Multi-warehouse multi-vehicle path planning method based on adaptive neighborhood ant colony system

Abstract

The application discloses a multi-warehouse multi-logistics vehicle path planning method based on an adaptive neighborhood ant colony system, proposes an adaptive warehouse selection strategy based on historical evolution information, autonomously selects appropriate warehouses for each logistics vehicle, and achieves the purpose of optimizing warehouse positions, designs a customer selection mechanism based on adaptive neighborhoods for the ant colony system, and effectively reduces the complexity of customer selection; through continuous iteration optimization, finally outputs high-quality and high-precision multi-logistics vehicle service paths, balances the service costs of each logistics vehicle while reducing the total service cost of all logistics vehicles. The application is solved through the ant colony optimization algorithm, adaptive warehouse selection strategies, a customer selection mechanism based on adaptive neighborhoods, an ant selection technology based on path estimated service costs, and a path merging and segmentation strategy fusing 2-opt, so that the quality of the path planning scheme is significantly improved, and the time consumed for constructing the planning scheme is greatly shortened.

Description

Technical Field

[0001] This invention belongs to the fields of logistics scheduling optimization, path planning and computational intelligence technology, and in particular relates to a multi-warehouse and multi-logistics vehicle path planning method based on an adaptive neighborhood ant colony system. Background Technology

[0002] With the deep application of cutting-edge technologies such as the Internet of Things, cloud computing, and artificial intelligence, the modern logistics industry is undergoing a transformation and upgrade from a traditional passive response model to an intelligent decision-making and refined operation model. In the field of logistics scheduling, current research largely focuses on vehicle scheduling optimization problems within a single warehousing center. However, with the evolution of business models, complex scheduling needs involving multiple warehousing nodes and uncertain warehouses have emerged, making existing logistics scheduling solutions based on single or fixed warehouses difficult to adapt to new business scenarios. Furthermore, conventional route planning methods often fail to fully consider the balance of task allocation among vehicles, leading to unreasonable planning results. This can not only extend delivery cycles but also increase operating costs, ultimately affecting the overall operational efficiency of the logistics system. Moreover, existing algorithms are mostly designed for small- to medium-scale problems, while the scale and complexity of real-world application scenarios continue to rise, making it difficult for traditional algorithms to guarantee both efficiency and quality. Summary of the Invention

[0003] Purpose of the invention: To address the problem of uncertain multi-warehouse and multi-vehicle route planning, a multi-warehouse and multi-vehicle route planning method based on an adaptive neighborhood ant colony system is provided.

[0004] Technical solution: The present invention provides a multi-warehouse, multi-vehicle route planning method based on an adaptive neighborhood ant colony system, comprising the following steps:

[0005] Step 1: Read data information, including logistics vehicle information, number of customers, customer location information, and service cost information between customers; construct an uncertain multi-warehouse, multi-logistic vehicle route planning model, and establish an optimization objective function to evaluate the quality of the route planning scheme;

[0006] Step 2: Based on customer location information and service cost information between customers, construct a service cost matrix between customers and initialize a heuristic information matrix; initialize the ant colony system algorithm parameters, and use a greedy algorithm to construct an uncertain multi-warehouse and multi-logistics vehicle route planning scheme, and initialize the pheromone matrix and warehouse weight vector;

[0007] Step 3: Construct an ant team, where each ant team contains the same number of ants as the logistics vehicle. Each ant is responsible for constructing a service path for one logistics vehicle. Each ant in each team adaptively selects a warehouse as the starting point of the path based on the warehouse weight vector. Subsequently, each ant constructs a service path for the logistics vehicle for each customer one by one, based on the proposed adaptive neighborhood method and the ant selection technique based on the estimated path service cost, combined with the heuristic information matrix and pheromone matrix. At the same time, the pheromone matrix is ​​locally updated during the path construction process.

[0008] Step 4: Based on the proposed path merging and segmentation strategy of fusion 2-opt, adaptively perform local search operation on the logistics vehicle path planning scheme constructed by the Ant team according to probability;

[0009] Step 5: Set optimization objectives, evaluate each path planning scheme, and update the globally optimal planning scheme; use the globally optimal planning scheme to globally update the pheromone matrix; then, update the warehouse weight vector according to each path planning scheme.

[0010] Step 6: Repeat steps 3 to 5 until the iteration termination condition is met, and finally output the globally optimal logistics vehicle route planning scheme.

[0011] Furthermore, step 1 specifically involves: the information read includes the number of customers N, and the customer coordinates (x, y). i , y i ), where x i , y i Let x and y represent the x and y coordinates of the i-th customer, respectively; the service cost e between customers. ij Number of logistics vehicles: M;

[0012] The specific steps for constructing an uncertain multi-warehouse, multi-vehicle route planning model are as follows: Given N customers and M logistics vehicles, the vehicle service map is modeled as a completely undirected graph G = (V, E), where V = {(x i , y i Let { ) | 1≤i≤N} represent the customer set; E = { e ij | i, j∈V} is the set of edges between all customers; each edge is assigned a weight, which represents the service cost from customer i to customer j;

[0013] Define the optimization goal:

[0014]

[0015]

[0016]

[0017] Where R represents the planning scheme, E k R represents the total service cost of the k-th path. k This represents the k-th path in the planning scheme. This represents the service cost between the i-th customer on the k-th path and the (i+1)-th customer on the k-th path. Represents the first customer and the k-th path. The service cost between n customers k Let n represent the number of customers along the k-th path. i This represents the number of customers along the i-th path.

[0018] Furthermore, step 1 specifically includes: the ant colony system algorithm includes the number of ant teams NP, pheromone factor α, heuristic pheromone factor β, local pheromone evaporation rate ξ, global pheromone evaporation rate ρ, and pseudo-random probability q0;

[0019] Heuristic information is generated based on the service costs between customers. The heuristic information η between two customers i and j is... i,j The calculation formula is:

[0020]

[0021] e i,j This represents the service cost between customer i and customer j.

[0022] Constructing a greedy multi-vehicle route planning scheme using a greedy algorithm includes the following steps:

[0023] Step 2-1: Initialize the ant colony, which consists of M ants;

[0024] Step 2-2: Each ant in the ant team randomly selects a customer from the customer set as the starting point, selects the next customer based on the heuristic information between customers, and begins to greedily construct a logistics vehicle route planning scheme; the ant team initializes a customer tabu table tabu, which contains all the selected initial customers, and initializes the path service cost of each ant to 0;

[0025] Steps 2-3: Each ant in the ant team obtains its own accessible customer set (candidates) based on the customer taboo list. Then, each ant pre-selects customer j from its own candidates that has the most heuristic information about its current customer i. Then, each ant adds the pre-selected customer to its own constructed service path and calculates the estimated path service cost for each ant.

[0026] Steps 2-4: Select the ant with the lowest estimated path service cost to build the path; add the customer pre-selected by this ant to the constructed logistics vehicle path, add it to the taboo list, and update the corresponding path service cost; other ants should abandon their pre-selected customers.

[0027] Steps 2-5: Repeat steps 2-3 to 2-4 until each ant in the ant team can no longer select customers, that is, all customers are served by logistics vehicles, thus obtaining a greedy uncertain multi-warehouse multi-logistic vehicle route planning scheme.

