A logistics path planning method based on an improved hybrid ant colony algorithm

Abstract

The application belongs to the technical field of logistics and artificial intelligence, and particularly relates to a logistics path planning method based on a hybrid ant colony algorithm, which comprises the following steps: establishing a logistics distribution task scheduling network model; comprehensively considering three objectives of minimizing cost, minimizing customer disappointment degree and minimizing driver load imbalance to establish an objective function; designing an improved hybrid ant colony algorithm according to the path planning of logistics distribution; solving the objective function by using the improved hybrid ant colony algorithm, and obtaining better non-dominated solutions by using a fast non-dominated sorting algorithm, and obtaining an optimal solution of the logistics distribution path planning from the non-dominated solution set. The application integrates a multi-objective optimization algorithm, comprehensively considers three objectives of minimum cost, minimum disappointment degree and minimum load imbalance, and is more in line with a real scene and has better convergence and ideal diversity compared with a traditional single-objective or double-objective path planning.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of logistics path planning and artificial intelligence, and particularly relates to a logistics path planning method based on an improved hybrid ant colony algorithm. BACKGROUND

[0002] With the development of the times, the production capacity of various industries is increasing, and the transportation and deployment of products and raw materials are in high demand. In addition, the rise of express delivery and take-out has made the logistics industry play an increasingly important role in China's market economy. In the face of rapid growth in logistics demand, an effective method to optimize logistics transportation costs can enable the long-term and stable development of the modern logistics industry. With the gradual improvement of China's basic network infrastructure, traditional railway transportation, port transportation, airport transportation, and the four-way highway and village-to-village highway have been greatly developed, and the logistics distribution environment has been greatly optimized. The state has vigorously promoted the modern logistics industry. In the industrial scene, the transportation and deployment of production materials in the industrial park affect the industrial production efficiency. Planning a multi-objective optimal route solution for the vehicle can effectively solve this problem, such as establishing a suitable scheduling model and calculating the optimal path with a hybrid ant colony algorithm, which is of great significance to the industrial park logistics transportation.

[0003] In logistics transportation, vehicle path planning is a key problem in logistics resource scheduling. The vehicle routing problem (VRP) is a well-known classic combinatorial optimization problem in transportation logistics and supply chain management. It involves providing efficient and economical goods / services from a central distribution / collection center to a limited number of geographically dispersed customers. Traditional path planning is generally single-objective or double-objective path planning, and in real logistics scenarios, there are usually multiple optimization objectives. The existing path planning algorithm that integrates NSGA-Ⅱ has limitations in exploring three-dimensional (three objectives) and higher latitude (four objectives and more objectives) problems, and the convergence and diversity are worse than NSGA-III. SUMMARY

[0004] To solve the above technical problems, the application provides a logistics path planning method based on an improved hybrid ant colony algorithm, which comprises:

[0005] S1: generating a logistics distribution task scheduling network model according to the route relationship between the distribution points and the distribution points, the route relationship between the distribution points and the distribution center, and the constraint information of the distribution vehicles of each distribution point and distribution route;

[0006] S2: comprehensively considering the three objectives of minimizing cost, minimizing customer dissatisfaction, and minimizing driver load imbalance, establishing an objective function of the logistics distribution task scheduling network model;

[0007] S3: designing an improved hybrid ant colony algorithm according to the path planning of logistics distribution;

[0008] S4: An improved hybrid ant colony algorithm is used to solve the objective function, and a NSGA-III fast non-dominated sorting algorithm is used to obtain better non-dominated solutions, and a non-dominated solution set is obtained, and an optimal solution of the logistics distribution path planning is obtained from the non-dominated solution set.

[0009] Preferably, considering the three objectives of minimizing cost, minimizing customer dissatisfaction and minimizing driver load imbalance, a target function of the logistics distribution task scheduling network model is established, including:

[0010] The cost function of minimizing cost is:

[0011]

[0012] Wherein, f1 represents the objective function of minimizing total cost, represents the transportation cost generated during distribution and the penalty cost generated by failing to complete the distribution task within the time window; c represents the distribution cost generated per unit distance, d ij represents the distance between distribution points i and j, represents whether the logistics vehicle k serves the distribution point j after serving the distribution point i, p i represents the penalty cost generated by failing to serve within the time window, and n represents the number of distribution points;

[0013] The cost function of minimizing customer dissatisfaction is:

[0014]

