery is the starting point of the pickup route, and the outlet or transfer center is the ending point. When the number of delivery vehicles is insufficient to complete the pickup task, the outlet will supplement the required pickup vehicles. When the last last-mile delivery point is the starting point of the route, its in-degree is zero and its out-degree is 1. When the outlet or transfer center is the ending point of the route, its out-degree is zero and its in-degree is the sum of the total number of routes. If the total capacity of vehicles completing last-mile delivery is insufficient to complete the pickup task for all major customers, the outlet will dispatch additional vehicles, which will depart from the outlet and pass through major customers to cooperate with the vehicles completing last-mile delivery to complete the pickup task.

[0034] Furthermore, all major customer stops along the route must have an equal in-degree and out-degree of 1 to ensure the continuity and integrity of the delivery path. Constraints include ensuring that the overall route, after individually numbering each start and end point, forms a directed acyclic graph to prevent loop paths and logical errors when a vehicle is added to a stop and the endpoint is also a stop. Considering the actual carrying capacity of the vehicles, constraints also include that the total number of packages collected on each route must be less than the vehicle's carrying capacity to ensure safety and efficiency during the collection process. To meet the timeliness requirements for collection, constraints also include a time window limit for major customer collection, meaning that vehicles must complete the collection of major customer packages within a specified time window.

[0035] The route planners input the detailed address of the vehicle after the last-mile delivery is completed, as well as the location of major customers within the service area of ​​the branch and related ticket volume data, to obtain the address information of the branch or transfer center and major customers.

[0036] The distance matrix is ​​obtained by using the Open Street Map open-source map service to obtain the shortest driving distance between the network points or transfer centers and each major customer in the actual road network. The calculated shortest driving distances between the network points or transfer centers and each major customer are then organized into a distance matrix, which intuitively shows the spatial relationship between different locations.

[0037] During the iterative calculation process, the ALNS performs a specific initial solution algorithm and destruction / repair operators; the initial solution algorithm is as follows:

[0038] Place each starting point in the starting point list and each major customer point in the unvisited list. Iterate through each starting point to initialize the route. Each time, select the unvisited point closest to the last point of the route. If the point meets constraints such as vehicle capacity and pickup time, insert the unvisited point into the route and delete it from the unvisited list. Otherwise, select the next major customer point that is relatively close to the last transit point of the route. If no point meets the constraints, insert a network point or transfer center into the route as the endpoint and proceed with the next route planning until all starting points are routes or the unvisited point list is not empty. If all starting points have been visited and there are still unvisited points, set the starting point as a network point and repeat the above steps until the unvisited point list is empty. If there are still unvisited starting points but the unvisited point list is empty, set the starting point of the route as the unvisited starting point and the endpoint as a network point or transfer center.

[0039] The destruction and repair operators are as follows: In the destruction operator, the start and end points of a route cannot be deleted; in the repair operator, if a route has a start point that is a network point, an end point that is a network point or a transfer center, and no intermediate points, this route is a network point supplementary route and has no actual physical meaning, so the route is deleted; if a route has a start point that is one of the set multiple start points, no intermediate points, and an end point that is a network point or a transfer center, then the vehicle must return empty, this route has actual physical meaning and cannot be deleted.

[0040] The route-vehicle matching strategy is as follows: When the number of routes exceeds the number of vehicles, it is confirmed whether the route capacity meets the vehicle carrying capacity requirements. When the number of routes decreases to less than or equal to the number of vehicles, it is determined whether each route can find a vehicle with matching capacity. If a route can be matched with a vehicle of suitable carrying capacity, the iterative calculation continues. If no match can be found, the large customer point is designated as a separate route, with the starting point as the branch point and the ending point as the branch point or transfer center, and the iterative calculation is performed again. The ticket volume of all routes is sorted in ascending order. Next, for each route, the number of available vehicles within the vehicle carrying capacity range corresponding to its ticket volume is calculated. Only when the number of these available vehicles is greater than or equal to the index of the current route (numbered starting from 1) can it be determined that the route can be matched with a vehicle of suitable carrying capacity.