[0028] Steps 2-6: Use the fitness value of the greedy uncertain multi-warehouse multi-logistics vehicle route planning scheme to assign the initial value of pheromone τ0, and then initialize the pheromone in all pheromone matrices to τ0.

[0029] Furthermore, in steps 2-3, the accessible customer set `candidates` is obtained using the following formula:

[0030]

[0031] Where V represents the set of all customers, and tabu represents the tabu list;

[0032] In steps 2-3, the estimated service cost of Ant Group is calculated using the following formula:

[0033]

[0034] Among them, future_expense i Represents the i-th path R i The estimated service cost, n i The representative represents the number of customers along the i-th path, pre i Represents path R i Corresponding to the customers pre-selected by Ant Group, On behalf of clients With the client i Service costs between On behalf of clients With the client i Service costs between E i Indicates the cost of path services;

[0035] In steps 2-4, the ants that construct the logistics vehicle route are selected according to the following formula: the ant with the lowest estimated service cost is selected for route construction.

[0036]

[0037] Where M represents the number of logistics vehicles; after the selected ant places the pre-selected customer into its path, the corresponding path service cost is updated using the following formula. E i Number of customers along the path n i :

[0038]

[0039]

[0040] In steps 2-6, the initial pheromone value τ0 is assigned using the maximum path service cost in the greedy path planning scheme. The calculation formula is as follows:

[0041]

[0042] Where N represents the number of customers, E greedy This represents the maximum path service cost in the solution constructed by the greedy algorithm.

[0043] Furthermore, step 3 specifically includes the following steps:

[0044] Step 3-1: Construct NP ant teams, each team containing M ants;

[0045] Step 3-2: Use an adaptive warehouse selection strategy to determine the starting customers of all ants in the ant team; then start building a logistics vehicle route planning scheme; initialize a customer tabu table tabu, which contains the starting customers of all routes, and initialize the route service cost of each ant to 0;

[0046] Step 3-3: Each ant in the ant team pre-selects the next customer based on an adaptive neighborhood-based customer selection mechanism;

[0047] Steps 3-4: Use ant selection technology based on estimated path service cost to select the logistics vehicle path for ants; add the pre-selected customers of the selected ants to the logistics vehicle path they build, and add them to the customer taboo list of the ant team, and update the corresponding path service cost; at the same time, call the local pheromone update rule to update the pheromone on the edge that the ant is currently passing through in real time, and other ants abandon their pre-selected customers.

[0048] Steps 3-5: Repeat steps 3-3 and 3-4 until all customers are selected and enter the tabu, thus obtaining an uncertain multi-warehouse and multi-vehicle route planning scheme;

[0049] Steps 3-6: Repeat steps 3-2 to 3-5 until all Ant teams have built uncertain multi-warehouse and multi-vehicle route planning schemes.

[0050] Furthermore, step 3-2 specifically includes the following steps:

[0051] Step 3-2-1: Initialize candidates, which includes all customers; initialize each customer as the initial weight w of the warehouse. i ;

[0052] Step 3-2-2: Based on the pheromone concentration, calculate the probability that all customers in the set are selected. The probability is calculated as follows:

[0053]

[0054] Among them, w i ,w j Let i and j represent the weights of customers i and j in the warehouse weight vector, respectively, and α represent the pheromone factor.

[0055] A roulette wheel selection mechanism is used to select a customer as a warehouse, which is then removed from the candidates, and a local weight update is performed. The method for the local weight update is as follows:

[0056]

[0057] Where τ0 is the initial value of the pheromone. This refers to the local evaporation rate.

[0058] Step 3-2-3: Repeat step 3-2-2 until M warehouses are found. Assign each warehouse to one of the M ants in the ant team as its warehouse.

[0059] Furthermore, in step 3-3, the customer selection mechanism is as follows: taking the last customer i of the current path k as the current customer, the probability of selecting customer j is calculated according to the following formula:

[0060]

[0061] in, For the pheromone concentration from customer i to customer j, η i,j For heuristic information from customer i to customer j, For the pheromone concentration from customer i to customer l, η i,l For heuristic information from customer i to customer l, α and β are the pheromone factor and heuristic information factor, respectively. `candidates` represents the candidate customer set, constructed based on a tabu list and an adaptive domain. After obtaining the probability of each customer in the candidate customer set, a roulette wheel selection mechanism is used to select a customer. Based on the probability of each customer in the candidate customer set, the next visiting customer is selected using the following formula:

[0062]

[0063] Where j* represents the customer selected using the roulette wheel strategy according to the calculated customer selection probability, rand(0,1) represents a random number between 0 and 1, q0 represents a probability parameter indicating that the ant will greedily select the customer with the strongest pheromone concentration and heuristic information with probability q0, and p i,l Let i be the probability of selecting customer l as the next node with i as the current node, and let l be the probability of selecting a customer according to the roulette wheel selection mechanism with a probability of (1-q0).

[0064] Furthermore, step 4 specifically includes the following steps:

[0065] Step 4-1: Based on the fitness values ​​of all the solution schemes constructed by the ant teams, calculate the probability of each solution scheme performing a local search, as shown in the following formula:

[0066]

[0067] Where, fit(R) k Let be the fitness value of the k-th solution, and max(fit) and min(fit) be the maximum and minimum fitness values ​​among all ant teams.

[0068] For each solution, a random number is generated and compared with the corresponding probability to determine whether a local search operation will be performed.

[0069] Step 4-2: For each solution that requires a local search operation, list two different logistics vehicle service paths within the solution.

[0070] Step 4-3: Randomly select a breakpoint for each of the two paths, connect the two paths together at the breakpoint, and perform 2-opt optimization on the merged path;

[0071] 2-opt optimization specifically includes the following steps:

[0072] Step 4-3-1: List any two customers in the merging path, swap these two customers and the customer sequence between them, and calculate the resulting reduction in service costs. If the reduction is positive, then apply this swap. Let t be the customer preceding customer i in the merging path, and l be the customer following customer j. The formula for calculating the total path cost reduction after swapping customer i and customer j and the customer sequence between them is:

[0073]

[0074] Among them, Δreduce_expense i,j e represents the reduction in service costs resulting from swapping customer i, j and the customer sequence between customers i, j.t,i e represents the service cost between customers t and i. j,l Represents the service cost between customers j and l, e t,j e represents the service cost between customers t and j. i,l This represents the service cost between customers i and l;

[0075] Δreduce_expense i,j If the result is positive, then this swap will be applied, and the total service cost of the merged path will be updated. The calculation formula is as follows:

[0076]

[0077] Among them, E merge This represents the total service cost of the merged path;

[0078] Step 4-3-2: Repeat step 4-3-1 until swapping any two customers in the merge path and the customer sequence between the two customers will not reduce the service cost.

[0079] Step 4-4: Find the optimal split point on the optimized path so that the maximum service cost of the two split paths is minimized. If the maximum service cost of the split path is less than the maximum service cost of the original two paths, the local search is successful. Therefore, the two paths after the local search are used to replace the original two paths. Otherwise, the local search is unsuccessful, the optimization is abandoned, and no replacement is made.