[0015] Wherein, f2 represents the objective function of minimizing the average dissatisfaction of customers, down i represents the dissatisfaction of customers, [ET i ,LT i ] represents the time window range of node i, represents the optimal service time window of distribution point i, t i represents the service time provided for the customer, if the service time t i provided for the customer i is within the optimal service time window, the dissatisfaction of the customer is 0, the farther the service time provided for the customer i is from the optimal service time window, the higher the dissatisfaction of the distribution point i is, and the dissatisfaction of the distribution point i is 1 when the service time provided for the customer i is outside the service time window [ET i ,LT i ], and a penalty cost is generated, and n represents the number of distribution points;

[0016] The cost function of minimizing driver load imbalance is:

[0017]

[0018] wherein W k denotes the delivery distance of the delivery vehicle k, d ij denotes the distance between the delivery points i and j, denotes whether the vehicle k serves the delivery point j after serving the delivery point i, the distance between the vehicles is unbalanced, and K denotes the number of delivery vehicles.

[0019] Preferably, the improved hybrid ant colony algorithm is designed according to the path planning of logistics distribution, comprising:

[0020] In the initial stage of the algorithm, a saving matrix updating probability formula is introduced to guide the ants to search for paths; a piecewise function is used to improve the pheromone evaporation factor to obtain a piecewise pheromone evaporation factor, thereby optimizing the convergence speed of the algorithm; the elite ant colony uses an ant cycle model to update the pheromone, and the ordinary ant uses an ant quantity model to update the pheromone; and a 3-opt algorithm is used to improve the local search ability of the hybrid ant colony algorithm.

[0021] Further, the saving matrix updating probability formula is introduced to guide the ants to search for paths, comprising:

[0022]

[0023] wherein, denotes the probability of the ant k serving the customers of the delivery point j after serving the customers of the delivery point i, denotes the set of customers that the ant k is allowed to serve at the time t (i.e., customers that have not been served); ξ and respectively denote the first and second weight coefficients, and satisfy 0≤ξ, and α denotes an information heuristic factor, reflecting the relative importance of the residual information in the exploration process of the ant colony, β denotes an expected heuristic factor (visibility coefficient), reflecting the relative importance of the path expectation in the exploration process of the ant colony; γ denotes a saving matrix heuristic factor, reflecting the relative importance of the saving matrix in the exploration process of the ant colony; η ij denotes the visibility from the delivery point i to the delivery point j, τ ij denotes the pheromone concentration from the delivery point i to the delivery point j, ET j denotes the earliest service time allowed for the node j, LT j denotes the latest service time allowed for the node LT j , U ij denotes the saving matrix, t j denotes the service time of the delivery vehicle for the customers of the delivery point j.

[0024] Further, a piecewise function is used to improve the pheromone evaporation factor, comprising:

[0025]

[0026] wherein, p represents pheromone evaporation factor, NC represents current iteration number, NC max represents the maximum iteration number set initially.

[0027] Further, the elite ant colony adopts the ant cycle model to update the pheromone, and the common ant adopts the ant quantity model to update the pheromone, including:

[0028] The elite ant colony adopts the ant cycle model to update the pheromone:

[0029]

[0030] wherein, τ ij represents the pheromone concentration on the path from the distribution point i to the distribution point j, p represents the pheromone evaporation factor, represents the additional pheromone content of the elite ant compared with the common ant, represents the additional pheromone content of the elite ant k after passing through the path i→j, ω represents the pheromone increment coefficient of the elite ant, and t represents time;

[0031] The common ant adopts the ant quantity model to update the pheromone:

[0032]

[0033] wherein, τ ij represents the pheromone concentration on the path from the distribution point i to the distribution point j, p represents the pheromone evaporation factor, Δτ ij represents the total sum of the additional pheromone content of all common ants after passing through the path i→j, represents the additional pheromone content of the elite ant k after passing through the path i→j.

[0034] Further, the 3-opt algorithm is used to improve the local search ability of the hybrid ant colony algorithm, including:

[0035] The 3-opt algorithm is used to delete three pairs of non-adjacent distribution point connections in the path, and then the path lengths of the other seven different connection modes are calculated. After comparison, the shortest path is selected as the new connection mode. The 3-opt has stronger local search ability than the 2-opt, but since the 3-opt produces more new connection modes than the 2-opt, it leads to more time-consuming. Therefore, the condition for using the local search algorithm in the algorithm is modified. When the ant colony is iterated, if the optimal path of the new generation is the same as that of the last generation, the local search algorithm is not used, and if it is different, the 3-opt local optimization algorithm is used to improve the local search ability of the algorithm.