[0041] This invention combines route capacity with vehicle matching strategies to ensure optimal resource utilization. In terms of data acquisition, it utilizes the open-source Open Street Map to construct a distance matrix from the pickup location to the distribution point / transfer center, providing data support for subsequent route planning. During the planning process, this invention rationally plans the pickup sequence for large customers based on the different capacities of vehicles at each distribution point. In this way, drivers can complete last-mile deliveries according to the planned route and then directly proceed to the location of large customers within the distribution point's area, efficiently delivering their packages to the distribution point or transfer center. Compared to the traditional method of relying on couriers for pickup, this invention significantly improves delivery efficiency. By constructing a mathematical model that meets actual business needs and combining it with Open Street Map information, this invention achieves rapid, bulk pickup of large customer packages and significantly improves pickup efficiency, providing the logistics industry with an efficient and intelligent multi-vehicle-yard pickup solution after completing last-mile delivery.

[0042] Example 2

[0043] A multi-parking lot open vehicle route planning method based on the ALNS algorithm, the method specifically includes the following steps:

[0044] The first step is to construct a mathematical model for the Single-trip, Multi-capacity, and Multi-Deport Open Vehicle Routing Problem with a Time Window, with the optimization objective of minimizing the total path length. This model includes a series of constraints, such as graph flow balance, limits on the number of large customer shipments along the route, vehicle carrying capacity limits, and vehicle pickup time window requirements, to ensure the rationality and feasibility of the planning. The specific model, variables, and constraints are explained below:

[0045]

[0046]

[0047] ∑ i ∑ j t ij x ijk ≤T, (16)

[0048] ∑ i ∑ j l ij x ijk ≤D, (17)

[0049]

[0050] Table 1 shows the meaning of each parameter in the above formula.

[0051]

[0052]

[0053] For convenience, this model assigns the network point number 0 and the destination number N+1. The location after the last delivery is completed is recorded from 1 to 1+|K|. Furthermore, the route of each vehicle is a directed acyclic graph.

[0054] The objective function is to minimize the total cost of vehicle travel. Constraint (1) ensures that there is vehicle transportation between each major customer; constraint (2) ensures that the total number of tickets delivered by the k-th vehicle to major customers does not exceed the capacity of the k-th vehicle; constraint (3) states that the k-th vehicle can choose to depart or not depart; constraint (4) describes that when the vehicle that has completed the last-mile delivery can complete the pickup task, the network point will not add more vehicles; constraint (5) ensures graph flow balance, that for each vehicle, the number of edges reaching the h-th major customer is equal to the number of edges leaving the h-th major customer; constraint (6) states that the k-th vehicle needs to go to the network point or transfer center; constraint (7) constrains the in-degree of the starting point to be 0; constraint (8) constrains the out-degree of the network point or transfer center N+1 as the destination to be 0; constraints (9) and (10) stipulate that for each vehicle, the number of paths between major customers does not exceed the starting point ( The out-degree (in-degree) of the destination; Constraint (11) stipulates that for vehicles replenished from the network point, the out-degree of the starting point and the in-degree of the destination must be equal, ensuring that the vehicles departing for delivery start from the starting point and finally return to the destination; Constraint (12) stipulates that the out-degree of the vehicle completing the last delivery must be 1, that is, a path must be generated, even though some paths start from the starting point and go directly to the network point or transfer center empty; Constraint (13) stipulates that all vehicles completing the last delivery must go to the network point or transfer center; Constraint (12) is a Miller-Tucker-Zemlin constraint, ensuring that the route trajectory of each vehicle is a directed acyclic graph; Constraint (13) is to ensure timeliness, if the time for the k-th vehicle to reach the i-th major customer point plus the travel time from the i-th major customer point to the j-th major customer point is greater than the time to reach the j-th major customer point, that is, s ik +t ij -s jk ≥0 indicates that the delivery requirements are not met. In this case, the pathway from customer i to large customer j is not connected, i.e., x ijk = 0; otherwise, the path can be chosen to be connected or disconnected based on other constraints, i.e., x ijk =0 or 1; constraints (14) and (15) specify the maximum driving time and maximum driving distance for a vehicle to complete a delivery loop; constraint (16) specifies the time window for each vehicle to arrive at each major customer's location to pick up the express delivery.

[0055] The second step involves the route planning personnel inputting the detailed addresses of each station where the vehicle is located after the last-mile delivery is completed, as well as the locations of major customers within the service area of ​​the network and related ticket volume data, to provide necessary information support for subsequent route planning.