[0080] Steps 4-5: Repeat steps 4-2 to 4-4 until any two paths in the plan can no longer be optimized. At this point, the current planning plan has been optimized.

[0081] Steps 4-6: Repeat steps 4-1 to 4-5 until all scheduling schemes requiring local search have completed the designed local search strategy.

[0082] Furthermore, step 5 specifically includes the following steps:

[0083] Step 5-1: Compare the fitness values ​​of all multi-logistics vehicle route planning schemes of the Ant Team and find the optimal solution in the iteration.

[0084] Step 5-2: Compare the fitness value of the found iterative optimal solution with the fitness value of the global optimal solution; if the iterative optimal solution is better, then replace the global optimal solution with the iterative optimal solution.

[0085] During the pheromone update phase, a globally optimal solution is used to update the pheromone concentration, and the pheromone concentration update formula is as follows:

[0086]

[0087] Where ρ is the global pheromone evaporation coefficient, and R best This is the globally optimal solution. The pheromone concentration from customer i to customer j;

[0088] In step 5, the warehouse weight vector update formula is:

[0089]

[0090] w i This indicates that customer i is the weight of the warehouse in the warehouse weight vector.

[0091] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method of the present invention.

[0092] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0093] 1. This invention optimizes customer selection and service order for multiple logistics vehicles by using an adaptive warehouse selection strategy, an adaptive neighborhood-based customer selection mechanism, ant selection technology based on path prediction service cost, and a path merging and splitting strategy that integrates 2-opt. This minimizes the total service cost of all logistics vehicle paths while maintaining the balance of service costs among different logistics vehicles to the greatest extent, thereby improving the efficiency of logistics scheduling optimization and reducing transportation costs.

[0094] 2. This invention addresses the multi-vehicle route planning problem under uncertain multi-warehouse conditions, solving the problem of route planning for logistics vehicles when warehouse locations are uncertain. By minimizing the maximum route service cost, it can minimize the total service cost of all routes while ensuring a balance in service costs among all routes.

[0095] 3. This invention employs an ant colony optimization algorithm, designing an adaptive warehouse selection strategy, a customer selection mechanism based on adaptive neighborhood, ant selection technology based on path prediction service cost, and a path merging and splitting strategy incorporating 2-opt. These significantly improve the quality of the path planning scheme while greatly reducing the time consumed in constructing the scheme. Applying these innovations to the uncertain multi-warehouse, multi-vehicle path planning problem can yield high-quality multi-vehicle path planning schemes in a shorter time, thereby improving the efficiency of logistics scheduling optimization and reducing transportation costs. Attached Figure Description

[0096] Figure 1This is a customer distribution map of an example of uncertain multi-warehouse, multi-vehicle route planning in this invention; the example includes 3 logistics vehicles and 150 customers.

[0097] Figure 2 This is a flowchart illustrating an uncertain multi-warehouse, multi-logistics vehicle planning method based on ant colony systems proposed in this invention.

[0098] Figure 3 This invention uses adaptive warehouse selection technology to select three warehouses for three logistics vehicles during the warehouse selection phase in an example.

[0099] Figure 4 This is a schematic diagram of the customer selection mechanism based on adaptive neighborhood proposed in this invention in an example.

[0100] Figure 5 This is a schematic diagram of the ant selection technology based on path prediction service cost proposed in this invention in an example.

[0101] Figure 6 This is an uncertain multi-warehouse, multi-vehicle planning scheme diagram constructed by the ant colony system algorithm proposed in this invention for an instance during the path construction stage.

[0102] Figure 7 This is a diagram of an uncertain multi-warehouse, multi-logistics vehicle planning scheme after employing local search, as an example of the present invention.

[0103] Figure 8 This invention provides the globally optimal, uncertain multi-warehouse, multi-vehicle planning scheme that the method proposed in this invention ultimately outputs for the example. Detailed Implementation

[0104] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. This embodiment is based on the technical solution of the present invention and provides detailed implementation methods and specific operating procedures; however, the scope of protection of the present invention is not limited to the following embodiment. This example uses the distance between customers as the service cost and the exhaustion of the maximum fitness evaluation count as the algorithm iteration termination condition.

[0105] Figure 2 This invention demonstrates a multi-warehouse, multi-vehicle path planning method based on an adaptive neighborhood ant colony system. For example... Figure 1 This paper presents a customer distribution map for an uncertain multi-warehouse, multi-vehicle route planning problem in an example of the present invention; the example includes 150 customers and 3 logistics vehicles. The goal is to minimize the maximum service cost among the 3 logistics vehicles while visiting all customers. The method of the present invention specifically includes the following steps:

[0106] Step 1: Obtain instance information for uncertain multi-warehouse, multi-vehicle route planning; construct an uncertain multi-warehouse, multi-vehicle route planning model integrating customer information and logistics vehicle information; establish an optimization objective function to evaluate the quality of the route planning scheme. The instance information includes the number of customers N, customer coordinates (x, y, y), and other relevant information. i , y i And the number of logistics vehicles M, where, x i , y i These represent the x and y coordinates of the customer, respectively. N=150, M=3, and the specific data for the customer's coordinates are shown in Table 1.

[0107] Table 1

[0108]

[0109] In step 1, the algorithm's optimization objective is to find the optimal route planning scheme that minimizes the maximum service cost among the three logistics vehicles, ensuring that each customer must be visited by exactly one logistics vehicle. Each logistics vehicle route is a sequence of customers, with the same customer as the starting and ending point of each sequence.

[0110] In step 2, the ant colony optimization algorithm parameters are first initialized, including the number of ant colonies NP, the global pheromone evaporation rate ρ, and the local pheromone evaporation rate. Heuristic information factor α, pheromone factor β, pseudo-random parameter q0, upper and lower bounds of the adaptive domain size max and size min The maximum fitness evaluation count is set to `max_Eval`. In this example, the ant colony size is set to 10, the global pheromone evaporation rate `ρ` and the local pheromone evaporation rate are both set to 0.1, the heuristic pheromone factor `α` and the pheromone factor `β` are both set to 1, the pseudo-random parameter `q0` is set to 0.9, and the upper limit of the adaptive domain is set to `size`. max and lower limit size min The parameters were set to 40 and 10 respectively, and the maximum fitness evaluation count was set to 5000. All of these parameters were experimentally optimized and represent the best parameter combination among all available options.

[0111] The heuristic information is generated based on the service cost between customers. In this example, the service cost between customers is set as the Euclidean distance between two customers. Subsequently, the greedy algorithm constructs a greedy planning scheme by including the following steps:

[0112] Step 2-1: Initialize an ant team consisting of 3 ants; then, initialize a multi-vehicle route planning scheme, represented as greedy_solution = {tour_route1, tour_route2, tour_route3}.

[0113] Step 2-2: Each ant in the ant team randomly selects a customer from the customer set as its starting point. Then, based on heuristic information among the customers, it selects the next customer to visit and begins greedily constructing a logistics vehicle route planning scheme. In this example, the three ants randomly select starting customers 12, 36, and 71, respectively. Therefore, the initial planning scheme is greedy_solution = {

[12] ,

[36] ,

[71] }. Subsequently, the ant team initializes the customer tabu list tabu = {12,36, 71}, and initializes the service cost of each ant to 0.