[0036] Preferably, S4 specifically comprises:

[0037] Step 1: initialize parameters: input the number of ant colony k, pheromone heuristic factor a, visibility heuristic factor b, saving matrix heuristic factor g, two weight coefficients and The number of iterations of the ant colony algorithm NC, the population size n of the NSGA-III algorithm;

[0038] Step 2: put K ants into the distribution center;

[0039] Step 3: take the distribution points in the logistics distribution task scheduling network model as nodes, and use the roulette method to select the next node from the node set, and store the nodes passed into the taboo table;

[0040] Step 4: judge whether the ant has completed the traversal of all nodes, if not, go to step 3, otherwise go to step 5;

[0041] Step 5: judge whether all ants in this iteration have completed the traversal, if not, add 1 to the number of ants and return to step 3, otherwise jump to step 6;

[0042] Step 6: the ants that have completed the traversal calculate the path of each ant according to the objective function of the logistics distribution task scheduling network model, obtain the optimal path under the three objectives, use the 3-opt local optimization algorithm for the optimal path of each objective, and analyze the feasibility of the 7 new paths generated by the 3-opt local optimization algorithm, if feasible, recalculate the path according to the objective function of the logistics distribution task scheduling network model through the new path, if better, replace the original path, otherwise use the original path;

[0043] Step 7: update the pheromone according to the segmented pheromone evaporation factor, pheromone update method and iteration number of the improved hybrid ant colony algorithm;

[0044] Step 8: use the NSGA-III algorithm to perform non-dominated sorting on the population according to the updated pheromone, and obtain a non-dominated solution set in the population;

[0045] Step 9: judge whether the maximum number of iterations set at the initialization has been reached, if not, return to step 2, otherwise jump to step 10;

[0046] Step 10: select the optimal solution that optimizes the three objectives simultaneously from the non-dominated solution set, and obtain the specific transportation tool distribution route.

[0047] Further, the 3-opt local optimization algorithm is used for the optimal path of each objective, and the feasibility of the 7 new paths generated by the 3-opt local optimization algorithm is analyzed, including:

[0048] Delete three pairs of non-adjacent distribution point connections in the path, arrange and combine the remaining paths to obtain 7 different connection mode paths, and select the shortest path as the new connection mode after comparing the original path with the 7 new paths.

[0049] Further, the NSGA-III algorithm is used for non-dominated sorting of the population, including:

[0050] The population with a population size of n' is layered by using the non-dominated sorting algorithm, and the solutions in the solution set of the new population after iteration are non-dominated sorted, the first layer is denoted as f 1 , the non-dominated sequence value i rank of which is 1, indicating that the solution is not dominated by all other solutions, and is the non-dominated solution set of the entire population; then f 2 is deleted, and the remaining population is non-dominated sorted to obtain the second layer of non-dominated solutions f rank with the sequence value i 2 being 2, and f 2 is the non-dominated solution set of the next layer, and so on, all solutions are layered, and the solutions in the first layer are selected as the non-dominated solution set of the whole.

[0051] The beneficial effects of the present application are:

[0052] The present application is directed to the scheduling management technology of logistics distribution vehicles between distribution points and between distribution points and distribution centers in the logistics industry, and proposes a logistics path planning method based on a hybrid ant colony algorithm, which provides an optimal path planning solution for a three-objective path planning model requiring the lowest cost, the smallest disappointment degree and the smallest load imbalance, has good practical significance and application value, and compared with the traditional single-objective and double-objective path planning, is closer to the actual scene demand. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 is a flow chart of the logistics path planning method based on the hybrid ant colony algorithm of the present application;

[0054] Figure 2 is a pheromone updating flow chart in the embodiment of the present application;

[0055] Figure 3 The hybrid ant colony algorithm flow chart in the embodiment of the present application. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0057] A logistics path planning method based on an improved hybrid ant colony algorithm, as shown in the formula (1), comprising the steps of: Figure 1

[0058] S1: generating a logistics distribution task scheduling network model according to the route relationship between distribution points and distribution points, the route relationship between distribution points and distribution centers, and the constraint information of each distribution point, distribution route and distribution vehicle;

[0059] S2: comprehensively considering the three objectives of minimizing cost, minimizing customer dissatisfaction and minimizing driver load imbalance, and establishing an objective function of the logistics distribution task scheduling network model;

[0060] S3: designing an improved hybrid ant colony algorithm according to the path planning of logistics distribution;

[0061] S4: solving the objective function by using the improved hybrid ant colony algorithm, obtaining better non-dominated solutions by using the NSGA-III fast non-dominated sorting algorithm, obtaining a non-dominated solution set, and obtaining an optimal solution of the logistics distribution path planning from the non-dominated solution set.