[0056] The third step is to use Open StreetMap service to generate a distance matrix between the location of the outlet or transit center and the location of the major customer, based on the address information of the outlet or transit center and the location of the major customer, so as to quantify the spatial relationship between different locations.

[0057] The fourth step is to initialize the model solution based on the Multi-Deport scenario during the iteration process of the ALNS algorithm, and to perform specific labeling and identification of vehicle yards and destinations in the destruction and repair operators, and to perform iterative calculations for each route to ensure that the total pickup distance is short.

[0058] The fifth step, during the iteration of the ALNS algorithm, involves implementing a route-vehicle matching strategy for vehicles using the Single-trip strategy (i.e., each vehicle performs only one delivery task) to ensure efficient resource utilization. The vehicle matching strategy is as follows: First, it is determined whether all routes have suitable vehicle matches. The number of planned routes changes from more routes than vehicles to less than vehicles in each iteration of the ALNS algorithm. When the number of routes exceeds the number of vehicles, the focus is on whether the route capacity meets the vehicle's carrying capacity requirements, without strictly limiting the matching of specific routes and vehicles. When the number of routes decreases to less than or equal to the number of vehicles, it is necessary to carefully determine whether each route can find a vehicle matching its capacity. If a route can be matched with a vehicle of suitable carrying capacity, the iteration continues; if no match is found, the stop is treated as a separate route and the iteration is repeated. Regarding the route-vehicle matching strategy, this invention employs the following method: First, the ticket volume of all routes is sorted in ascending order. Then, for each route, the number of available vehicles within the vehicle carrying capacity range corresponding to its ticket volume is calculated. Only when the number of available vehicles is greater than or equal to the index of the current route (numbered starting from 1) can it be determined that a route can be matched with vehicles of suitable carrying capacity. Finally, after determining that the calculated delivery volume of the route can be delivered by the existing vehicle capacity, vehicles of the corresponding capacity are matched for each route according to the general box capacity matching rules. For example, if the calculated delivery volume of a route is 1300, 1500, 2000, and the vehicle capacity is 1400, 2000, 2400, then the vehicles matched for each route will be 1400, 2000, 2400 or 1400, 2400, 2000.

[0059] The first step is to establish a mathematical model for the Single-trip, Multi-capacity, and Multi-Deport Vehicle Routing Problem with a Time Window. This model possesses the following characteristics: the objective function is set to minimize the total travel distance of all vehicles to optimize overall delivery efficiency. Constraints cover multiple aspects. Firstly, there is the graph flow balance condition, ensuring that the final delivery endpoint serves as the starting point of the pickup route, and the network point or transfer center as the endpoint. When vehicles are insufficient, the network point serves as the starting point to supplement vehicles, maintaining flow balance at each major customer point. Specifically, when the final delivery endpoint serves as the starting point of the pickup route, its in-degree is zero; while when the network point or transfer center serves as the route endpoint, its out-degree is zero, and its in-degree is the sum of the total number of routes. Furthermore, intermediate major customer points along the route must satisfy the condition that their in-degree and out-degree are equal and both are 1, ensuring the continuity and integrity of the delivery path. Constraints include the requirement that the overall route, after the starting and ending points are individually numbered, must form a directed acyclic graph to avoid cyclic paths and logical errors. Considering the actual carrying capacity of vehicles, constraints also include that the sum of major customer tickets on each route must be less than the vehicle's carrying capacity to ensure safety and efficiency during the delivery process. To meet delivery time requirements, the constraints also include a time window limit for picking up shipments from major customers, meaning that vehicles must complete the pickup of shipments from major customers within a specified time window.

[0060] In the second step, the route planning personnel input the detailed address of the vehicle after the last-mile delivery is completed, as well as the location of major customers within the service area of ​​the branch and related ticket volume data, to obtain the address information of the branch or transfer center and major customers.

[0061] The third step uses the following method to obtain the distance matrix: Through the Open Street Map open-source map service, the shortest driving distance between the network points or transfer centers and each major customer in the actual road network is obtained. This distance data not only considers the actual direction and length of the roads, but also fully takes into account traffic rules and restrictions, thus ensuring the accuracy and practicality of the calculation results. The calculated shortest driving distances between the network points or transfer centers and each major customer are then organized into a distance matrix, intuitively displaying the spatial relationships between different locations.