[0114] In steps 2-3, each ant in the ant team obtains its own accessible customer set `candidates` based on the customer taboo list. Then, each ant pre-selects a customer `j` from its accessible customer set `candidates` that has the most heuristic information about its current customer `i`. Next, each ant adds its pre-selected customer to its constructed service path and calculates the estimated service cost for each ant.

[0115] In this example, after several customer selections, the current planning solution of the three ants in the ant team is: greedy_solution = {[71, 20, 73, 140, 58, 126, 35, 37, 23, 17, 136, 78, 105, 89, 48, 5, 62, 0, 129, 91], [12, 32, 145, 102, 75, 94, 125, 81, 115, 49, 43, 113, 143, 63, 39, 53, 1, 72], [36, 4, 77, 51, 95, 38, 100, 120, 29, 106, 111, 131, 28, 45,} [2, 135, 42, 122, 114, 119]}, the next customers pre-selected by the three ants are 7, 67, and 148 respectively. Adding the pre-selected customers to the paths of their respective ants, the estimated service costs for the three ants are calculated to be 4362.85, 4371.01, and 4357.93 respectively.

[0116] Steps 2-4: Select the ant with the lowest estimated service cost to construct the path. Add the customer pre-selected by this ant to its service path, add it to the taboo list, and update the corresponding service cost. Other ants abandon their pre-selected customers.

[0117] In the above example, Ant 3 has the lowest estimated service cost, so Ant 3 was chosen for path building.

[0118] After the selected ant puts its pre-selected customers into the path, update the corresponding service cost and number of customers;

[0119] In this example, the original service cost of Ant 3 was 3883.35. After adding the pre-selected customers of Ant 3 to the path, the service cost increased by 271.45. Therefore, the updated service cost of Ant 3 is 4154.80; other Ants abandoned their pre-selected customers. Subsequently, the current planning scheme `greedy_solution` is defined as: {[71, 20, 73, 140, 58, 126, 35, 37, 23,17, 136, 78, 105, 89, 48, 5, 62, 0, 129, 91], [12, 32, 145, 102, 75, 94, 125,81, 115, 49, 43, 113, 143, 63, 39, 53, 1, 72], [36, 4, 77, 51, 95, 38, 100,120, 29, 106, 111, 131, 28, 45, 2, 135, 42,} 122, 114, 119, 148]};

[0120] Steps 2-5 are repeated, from steps 2-3 to 2-4, until all customers are selected, thus obtaining a greedy, uncertain multi-warehouse, multi-vehicle route planning scheme. In this example, the greedy programming solution is: greedy_solution = {[71,20, 73, 140, 58, 126, 35, 37, 23, 17, 136, 78, 105, 89, 48, 5, 62, 0, 129,91, 7, 41, 121, 79, 30, 88, 132, 137, 147, 141, 104, 66, 27, 92, 130, 46,108, 90, 97, 22, 44, 31, 10, 14, 16, 112, 9, 83, 103, 101, 110, 98, 93, 21,15, 87, 133, 52, 18, 74, 96, 142, 55, 138, 118, 3, 64, 65, 128, 25, 69], [12,32, 145, 102, 75, 94, 125, 81, 115, 49, 43, 113, 143, 63, 39, 53, 1, 72, 67,84, 80, 24, 60, 57, 107, 68, 146, 47, 99, 70, 40, 117, 123], [36, 4, 77, 51,95, 38, 100, 120, 29, 106, 111, 131, 28, 45, 2, 135, 42, 122, 114, 119, 148,54, 82, 33, 134, 139, 116, 6, 8, 56, 144, 86, 124, 50, 19, 11, 26, 85, 34,149, 61, 59, 76, 109, 127, 13]}.

[0121] Steps 2-6: The fitness value of the greedy, uncertain multi-warehouse, multi-vehicle route planning scheme is used to assign the initial pheromone value τ0; then, the pheromone values ​​in the pheromone matrix are initialized to τ0. The pheromone matrix is ​​a 150×150 two-dimensional matrix [τ...]. ij ] 150×150， τ ij τ represents the pheromone concentration between clients i and j. At the initial stage of the algorithm, τ ij =τ0.

[0122] Simultaneously, the warehouse weight vector is also initialized in the same way. In this example, the warehouse weight vector is a 150*1 vector, [wi ] 150 , w i Let customer i be the weight of the warehouse. At the beginning of the algorithm, w... i =τ0.

[0123] In step 3, the process of maintaining an ant colony and constructing a planning scheme for each ant colony includes the following steps:

[0124] Step 3-1: Maintain NP ant teams, each team containing M ants. Each ant team is responsible for constructing an uncertain multi-warehouse and multi-vehicle route planning scheme.

[0125] Step 3-2: Adaptive warehouse selection strategy is adopted for warehouse selection. The adaptive warehouse selection strategy includes: using a probabilistic selection mechanism to determine the starting customers of all ants in the ant team based on the warehouse weight vector, and updating the warehouse weight vector locally according to the determined warehouses. Then, the logistics vehicle route planning scheme is constructed. A customer tabu table tabu is initialized, which contains warehouses for all routes, and the service cost of each ant is initialized to 0. In this example, the three ants select warehouses 0, 1, and 2 according to the adaptive warehouse selection strategy. At this time, the planning scheme solution={[0], [1], [2]}, tabu={0, 1, 2}, and the route service cost of each ant is initialized to 0. At this time, the multiple logistics vehicle warehouses are as follows: Figure 3 As shown.

[0126] The aforementioned adaptive warehouse selection strategy utilizes historical evolution information to autonomously select suitable warehouses for each logistics vehicle, and continuously updates the strategy during iterations to optimize warehouse locations.

[0127] Step 3-3: An adaptive neighborhood-based customer selection mechanism is used to determine the ants' pre-selected customers. This mechanism involves each ant in the ant colony first obtaining its own set of accessible customers (candidates) based on a customer taboo list and an adaptive neighborhood technique. Then, a probabilistic selection mechanism is used to pre-select the next customer to visit from these candidates. The adaptive neighborhood technique is used to dynamically adjust the range of candidate customers. Initially, the neighborhood range is set to the maximum value, size. max As the number of unvisited customers decreases, the neighborhood shrinks according to the following formula:

[0128]

[0129] Among them, V free Let V be the set of unvisited customers, and let V be the set of all customers, with size... max and size minThese are the upper and lower bounds of the neighborhood range, respectively. Finally, based on the size of the neighborhood range, the candidates are determined using the following formula:

[0130]

[0131] This refers to the number of nodes that have the lowest service cost with the current customer. In this example, after several iterations, the solution = {[0, 129, 91, 7, 41, 121, 79, 30, 88, 132, 137, 147, 141, 104, 66, 107, 57], [1, 143, 113, 43, 49, 72, 67, 84, 134, 139, 116, 6, 8, 56, 144, 86, 124, 50, 60], [2, 45, 28, 131, 111, 106, 120, 29, 100, 38, 95, 135, 42, 122, 114, 119, 148,} 54]}, at this time, for path 1 constructed for ant 1, with 57 as the current customer, the calculated size is 29, and the candidates at this time are [27, 92, 130, 24, 80, 68, 109, 76, 97, 22, 46, 90, 59, 63, 19, 61, 39, 108, 149, 53, 11, 44, 62, 85, 31, 26, 82, 34, 5], as Figure 4 As shown, the circles represent the candidate customers of Ant 1.