[0062] According to the path intercommunication and length information between distribution points and distribution points and between distribution points and distribution centers, the distribution center and the distribution point are abstracted as a vertex, the path between the vertices is abstracted as an edge, and the entire road network relationship is abstracted as an undirected graph G(V,K). In the vehicle path problem with time window, K represents the vehicle set owned by the distribution center; N represents the set of n distribution points that need logistics distribution service, N={1,2,···,n}; V={0}∪N; wherein 0 is the distribution center; V k is the set of logistics vehicles K visiting the distribution points; i,j represent the number of distribution points; represents whether the logistics vehicle k serves the distribution point j after serving the distribution point i; represents whether the distribution point i is served by the logistics vehicle k; Q max represents the loading capacity constraint of the logistics vehicle; q i represents the demand of the distribution point i, and q0=0; C is the unit transportation cost of the logistics transportation vehicle; the soft time window of the distribution point is [ET i ,LT i ], wherein ET i represents the earliest service time allowed by the logistics distribution point i, LT i represents the latest service time allowed by the logistics distribution point i, and the constraint condition of the mathematical model of VRPTW is:

[0063] represents that the total demand transported by each logistics transportation vehicle k does not exceed the maximum loading limit of the vehicle;

[0064] ​ This indicates that logistics delivery vehicles need to return to the distribution center after leaving it;

[0065] This means that each logistics delivery point can only be served by one vehicle.

[0066] Taking into account the three objectives of minimizing cost, minimizing customer disappointment, and minimizing driver load imbalance, the objective function of the logistics delivery task scheduling network model is established, including:

[0067] Minimize cost function:

[0068]

[0069] Where f1 represents the objective function to minimize the total cost, represents the transportation cost incurred during delivery and the penalty cost for failing to complete the delivery task within the time window; c represents the delivery cost per unit distance, and d represents the cost per unit distance. ij This represents the distance between delivery points i and j. Indicates whether logistics vehicle k serves delivery point i and then serves delivery point j, p i This represents the penalty cost incurred for failing to provide service within the time window, where n represents the number of delivery points;

[0070] The penalty cost of failing to provide service within the time window:

[0071]

[0072] Where, p i This represents the penalty cost incurred for failing to provide service within the time window. No penalty cost is incurred if the delivery task is completed within the time window. The time window range for node i is known to be [ET]. i ,LT i ], where ET i LT represents the earliest allowed service time for node i. i p' represents the latest allowed service time for node i, p' represents the penalty coefficient for a delivery time earlier than the earliest service time, and p″″ represents the penalty coefficient for a delivery time later than the latest service time.

[0073] Minimize the cost function of customer disappointment:

[0074]

[0075] Where f2 represents the objective function that minimizes the average customer disappointment level, down i Indicates the level of customer disappointment. [ET i ,LT i The symbol ] represents the time window range of node i. denotes the optimal service time window of the delivery point i, t i denotes the service time provided for the customer, if the service time t i is provided for the customer i, the dissatisfaction degree of the customer i is 0, the farther the service time provided for the customer i is from the optimal service time window, the higher the dissatisfaction degree of the delivery point i is, the service time window [ET i , LT i ] outside the optimal service time window, the dissatisfaction degree is 1, and a penalty cost is generated, and n denotes the number of delivery points;

[0076] minimize the driver load imbalance cost function:

[0077]

[0078] wherein, W k denotes the delivery distance of the delivery vehicle k, d ij denotes the distance between the delivery points i and j, denotes whether the vehicle k serves the delivery point j after serving the delivery point i, the distance between the vehicle delivery is not balanced, and k denotes the number of delivery vehicles.

[0079] The improved hybrid ant colony algorithm is designed according to the path planning of logistics distribution, including:

[0080] In the initial stage of the algorithm, the saving matrix updating probability formula is introduced to guide the ants to search the path; the piecewise function is used to improve the pheromone evaporation factor, so as to obtain the piecewise pheromone evaporation factor, and the convergence speed of the algorithm is optimized; the elite ant colony uses the ant cycle model to update the pheromone, and the ordinary ant uses the ant quantity model to update the pheromone; the 3-opt algorithm is used to improve the local search ability of the hybrid ant colony algorithm.