[0062] In the fourth step, ALNS requires a specific initial solution algorithm and destruction / repair operators during iterative calculations. Let's first introduce the initial solution algorithm: First, place each starting point in a starting point list and each major customer point in an unvisited list. Then, iterate through each starting point to initialize the route. Each time, select the unvisited point closest to the last point on the route. If the point satisfies constraints such as vehicle capacity or pickup time, insert the unvisited point into the route and remove it from the unvisited list. Otherwise, select the next major customer point relatively close to the last transit point on the route. If no point satisfies the constraints, insert a network point or transfer center as the endpoint into the route and proceed with the next route planning until all starting points are routes or the unvisited point list is not empty. If all starting points have been visited and there are still unvisited points, set the starting point as a network point and repeat the above steps until the unvisited point list is empty. If there are still unvisited starting points but the unvisited point list is empty, directly set the starting point of the route as the unvisited starting point and the endpoint as a network point or transfer center. Next, let's introduce the destruction and repair operators: There are many operators designed for the ALNS algorithm, but it needs to be specifically noted that in the destruction operator, the start and end points of the route cannot be deleted; in the repair operator, if a route has a start point that is a network point, an end point that is a network point or a transfer center, and no intermediate points, this route is a network point supplementary route and has no actual physical meaning, so it should be deleted; however, if a route has a start point that is one of the set multiple start points, no intermediate points, and an end point that is a network point or a transfer center, then the vehicle must return empty, this route has actual physical meaning, and it cannot be deleted.

[0063] In step five, the relationship between the number of routes and the number of vehicles dynamically changes during the iterative calculation of ALNS. Adopting a single-trip strategy, where each vehicle delivers only one route, the number of planned routes will shift from being greater than the number of vehicles to being less than the number of vehicles during the iteration process. When the number of routes exceeds the number of vehicles, the focus is primarily on whether the route capacity meets the vehicle's carrying capacity requirements, without strictly limiting the matching of specific routes and vehicles. When the number of routes decreases to less than or equal to the number of vehicles, it is necessary to carefully determine whether each route can find a vehicle with a matching capacity. If a route can be matched with a vehicle of suitable carrying capacity, the iterative calculation continues; if not, the large customer point is designated as a separate route, with the starting point as the network point and the ending point as the network point or transfer center, and the iterative calculation is repeated. Regarding the route-vehicle matching strategy, this invention employs the following method: First, all routes are sorted by ticket volume in ascending order. Then, for each route, the number of available vehicles within the vehicle carrying capacity range corresponding to its ticket volume is calculated. Only when the number of available vehicles is greater than or equal to the index of the current route (numbered starting from 1) can it be determined that the route can be matched with vehicles of appropriate capacity.

[0064] 1. Compared to commercial solvers (such as Gurobi), the method of this invention achieves comparable solution quality, with a difference in solution quality ranging from -7% to 10%. However, given the typically high subscription fees required for commercial solvers, the method of this invention utilizes the open-source ALNS package, achieving a significantly cheaper computational method compared to commercial solvers and GPUs by requiring only server CPU computing resources. This not only reduces operating costs for enterprises but also makes the method more competitive in practical applications.

[0065] 2. By utilizing the Adaptive Large Neighborhood Search algorithm, the method of this invention can find high-quality solutions in a shorter time. This ensures that vehicle route planning can be completed quickly and accurately in practical applications, thereby improving the efficiency and accuracy of logistics and distribution.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation of the invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications and substitutions should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A multi-depot open vehicle route planning method based on the ALNS algorithm, characterized in that, The method specifically includes the following steps: S1. Construct a mathematical model for vehicle routing with the optimization objective of minimizing the total path length. This model includes several constraints, specifically: graph flow balance, limits on the number of large customer shipments along the route, vehicle carrying capacity limits, and vehicle pickup time window requirements. S2: The route planning personnel at the network point input the detailed addresses of each station where the vehicle is located after the last-mile delivery is completed, as well as the location of major customers within the network point's service area and related ticket volume data, to provide necessary information support for subsequent route planning. S3, based on the address information of the locations of outlets or transit centers and major customers, uses Open Street Map service to generate a distance matrix between the locations of outlets or transit centers and major customers, in order to quantify the spatial relationship between different locations; S4. During the iteration of the ALNS algorithm, the model solution is initialized based on the Multi-Deport scenario, and specific labels and identifications are made for the parking lot and the destination in the destruction and repair operators. Iterative calculations are performed on each route to ensure that the total pickup distance is short. S5. During the iteration of the ALNS algorithm, the route adopts the Single-trip strategy to match vehicles, that is, each vehicle only performs a single delivery task, and the route and vehicle matching strategy is implemented.