[0132] The aforementioned customer selection mechanism based on adaptive neighborhood effectively reduces the complexity of customer selection and eliminates redundant searches, thereby effectively improving the efficiency and accuracy of the algorithm.

[0133] Subsequently, the last customer on the current path is taken as the current customer. After obtaining the candidates, a probabilistic selection mechanism is used to select the pre-selected customers for this path. In this example, when the current customer is 57, based on the aforementioned candidate customer set and according to the probabilistic selection mechanism, the final selected customer is 27.

[0134] Steps 3-4 involve ant selection using ant selection based on path prediction service cost. In this example, solution1 = {[0, 129, 91, 7, 41, 121, 79, 30, 88, 132, 137, 147, 141, 104,66, 107, 57], [1, 143, 113, 43, 49, 72, 67, 84, 134, 139, 116, 6, 8, 56, 144,86, 124, 50, 60], [2, 45, 28, 131, 111, 106, 120, 29, 100, 38, 95, 135, 42,122,} Given the numbers 114, 119, 148, and 54, the next customers pre-selected by the three ants are 27, 24, and 82 respectively. Therefore, the estimated service costs for the three paths are 3817.554, 3855.483, and 3825.8982 respectively. The three paths are then as follows: Figure 5 As shown, the triangle represents the customers pre-selected by the three ants.

[0135] Select the ant in the ant team with the lowest estimated service cost to build a service path. Add the ant's pre-selected customer to its built logistics vehicle path, and simultaneously add the customer to the ant team's customer taboo list, updating the ant's service cost. Call the local pheromone update method to instantly update the pheromone along the edge from the current customer to the pre-selected customer. Other ants abandon their pre-selected customers. In this example, since path 1 has the lowest future service cost, path construction is performed on the ants that choose this path. The resulting planning scheme is: solution1 = {[0, 129,91, 7, 41, 121, 79, 30, 88, 132, 137, 147, 141, 104, 66, 107, 57, 27], [1,143, 113, 43, 49, 72, 67, 84, 134, 139, 116, 6, 8, 56, 144, 86, 124, 50, 60],[2, 45, 28, 131, 111, 106, 120, 29, 100, 38, 95, 135, 42, 122, 114, 119, 148,54]}.

[0136] The ant selection technique based on path prediction service cost described above can fully consider changes in the search environment and balance the service cost of each path as much as possible while minimizing the total service cost of all paths.

[0137] Subsequently, the pheromone matrix is ​​updated by calling the local pheromone update rule. In this example, Ant 1's logistics vehicle path has added a new customer 27, so the pheromone on the edge from customer 47 to customer 27 needs to be updated.

[0138] Steps 3-5 are repeated from steps 3-3 to 3-4 until all customers have been served, thus obtaining an uncertain multi-warehouse and multi-vehicle route planning scheme. In this example, an uncertain multi-warehouse, multi-vehicle route planning scheme is obtained as follows: solution1 = {[0, 129, 91, 7, 41, 121, 79, 30, 88, 132, 137, 147, 141, 104, 66, 107, 57, 27, 92, 130, 46, 108, 90, 97, 22, 44, 31, 10, 14, 16, 140, 58, 73, 20, 71, 112, 9, 83, 35, 37, 23, 17, 136, 78, 105, 89, 55, 142, 118, 96, 74, 18, 52,} 133, 87, 15, 93, 21, 69, 65, 64, 3, 117, 123, 25, 128],[1, 143,113, 43, 49, 72, 67, 84, 134, 139, 116, 6, 8, 56, 144, 86, 124, 50, 60, 24,80, 68, 63, 39, 53, 115, 81, 125, 94, 12, 32, 145, 102, 75, 36, 4, 77, 51,146, 138, 48, 5, 62, 103],[2, 45, 28, 131, 111, 106, 120, 29, 100, 38, 95, 135, 42, 122, 114, 119, 148, 54, 82, 33, 11, 26, 85, 34, 149, 61, 59, 76, 109, 19, 127, 40, 70, 99, 47, 13, 126, 98, 101, 110]}. For example Figure 6 This example demonstrates the multi-vehicle route planning scheme described above, which includes vehicle route 1, vehicle route 2, and vehicle route 3, with service costs of 14054.36, 14579.40, and 14766.13, respectively.

[0139] Steps 3-6, repeat steps 3-2 to 3-5, until all Ant teams have built uncertain multi-warehouse and multi-logistics vehicle route planning schemes.

[0140] Step 4: Perform local search using a path merging and segmentation strategy based on fusion 2-opt. The specific process of the local search is as follows:

[0141] Step 4-1: Based on the fitness values ​​of all path planning schemes built by the ant teams, calculate the probability of each planning scheme to perform a local search, generate a random number for each planning scheme, and compare it with the corresponding probability to determine whether a local search operation is required.

[0142] Step 4-2: For a planning scheme that requires a local search operation, enumerate the paths of any two different logistics vehicles within the scheme. In this example, the two enumerated paths are: [1, 143, 113, 43, 115, 49, 72, 67, 84, 134, 139, 116, 6, 8, 56, 144, 86, 124, 50, 60, 24, 80, 68, 63, 39, 53, 138, 55, 142, 118, 96, 74, 18, 52, 133, 17, 23, 37, 103, 110, 101, 98, 9, 83, 35, 126], [0, 129, 91, 7, 41, 121, 79, 30, 88, The service costs for the following numbers are 12553.95 and 11466.47 respectively: 132, 137, 147, 141, 104, 66, 107, 57, 27, 92, 130, 46, 108, 90, 97, 22, 44, 31, 10, 14, 16, 140, 58, 73, 20, 71, 112, 48, 5, 62, 89, 105, 78, 136, 87, 15, 21, 93, 69, 65, 64, 3, 117, 123, 25, 128.

[0143] Step 4-3: Randomly select a breakpoint for these two paths, then break the two paths at the breakpoint and merge them together, and then perform 2-opt optimization on the merged path. In this example, the selected breakpoints are 128 and 1, and the final merged path is: [0, 129, 91, 7, 41, 121, 79, 30, 88, 132, 137, 147, 141, 104, 66, 107, 57, 27, 92, 130, 46, 108, 90, 97, 22, 44, 31, 10, 14, 16, 140, 58, 73, 20, 71, 112, 48, 5, 62, 89, 105, 78, 136, 87, 15, 21, 93, 69, 65, 64, 3, 117, 123, 25, 128, 1, 143, 113, 43, 115, 49, 72, 67, 84, 134, 139, 116, 6, 8, 56, 144, 86, 124, 50, 60, 24, 80, 68, 63, 39, 53, 138, 55, 142, 118, 96, 74, 18, 52, 133, 17, 23, 37, 103, 110, 101, 98, 9, 83, 35, 126).