[0081] The saving matrix updating probability formula is introduced to guide the ants to search the path, including:

[0082]

[0083] wherein, denotes the probability of the ant k serving the customers of the delivery point j after serving the customers of the delivery point i, denotes the customer set allowed to be served by the ant k at the t time (i.e. the customers not yet served); ξ and respectively denote the first and second weight coefficients, and satisfy 0≤ξ, and α represents an information heuristic factor, reflecting the relative importance of the residual information in the exploration process of the ant colony, β represents an expected heuristic factor (visibility coefficient), reflecting the relative importance of the path expectation in the exploration process of the ant colony; γ represents a saving matrix heuristic factor, reflecting the relative importance of the saving matrix in the exploration process of the ant colony; η ij represents the visibility from the distribution point i to the distribution point j, τ ij represents the pheromone concentration of the distribution point i to the distribution point j, ET j , LT j represents the start point of the time window, U ij represents the saving matrix, t j represents the service time of the distribution vehicle for the customer of the distribution point j.

[0084] In the initial stage, the global search ability of the algorithm is improved by setting a larger evaporation factor p, and in the later stage of the algorithm, a smaller evaporation factor p is set to achieve the purpose of rapid convergence. A segmented function is used to improve the pheromone evaporation factor, including:

[0085]

[0086] wherein p represents the pheromone evaporation factor, NC represents the current iteration number, NC max represents the maximum iteration number set initially.

[0087] As Figure 2 shown: when updating the pheromone, the iteration number is calculated to select the appropriate pheromone evaporation factor, and the pheromone on the path is updated by bringing in the pheromone update formula.

[0088] The elite ant colony adopts the ant cycle model to update the pheromone, and the ordinary ant adopts the ant quantity model to update the pheromone, including:

[0089] The elite ant colony adopts the ant cycle model to update the pheromone:

[0090]

[0091] wherein τ ij represents the pheromone concentration on the path from the distribution point i to the distribution point j, p represents the pheromone evaporation factor, represents the additional pheromone content of the elite ant compared with the ordinary ant, represents the additional pheromone content of the elite ant k after passing through the path i→j, ω represents the pheromone increment coefficient of the elite ant, and t represents the time;

[0092] The ordinary ant adopts the ant quantity model to update the pheromone:

[0093]

[0094] wherein τ ij denotes the pheromone concentration on the path from distribution point i to distribution point j, p denotes the pheromone evaporation factor, Δτ ij denotes the total amount of newly added pheromone content after all ordinary ants pass through the path i→j, denotes the amount of newly added pheromone content after elite ant k passes through the path i→j.

[0095] The 3-opt algorithm is used to improve the local search ability of the hybrid ant colony algorithm, including:

[0096] The 3-opt algorithm is used to delete three pairs of non-adjacent distribution point connections in the path, and then calculate the path lengths of the other seven different connection modes. After comparison, the shortest path is selected as the new connection mode. The 3-opt algorithm has stronger local search ability than the 2-opt algorithm, but the new connection mode generated by the 3-opt algorithm is more than that of the 2-opt algorithm, resulting in more time consumption of the 3-opt algorithm. Therefore, the condition for using the local search algorithm in the algorithm is modified. If the optimal path of the new generation is the same as that of the last generation, the local search algorithm is not used. If they are different, the 3-opt local optimization algorithm is used to improve the local search ability of the algorithm.

[0097] The S4, as shown in Figure 3 specifically comprises:

[0098] Step 1: initialize parameters: input the number of ant colonies k, the pheromone heuristic factor a, the visibility heuristic factor b, the saving matrix heuristic factor g, and the two weight coefficients ξ and the number of iterations of the ant colony algorithm NC, and the population size n of the NSGA-III algorithm;

[0099] Step 2: place K ants in the distribution center;

[0100] Step 3: take the distribution points in the logistics distribution task scheduling network model as nodes, and use the roulette method to select the next node from the node set. The nodes passed through are stored in the taboo table.

[0101] Step 4: determine whether the ants have completed the traversal of all nodes. If not, go to step 3, otherwise go to step 5.