2. The multi-parking lot open vehicle route planning method based on the ALNS algorithm according to claim 1, characterized in that, The mathematical model for vehicle routing is the Single-trip, Multi-capacity and Multi-Deport Open Vehicle Routing Problem with Time Window. The mathematical model is set to minimize the total travel distance of all vehicles in order to optimize overall delivery efficiency.

3. The multi-parking lot open vehicle route planning method based on the ALNS algorithm according to claim 2, characterized in that, The graph flow balancing is specifically as follows: ensure that the last point of last-mile delivery is the starting point of the pickup route, and the outlet or transfer center is the ending point. When the number of delivery vehicles is insufficient to complete the pickup task, the outlet will supplement the required pickup vehicles. When the last last-mile delivery point is the starting point of the route, its in-degree is zero and its out-degree is 1. When the outlet or transfer center is the ending point of the route, its out-degree is zero and its in-degree is the sum of the total number of routes. If the total capacity of vehicles completing last-mile delivery is insufficient to complete the pickup task for all major customers, the outlet will dispatch additional vehicles, which will depart from the outlet and pass through major customers to cooperate with the vehicles completing last-mile delivery to complete the pickup task.

4. The multi-parking lot open vehicle route planning method based on the ALNS algorithm according to claim 1, characterized in that, The route planners input the detailed address of the vehicle after the last-mile delivery is completed, as well as the location of major customers within the service area of ​​the branch and related ticket volume data, to obtain the address information of the branch or transfer center and major customers.

5. The multi-parking lot open vehicle route planning method based on the ALNS algorithm according to claim 1, characterized in that, The distance matrix is ​​obtained by using the Open Street Map open-source map service to obtain the shortest driving distance between the network points or transfer centers and each major customer in the actual road network. The calculated shortest driving distances between the network points or transfer centers and each major customer are then organized into a distance matrix, which intuitively shows the spatial relationship between different locations.

6. The multi-parking lot open vehicle route planning method based on the ALNS algorithm according to claim 1, characterized in that, During the iterative calculation process, the ALNS performs a specific initial solution algorithm and destruction / repair operators; the initial solution algorithm is as follows: Place each starting point in the starting point list and each major customer point in the unvisited list; iterate through each starting point to initialize the route. Each time, select the unvisited point closest to the last point of the route. If the point meets the constraints such as vehicle capacity and pickup time, insert the unvisited point into the route and delete it from the unvisited list. Otherwise, select the next major customer point that is relatively close to the last transit point of the route. If no point meets the constraints, insert a network point or transfer center into the route as the endpoint and proceed with the next route planning until all starting points are routes or the unvisited point list is not empty. If all starting points have been visited and there are still unvisited points, then set the starting points as network points and repeat the above steps until the list of unvisited points is empty. If there are still unvisited starting points but the list of unvisited points is empty, the route will directly set the starting point as the unvisited starting point and the destination as a network point or transfer center. Specifically, the destruction and repair operators are defined as follows: In the destruction operator, the start and end points cannot be deleted from the route; In the repair operator, if a route starts at a network point, ends at a network point or transfer center, and has no intermediate points, this route is a network point supplementary route and has no actual physical meaning, so the route is deleted; if a route starts at one of the set multiple starting points, has no intermediate points, and ends at a network point or transfer center, then the vehicle must return empty, this route has actual physical meaning and cannot be deleted.

7. The multi-parking lot open vehicle route planning method based on the ALNS algorithm according to claim 1, characterized in that, The specific route and vehicle matching strategy is as follows: when the number of routes is greater than the number of vehicles, it is confirmed whether the route capacity meets the requirements of the vehicle carrying capacity; when the number of routes is reduced to less than or equal to the number of vehicles, it is determined whether each route can find a vehicle that matches its capacity. If the route can be matched with vehicles of suitable carrying capacity, the calculation continues iteratively; If a match cannot be found, the major customer point will be treated as a separate route, with the starting point being the branch point and the ending point being either the branch point or the transit center, and the iterative calculation will be performed again.

Images