[0144] The merged path is then optimized using 2-opt. The specific steps of 2-opt optimization are as follows: Enumerate any two customers in the merged path, swap these two customers and the customer sequence between them, and calculate the resulting reduction in path service cost. If the reduction is positive, the swap is applied. Repeat the above process until swapping any two customers in the merged path and the node sequence between them no longer reduces the total service cost.

[0145] In this example, the resulting merged path is: [0, 129, 91, 7, 41, 121, 79, 30, 88, 132, 137, 147, 141, 104, 66, 107, 57, 27, 92, 130, 46, 108, 90, 97, 22, 44, 31, 10, 14, 16, 140, 58, 73, 20, 71, 112, 48, 5, 62, 89, 105, 78, 136, 87, 15, 21, 93, 69, 65, 64, 3, 117, 123, 25, 128, 1, [143, 113, 43, 115, 49, 72, 67,84, 134, 139, 116, 6, 8, 56, 144, 86, 124, 50, 60, 24, 80, 68, 63, 39, 53,138, 55, 142, 118, 96, 74, 18, 52, 133, 17, 23, 37, 103, 110, 101, 98, 9, 83,35, 126]; Attempt to swap the customer sequence [91, … , 53] within this sequence. After the swap... The value is 76.41, which is greater than 0. Therefore, this swap is applied, and the new merge path is [0, 129, 53, 39, 63, 68, 80, 24, 60, 50, 124, 86, 144, 56, 8, 6, 116, 139, 134, 84, 67, 72, 49, 115, 43, 113, 143, 1, 128, 25, 123, 117, 3, 64, 65, 69, 93, 21, 15, 87, 136, 78, 105, 89, 62, 5, 48, 112, 71, 20, 73, 58, 140, 16, 14, 10, 31, 44,22, 97, 90, 108, 46, 130, 92, 27, 57, 107, 66, 104, 141, 147, 137, 132, 88,30, 79, 121, 41, 7, 91, 138, 55, 142, 118, 96, 74, 18, 52, 133, 17, 23, 37,103, 110, 101, 98, 9, 83, 35, 126). After 2-opt optimization, the final merged path is [0, 129,91, 7, 41, 121, 79, 30, 88, 132, 147, 141, 104, 66, 107, 57, 27, 92, 130, 46,108, 90, 97, 22, 109, 76, 19, 11, 54, 82, 33, 45, 28, 131, 111, 106, 100, 38,115, 81, 125, 94, 75, 32, 145, 102, 12, 146, 36, 4, 51, 77, 95, 120, 29, 47,99, 40, 70, 13, 2, 135, 127, 42, 122, 114, 119, 148, 26, 85, 34, 149, 61, 59,44, 31, 10, 14, 16, 140, 58, 73, 20, 71, 9, 83, 126, 35, 98, 101, 110, 103,37, 23, 17, 136, 78, 105, 89, 48, 5, 112, 62).

[0146] Step 4-4: Find the optimal split point on the merged path after 2-opt optimization, such that the maximum service cost of the two split paths is minimized. If the maximum service cost of the split path is less than the maximum service cost of the original two paths, then replace the original two paths with the split path; otherwise, abandon the local search and do not replace them. In this example, the two split paths are [0, 129, 91, 7, 41, 121, 79, 30, 88, 132, 137, 53, 39, 63, 68, 143, 1, 113, 43, 115, 49, 72, 67, 84, 134, 139, 116, 8, 6, 56, 144, 86, 50, 124, 24, 80, 60, 57, 107, 66, 147, 141, 104, 27, 92, 130, 46, 62] and [108, 90, 97, 22, 44, 31, 10, 14, ] respectively. The service costs are 9846.42 and 9851.31 respectively. Compared with the maximum service cost of the two paths before the local search, the maximum service cost of the two new paths is lower. Therefore, these two new paths replace the two old paths.

[0147] Steps 4-5 are repeated from 4-2 to 4-4 until no path planning scheme can be improved after performing local search operations on any two paths. At this point, the path planning scheme has completed local search optimization. In this example, the final planning scheme is {[49, 72, 67, 84, 134, 139, 116, 8, 6, 56, 144, 86, 50, 124, 24, 80, 60, 57, 107, 66, 147, 141, 104, 27, 92, 130, 129, 0, 62, 91, 7, 41, 121, 79, 30, 88, 132, 137, 53, 39, 63, 68, 143, 1, 113, 43, 115], [97, 109, 76, 19, 11, 54, 82, 33, 28,} 131, 111, 106, 29, 120, 100, 38, 81, 125, 94, 75, 32, 145,102, 12, 146, 36, 4, 95, 77, 51, 47, 99, 40, 70, 13, 45, 2, 135, 127, 42,122, 114, 119, 148, 26, 85, 34, 149, 61, 59, 22], [46, 108, 90, 44, 31, 10,14, 16, 140, 58, 126, 73, 20, 71, 9, 83, The path diagram of this scheme is as follows: [35, 98, 101, 110, 103, 37, 23, 17,93, 87, 15, 21, 69, 128, 25, 123, 117, 64, 65, 3, 118, 142, 55, 138, 96, 74,18, 52, 133, 136, 78, 105, 89, 48, 5, 112]. Figure 7 As shown.

[0148] Steps 4-6 are repeated from 4-1 to 4-5 until all planning schemes requiring local search operations have completed local search optimization.

[0149] The aforementioned path merging and splitting strategy that integrates 2-opt can simultaneously optimize customer allocation and service order among customers, effectively improving the quality of multi-logistics vehicle route planning solutions.

[0150] In step 5, the update of the globally optimal planning scheme is divided into the following parts:

[0151] Step 5-1: Compare the fitness values ​​of all the path planning schemes built by the Ant team to find the optimal path planning scheme in this iteration.

[0152] Step 5-2: Compare the fitness value of the iterative optimal planning scheme with the fitness value of the global optimal planning scheme. If the iterative optimal planning scheme is better than the global optimal planning scheme, then replace the global optimal planning scheme.

[0153] In step 5, during the pheromone update phase, the pheromone matrix is ​​updated using a globally optimal planning scheme;

[0154] In this example, fit(R) best The value is 9412. The pheromone concentration of all edges on the three paths and the weight of the warehouses on the three paths are updated sequentially.