[0102] Step 5: determine whether all ants have completed the traversal in this iteration. If not, add 1 to the number of ants and return to step 3, otherwise go to step 6.

[0103] Step 6: The ants that complete the traversal calculate the path of each ant according to the objective function of the logistics distribution task scheduling network model, obtain the optimal path under the three objectives, use the 3-opt local optimization algorithm for the optimal path of each objective, and analyze the feasibility of the seven new paths generated by the 3-opt local optimization algorithm. If feasible, the path is recalculated through the new path according to the objective function of the logistics distribution task scheduling network model. If better, the original path is replaced. Otherwise, the original path is used.

[0104] Step 7: Update the pheromone according to the segmented pheromone evaporation factor, pheromone update method, and iteration number of the improved hybrid ant colony algorithm.

[0105] Step 8: Sort the population using the NSGA-III algorithm according to the updated pheromone to obtain the non-dominated solution set in the population.

[0106] Step 9: Determine whether the maximum number of iterations set at initialization has been reached. If not, return to step 2. Otherwise, go to step 10.

[0107] Step 10: Select the optimal solution that optimizes the three objectives simultaneously from the non-dominated solution set to obtain the specific transportation tool distribution route.

[0108] Further, the 3-opt local optimization algorithm is used for the optimal path of each objective, and the feasibility of the seven new paths generated by the 3-opt local optimization algorithm is analyzed, including:

[0109] Delete three pairs of non-adjacent distribution points in the path, and arrange and combine the remaining paths to obtain seven different connection mode paths. After comparing the original path with the seven new paths, select the shortest path as the new connection mode.

[0110] Sort the population using the NSGA-III algorithm, including:

[0111] Sort the population of size n' by layer through the non-dominated sorting algorithm. Sort the solutions in the solution set of the new population after iteration. The first layer is denoted as f 1 , whose non-dominated sequence value i rank is 1, indicating that this layer solution is not dominated by all other solutions, and is the non-dominated solution set of the entire population. Then, f 2 is deleted, and the remaining population is sorted by non-dominated sequence to obtain the second layer non-dominated solution f rank with sequence value i 2 of 2, and f 2 is the non-dominated solution set of the next layer. In this way, all solutions are layered, and the solutions in the first layer are selected as the non-dominated solution set of the whole.

[0112] The non-dominated sorting algorithm processing process includes:

[0113] To maintain the population size at the initial set size n, it is necessary to adjust the population size according to i. rank The set of individuals sorted from low to high after non-dominance is added to the new population. When 1 rank +2 rank +…+n-1 rank <n&&1 rank +2 rank +…+n rank When >n, where the sequence value i rank This represents the non-dominated solution at level i, where n is the population size, and we need to select from the sequence value n. rank Obtain from the set of individuals Individuals join the new population, among which... The number of non-dominated solutions in the i-th layer is represented by: calculating the minimum value of the objective function in each dimension to obtain the ideal point; then, using the ideal point as the origin and the objective function as the coordinate axis, calculating the extreme points in each dimension; constructing a hyperplane using the extreme points; then, uniformly constructing reference points on the hyperplane according to the number of partitions for each objective; connecting the ideal point and the reference points to construct a reference point vector; calculating the perpendicular distance between each individual and the reference vector; associating each individual with the reference point with the nearest corresponding perpendicular distance; and finally, starting from the sequence value n, calculating the number of individuals associated with each reference point. rank Select from the set of individuals Individuals join the population.

[0114] Calculate the extreme points for each dimension:

[0115]

[0116] in, f′ represents the minimum value of the ideal point in the target direction, x represents an individual in the population. i (x) represents the actual value f of target i. i (x) minus the ideal value Weight vector w i The size of target i is 1, and the size of other targets is 0.000001. M represents the number of targets, and w represents the weight vector.