[0155] In step 6, each evaluation of the ant team's path planning scheme consumes an evaluation count. The algorithm terminates when the number of evaluation counts is exhausted. The final globally optimal uncertain multi-warehouse, multi-vehicle path planning scheme is output. In this example, the final globally optimal path planning scheme is obtained as gbest = {[138, 96, 74, 18, 52, 133, 105, 83, 9, 89, 48, 5, 62, 112, 71, 20, 73, 10, 14, 16, 140, 58, 126, 35, 98, 101, 110, 103, 37, 23, 17, 78, 136, 87, 15, 93, 21, 69, 3, 64, 65, 128, 25, 123, 117, 118, 142, 55], [53, 1, 43, 49, 115, 81, ...} 125, 94, 12, 75, 32, 145, 102,146, 36, 4, 51, 77, 95, 38, 100, 120, 29, 106, 111, 131, 28, 45, 2, 13, 47,99, 70, 40, 135, 127, 42, 122, 114, 119, 148, 54, 82, 33, 134, 84, 67, 72,113, 143, 63, 39], [19, 11, 26, 85, 34, 149, 61, 59, 76, 109, 22, 97, 90, 44,31, 108, 46, 130, 92, 27, 129, 0, 91, 7, 41, 121, 79, 30, 88, 137, 132, 147,141, 104, 66, 57, 107, 68, 80, 24, 60, 50, 124, 86, 144, 139, 116, 8, 6,56]}. For example Figure 8 The diagram shows the global optimal uncertain multi-warehouse multi-logistics vehicle route planning scheme finally output by an example of the present invention, which includes logistics vehicle route 1, logistics vehicle route 2 and logistics vehicle route 3 constructed by Ant 1, Ant 2 and Ant 3 respectively, with service costs of 9012.05, 8992.47 and 9033.02 for each route respectively.

[0156] The multi-warehouse, multi-vehicle path planning method based on the adaptive neighborhood ant colony system yields feasible path planning schemes that satisfy the constraints. Furthermore, thanks to various mechanisms within the algorithm, the resulting schemes are all of high quality. There is no overlap in service areas between different logistics vehicles, and no path intersections within the same logistics vehicle's path. Simultaneously, the service costs among the logistics vehicles are relatively balanced.

Claims

1. A multi-warehouse, multi-vehicle route planning method based on an adaptive neighborhood ant colony system, characterized in that, Includes the following steps: Step 1: Read data information, including logistics vehicle information, number of customers, customer location information, and service cost information between customers; Construct an uncertain multi-warehouse, multi-vehicle route planning model and establish an optimization objective function to evaluate the quality of the route planning scheme; Step 2: Based on customer location information and service cost information between customers, construct a service cost matrix between customers and initialize a heuristic information matrix; initialize the ant colony system algorithm parameters, and use a greedy algorithm to construct an uncertain multi-warehouse and multi-logistics vehicle route planning scheme, and initialize the pheromone matrix and warehouse weight vector; Step 2 specifically involves the ant colony system algorithm, which includes the number of ant colonies NP, pheromone factor α, heuristic pheromone factor β, local pheromone evaporation rate ξ, global pheromone evaporation rate ρ, and pseudo-random probability q0. Heuristic information is generated based on the service costs between customers. The heuristic information η between two customers i and j is... i,j The calculation formula is: ； e i,j This represents the service cost between customer i and customer j. Constructing a greedy multi-vehicle route planning scheme using a greedy algorithm includes the following steps: Step 2-1: Initialize the ant colony, which consists of M ants; Step 2-2: Each ant in the ant team randomly selects a customer from the customer set as the starting point, selects the next customer based on the heuristic information between customers, and begins to greedily construct a logistics vehicle route planning scheme; the ant team initializes a customer tabu table tabu, which contains all the selected initial customers, and initializes the path service cost of each ant to 0; Steps 2-3: Each ant in the ant team obtains its own set of accessible customers (candidates) based on the customer taboo table. Then, each ant pre-selects customer j from its own set of accessible customers (candidates) that has the most heuristic information about its current customer i. Then, each ant adds the pre-selected customer to its own constructed service path and calculates the estimated path service cost for each ant. Steps 2-4: Select the ant with the lowest estimated path service cost to build the path; add the customer pre-selected by this ant to the constructed logistics vehicle path, add it to the taboo list, and update the corresponding path service cost; other ants should abandon their pre-selected customers. Steps 2-5: Repeat steps 2-3 to 2-4 until each ant in the ant team can no longer select customers, that is, all customers are served by logistics vehicles, thus obtaining a greedy uncertain multi-warehouse multi-logistic vehicle route planning scheme. Steps 2-6: Use the fitness value of the greedy uncertain multi-warehouse multi-logistics vehicle route planning scheme to assign the initial value of pheromone τ0, and then initialize the pheromone in all pheromone matrices to τ0; Step 3: Construct an ant team, where each ant team contains the same number of ants as the logistics vehicle. Each ant is responsible for constructing a service path for one logistics vehicle. Each ant in each team adaptively selects a warehouse as the starting point of the path based on the warehouse weight vector. Subsequently, each ant constructs a service path for the logistics vehicle for each customer one by one, based on the proposed adaptive neighborhood method and the ant selection technique based on the estimated path service cost, combined with the heuristic information matrix and pheromone matrix. At the same time, the pheromone matrix is ​​locally updated during the path construction process. Step 3 specifically includes the following steps: Step 3-1: Construct NP ant teams, each team containing M ants; Step 3-2: Use an adaptive warehouse selection strategy to determine the starting customers of all ants in the ant team; then start building a logistics vehicle route planning scheme; initialize a customer tabu table tabu, which contains the starting customers of all routes, and initialize the route service cost of each ant to 0; Step 3-3: Each ant in the ant team pre-selects the next customer based on an adaptive neighborhood-based customer selection mechanism; Steps 3-4: Use ant selection technology based on estimated path service cost to select the logistics vehicle path for ants; add the pre-selected customers of the selected ants to the logistics vehicle path they build, and add them to the customer taboo list of the ant team, and update the corresponding path service cost; at the same time, call the local pheromone update rule to update the pheromone on the edge that the ant is currently passing through in real time, and other ants abandon their pre-selected customers. Steps 3-5: Repeat steps 3-3 and 3-4 until all customers are selected and enter the tabu, thus obtaining an uncertain multi-warehouse and multi-vehicle route planning scheme; Steps 3-6: Repeat steps 3-2 to 3-5 until all Ant teams have built uncertain multi-warehouse and multi-vehicle route planning schemes; Step 4: Based on the proposed path merging and segmentation strategy of fusion 2-opt, adaptively perform local search operation on the logistics vehicle path planning scheme constructed by the Ant team according to probability; Step 5: Set optimization objectives, evaluate each path planning scheme, and update the globally optimal planning scheme; use the globally optimal planning scheme to globally update the pheromone matrix; then, update the warehouse weight vector according to each path planning scheme. Step 6: Repeat steps 3 to 5 until the iteration termination condition is met, and finally output the globally optimal logistics vehicle route planning scheme.

2. The multi-warehouse, multi-vehicle route planning method based on an adaptive neighborhood ant colony system according to claim 1, characterized in that, Step 1 specifically involves: the data information being read includes the number of customers N, and the customer coordinates (x, y, y). i , y i ), where x i , y i Let x and y represent the x and y coordinates of the i-th customer, respectively; the service cost e between customers. ij Number of logistics vehicles: M; The specific steps for constructing an uncertain multi-warehouse, multi-vehicle route planning model are as follows: Given N customers and M logistics vehicles, the vehicle service map is modeled as a completely undirected graph G = (V, E), where V = {(x i , y i Let { ) | 1≤i≤N} represent the customer set; E = { e ij | i, j∈V} is the set of edges between all customers; each edge is assigned a weight, which represents the service cost from customer i to customer j; Define the optimization goal: ； ； ； Where R represents the planning scheme, E k R represents the total service cost of the k-th path. k This represents the k-th path in the planning scheme. This represents the service cost between the i-th customer on the k-th path and the (i+1)-th customer on the k-th path. Represents the first customer and the k-th path. The service cost between n customers k Let n represent the number of customers along the k-th path. i This represents the number of customers along the i-th path.