[0117] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A logistics route planning method based on an improved hybrid ant colony algorithm, characterized in that, include: S1: Generate a logistics delivery task scheduling network model based on the route relationships between delivery points and delivery centers, the route relationships between delivery points and delivery centers, and the constraint information of delivery vehicles at each delivery point and along delivery routes. S2: Considering the three objectives of minimizing cost, minimizing customer disappointment, and minimizing driver load imbalance, establish the objective function of the logistics delivery task scheduling network model; Taking into account the three objectives of minimizing cost, minimizing customer disappointment, and minimizing driver load imbalance, the objective function of the logistics delivery task scheduling network model is established, including: Minimize cost function: Where f1 represents the objective function to minimize the total cost, represents the transportation cost incurred during delivery and the penalty cost for failing to complete the delivery task within the time window; c represents the delivery cost per unit distance, and d represents the cost per unit distance. ij This represents the distance between delivery points i and j. Indicates whether logistics vehicle k serves delivery point i and then serves delivery point j, p i This represents the penalty cost incurred for failing to provide service within the time window, where n represents the number of delivery points; Minimize the cost function of customer disappointment: Where f2 represents the objective function that minimizes the average customer disappointment level, down i Indicates the level of customer disappointment. [ET i ,LT i The symbol ] represents the time window range of node i. This indicates the optimal service time window for delivery point i. t i This indicates the service time provided to the customer. If the service time provided to the customer is t... i Within the optimal service time window, customer disappointment is 0. The further away the service time for customer i is from the optimal service time window, the higher the disappointment of delivery point i. Within the service time window [ET] i ,LT i When providing services to delivery point i, the disappointment level is 1 and a penalty cost will be incurred. n represents the number of delivery points. Minimize the driver load imbalance cost function: Among them, W k This represents the delivery distance of delivery vehicle k. d ij This represents the distance between delivery points i and j. This indicates whether vehicle k serves delivery point i and then delivery point j. The distances between vehicle deliveries are unbalanced, and K represents the number of delivery vehicles. S3: Improve the hybrid ant colony algorithm based on the route planning design for logistics distribution; S4: The objective function is solved by an improved hybrid ant colony algorithm, and a better non-dominated solution is obtained by using the NSGA-Ⅲ fast non-dominated sorting algorithm. The solution set of non-dominated solutions is obtained from the solution set of non-dominated solutions. The optimal solution for logistics distribution route planning is obtained from the solution set of non-dominated solutions.

2. The logistics route planning method based on the improved hybrid ant colony algorithm according to claim 1, characterized in that, The hybrid ant colony algorithm is improved based on the route planning design for logistics distribution, including: In the initial stage of the algorithm, a probability formula for updating the saving matrix is ​​introduced to guide the ants in path search; a piecewise function is used to improve the pheromone evaporation factor, resulting in a piecewise pheromone evaporation factor; elite ant colonies use the ant week model to update pheromones, while ordinary ants use the ant quantity model to update pheromones; and the 3-opt algorithm is used to improve the local search capability of the hybrid ant colony algorithm.

3. The logistics route planning method based on the improved hybrid ant colony algorithm according to claim 2, characterized in that, Introducing a probability formula for updating the savings matrix to guide ants in pathfinding includes: in, V represents the probability that ant k serves customers at delivery point j after serving customers at delivery point i. j k Let ξ represent the set of clients that ant k is allowed to serve next at time t (i.e., clients that have not yet been served); and ξ and Let ξ represent the first and second weighting coefficients, respectively, satisfying 0 ≤ ξ. and α represents the information heuristic factor, reflecting the relative importance of residual information in the ant colony's exploration process; β represents the expectation heuristic factor (visibility coefficient), reflecting the relative importance of path expectation in the ant colony's exploration process; γ represents the saving matrix heuristic factor, reflecting the relative importance of the saving matrix in the ant colony's exploration process; η ij τ represents the visibility from delivery point i to delivery point j. ij ET represents the pheromone concentration from delivery point i to delivery point j. j LT represents the earliest allowed service time for node j. j Represents node LT j The latest permitted service time, U ij Let t represent the savings matrix. j This indicates the service time for the delivery vehicle to serve customers at delivery point j.

4. The logistics route planning method based on the improved hybrid ant colony algorithm according to claim 2, characterized in that, Piecewise functions are used to improve the pheromone evaporation factor, including: Where ρ represents the pheromone evaporation factor, NC represents the current iteration number, and NC max This indicates the initial maximum number of iterations.

5. A logistics route planning method based on an improved hybrid ant colony algorithm according to claim 2, characterized in that, Elite ant colonies use a weekly ant model to update pheromones, while ordinary ant colonies use a population-based ant model, including: Elite ant colonies use a weekly ant model to update pheromones: Where, τ ij ρ represents the pheromone concentration along the path from delivery point i to delivery point j, and ρ represents the pheromone volatile factor. This indicates the higher pheromone content in elite ants compared to ordinary ants. ω represents the increase in pheromone content after elite ant k travels through the path i→j, ω represents the pheromone increment coefficient of elite ant, and t represents time. Common ants use an ant population model to update pheromones: Where, τ ij ρ represents the pheromone concentration along the path from delivery point i to delivery point j, ρ represents the pheromone evaporation factor, and Δτ represents the pheromone concentration along the path from delivery point i to delivery point j. ij This represents the total increase in pheromone content for all ordinary ants after traversing the path i→j. This indicates the increase in pheromone content after elite ant k travels along the path i→j.

6. A logistics route planning method based on an improved hybrid ant colony algorithm according to claim 2, characterized in that, The 3-opt algorithm is used to improve the local search capability of the hybrid ant colony algorithm, including: The 3-opt algorithm is used to remove three pairs of non-adjacent delivery point connections in the path. Then, the path lengths of the other seven different connection methods are calculated and compared, and the shortest path is selected as the new connection method. 3-opt has stronger local search capabilities than 2-opt. However, because 3-opt generates more new connection methods than 2-opt, it takes more time. Therefore, we modified the conditions for using the local search algorithm. During ant colony iteration, if the optimal path of the new generation is the same as the optimal path of the previous generation, the local search algorithm is not used. If they are different, the 3-opt local optimization algorithm is used to improve the local search capability of the algorithm.

7. The logistics route planning method based on the improved hybrid ant colony algorithm according to claim 1, characterized in that, S4 specifically includes: Step 1: Initialize parameters: Input the ant population size k, pheromone heuristic factor α, visibility heuristic factor β, saving matrix heuristic factor γ, and two weighting coefficients ξ and The number of iterations in the ant colony algorithm is NC, and the population size in the NSGA-Ⅲ algorithm is n. Step 2: Place K ants at the distribution center; Step 3: Using the delivery points in the logistics delivery task scheduling network model as nodes, the ants use the roulette wheel algorithm to select the next node from the node set, and then store the nodes they have passed through into the taboo list. Step 4: Determine if the ant has completed traversing all nodes. If not, go to step 3; otherwise, go to step 5. Step 5: Determine whether all ants have completed the traversal in this iteration. If not, increment the ant count by 1 and return to step 3; otherwise, jump to step 6. Step 6: After completing the traversal, each ant calculates its path according to the objective function of the logistics delivery task scheduling network model, obtaining the optimal path under the three objectives. The 3-opt local optimization algorithm is used for the optimal path of each objective, and the feasibility of the 7 new paths generated by the 3-opt local optimization algorithm is analyzed. If it is feasible, the path is recalculated according to the objective function of the logistics delivery task scheduling network model. If it is better, the original path is replaced; otherwise, the original path is used. Step 7: Update the pheromone according to the segmented pheromone evaporation factor, pheromone update method, and iteration number of the improved hybrid ant colony algorithm; Step 8: Use the NSGA-Ⅲ algorithm to perform non-dominated sorting of the population based on the updated pheromones to obtain the non-dominated solution set in the population; Step 9: Determine whether the maximum number of iterations set during initialization has been reached. If not, return to step 2; otherwise, jump to step 10. Step 10: Select the optimal solution that simultaneously optimizes the three objectives from the non-dominated solution set to obtain the specific transportation delivery route.

8. A logistics route planning method based on an improved hybrid ant colony algorithm according to claim 7, characterized in that, The optimal path for each objective is applied using the 3-opt local optimization algorithm, and the feasibility of the seven new paths generated by the 3-opt local optimization algorithm is analyzed, including: Delete three pairs of non-adjacent delivery point connections in the path, and then arrange the remaining paths to obtain 7 different connection methods. Compare the original path with the 7 new paths and select the shortest path as the new connection method.

9. A logistics route planning method based on an improved hybrid ant colony algorithm according to claim 7, characterized in that, The NSGA-III algorithm is used to perform non-dominated sorting of the population, including: A population of size n′ is stratified using a non-dominated sorting algorithm. The solutions in the solution set of the new population after iteration are then non-dominated and sorted. The first layer is denoted as f. 1 Its non-dominated ordinal value i rank A value of 1 indicates that the solution at this level is not dominated by any other solution, and is the non-dominated solution set of the entire population; then f is removed. 2 The remaining population is sorted using a non-dominated method to obtain the sequence value i. rank The second-level nondominated solution f is 2 2 f 2 The solution set of the next layer is selected as the non-dominated solution set of the whole. This process is repeated for all solutions, and the solution set of the first layer is selected as the non-dominated solution set of the whole.

Images