3. The multi-warehouse, multi-vehicle path planning method based on an adaptive neighborhood ant colony system according to claim 1, characterized in that, In steps 2-3, the accessible customer set (candidates) is obtained using the following formula: ； Where V represents the set of all customers, and tabu represents the tabu list; In steps 2-3, the estimated service cost of Ant Group is calculated using the following formula: ； Among them, future_expense i Represents the i-th path R i The estimated service cost, n i The representative represents the number of customers along the i-th path, pre i Represents path R i Corresponding to the customers pre-selected by Ant Group, On behalf of clients With the client i Service costs between On behalf of clients With the client i Service costs between, E i Indicates the cost of path services; In steps 2-4, the ants that construct the logistics vehicle route are selected according to the following formula: the ant with the lowest estimated service cost is selected for route construction. ； Where M represents the number of logistics vehicles; after the selected ant places the pre-selected customer into its path, the corresponding path service cost E is updated using the following formula. i Number of customers along the path n i : ； ； In steps 2-6, the initial pheromone value τ0 is assigned using the maximum path service cost in the greedy path planning scheme. The calculation formula is as follows: ； Where N represents the number of customers, E greedy This represents the maximum path service cost in the solution constructed by the greedy algorithm.

4. The multi-warehouse, multi-vehicle path planning method based on an adaptive neighborhood ant colony system according to claim 1, characterized in that, Step 3-2 specifically includes the following steps: Step 3-2-1: Initialize the accessible customer set `candidates`, which includes all customers; initialize each customer as the initial weight `w` for the repository. i ; Step 3-2-2: Based on the pheromone concentration, calculate the probability that all customers in the set are selected. The probability is calculated as follows: ； Among them, w i ,w j Let i and j represent the weights of customers i and j in the warehouse weight vector, respectively, and α represent the pheromone factor. A roulette wheel selection mechanism is used to select a customer as a warehouse, which is then removed from the accessible customer set (candidates), and a local weight update is performed. The method for the local weight update is as follows: ； in This is the initial value of pheromones. This refers to the local evaporation rate. Step 3-2-3: Repeat step 3-2-2 until M warehouses are found. Assign each warehouse to one of the M ants in the ant team as its warehouse.

5. The multi-warehouse, multi-vehicle route planning method based on an adaptive neighborhood ant colony system according to claim 1, characterized in that, In step 3-3, the customer selection mechanism is as follows: taking the last customer i of the current path k as the current customer, the probability of selecting customer j is calculated according to the following formula: ； in, For the pheromone concentration from customer i to customer j, η i,j For heuristic information from customer i to customer j, For the pheromone concentration from customer i to customer l, η i,l For the heuristic information from customer i to customer l, α and β are the pheromone factor and heuristic information factor, respectively. `candidates` represents the accessible customer set, constructed based on a tabu list and an adaptive domain. After obtaining the probability of each customer in the accessible customer set, a roulette wheel selection mechanism is used to select a customer. Based on the probability of each customer in the candidate customer set, the next accessible customer is selected using the following formula: ； Where j* represents the customer selected using the roulette wheel strategy according to the calculated customer selection probability, rand(0,1) represents a random number between 0 and 1, q0 represents a probability parameter indicating that the ant will greedily select the customer with the strongest pheromone concentration and heuristic information with probability q0, and p i,l Let i be the probability of selecting customer l as the next node with i as the current node, and let l be the probability of selecting a customer according to the roulette wheel selection mechanism with a probability of (1-q0).

6. The multi-warehouse, multi-vehicle path planning method based on an adaptive neighborhood ant colony system according to claim 1, characterized in that, Step 4 specifically includes the following steps: Step 4-1: Based on the fitness values ​​of all the solution schemes constructed by the ant teams, calculate the probability of each solution scheme performing a local search, as shown in the following formula: ； Where, fit(R) k Let be the fitness value of the k-th solution, and max(fit) and min(fit) be the maximum and minimum fitness values ​​among all ant teams. For each solution, a random number is generated and compared with the corresponding probability to determine whether a local search operation will be performed. Step 4-2: For each solution that requires a local search operation, list two different logistics vehicle service paths within the solution. Step 4-3: Randomly select a breakpoint for each of the two paths, connect the two paths together at the breakpoint, and perform 2-opt optimization on the merged path; 2-opt optimization specifically includes the following steps: Step 4-3-1: List any two customers in the merging path, swap these two customers and the customer sequence between them, and calculate the resulting reduction in service costs. If the reduction is positive, then apply this swap. Let t be the customer preceding customer i in the merging path, and l be the customer following customer j. The formula for calculating the total path cost reduction after swapping customer i and customer j and the customer sequence between them is: ； Among them, Δreduce_expense i,j e represents the reduction in service costs resulting from swapping customer i, j and the customer sequence between customers i, j. t,i e represents the service cost between customers t and i. j,l Represents the service cost between customers j and l, e t,j E represents the service cost between customers t and j. i,l This represents the service cost between customers i and l; Δreduce_expense i,j If the result is positive, then this swap will be applied, and the total service cost of the merged path will be updated. The calculation formula is as follows: ； Among them, E merge This represents the total service cost of the merged path; Step 4-3-2: Repeat step 4-3-1 until swapping any two customers in the merge path and the customer sequence between the two customers will not reduce the service cost. Step 4-4: Find the optimal split point on the optimized path so that the maximum service cost of the two split paths is minimized. If the maximum service cost of the split path is less than the maximum service cost of the original two paths, the local search is successful. Therefore, the two paths after the local search are used to replace the original two paths. Otherwise, the local search is unsuccessful, the optimization is abandoned, and no replacement is made. Steps 4-5: Repeat steps 4-2 to 4-4 until any two paths in the plan can no longer be optimized. At this point, the current planning plan has been optimized. Steps 4-6: Repeat steps 4-1 to 4-5 until all scheduling schemes requiring local search have completed the designed local search strategy.

7. The multi-warehouse, multi-vehicle path planning method based on an adaptive neighborhood ant colony system according to claim 1, characterized in that, Step 5 specifically includes the following steps: Step 5-1: Compare the fitness values ​​of all multi-logistics vehicle route planning schemes of the Ant Team and find the optimal solution in the iteration. Step 5-2: Compare the fitness value of the found iterative optimal solution with the fitness value of the global optimal solution; if the iterative optimal solution is better, then replace the global optimal solution with the iterative optimal solution. During the pheromone update phase, a globally optimal solution is used to update the pheromone concentration, and the pheromone concentration update formula is as follows: ； Where ρ is the global pheromone evaporation coefficient, and R best This is the globally optimal solution. The pheromone concentration from customer i to customer j; In step 5, the warehouse weight vector update formula is: ； w i This indicates that customer i is the weight of the warehouse in the warehouse weight vector.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method of claim 1.

Images