A method for planning a path of a three-level urban-rural common distribution network based on an adaptive hybrid algorithm

Abstract

The application discloses a three-level urban and rural common distribution network path planning method based on an adaptive hybrid algorithm, and particularly relates to a distribution path planning method which is optimized by combining a time window, vehicle capacity and path continuity as constraint conditions and an adaptive firework-quantum genetic hybrid algorithm. The application comprises the following steps: firstly, basic information such as coordinates, demand and time window of a city common distribution center, a county and township transfer station and a rural end self-pickup point is obtained, and vehicle resources are initialized; then, a mathematical model with the minimum total distribution cost as the target is constructed based on common distribution, and the model comprises fixed transportation cost, variable transportation cost and time penalty cost. The model is solved by using the adaptive firework-quantum genetic hybrid algorithm, which comprises quantum bit coding to generate an initial solution, an insertion algorithm to optimize the initial solution, a firework algorithm for global search and a quantum genetic algorithm for local optimization. In the optimization process, the diversity and convergence speed of the solution are improved by combining spark disturbance operation, quantum bit rotating gate dynamic adjustment and crossover mutation operator, and finally, the optimal distribution path meeting the time window constraint is output. The application is suitable for path planning of a complex three-level urban and rural distribution network, can significantly reduce transportation cost, improve distribution efficiency and optimize urban and rural logistics resource allocation.

Description

Technical Field

[0001] This invention relates to logistics distribution optimization technology, specifically to a method for route planning in a three-tiered urban-rural joint distribution network based on an adaptive hybrid algorithm. Background Technology

[0002] To address these issues, a three-tiered urban-rural distribution network based on collaborative distribution is gradually emerging as an innovative modern logistics model. By sharing resources, vehicles, and information platforms among urban collaborative distribution centers, county and township transit stations, and rural last-mile self-pickup points, collaborative distribution can effectively integrate resources at all levels of distribution nodes from urban to rural areas, reducing transportation costs and improving distribution efficiency. However, while there is considerable research on the vehicle routing problem in collaborative distribution, there is a lack of optimization models specifically for three-tiered urban-rural distribution. Most studies plan the two-stage delivery routes separately, without considering the impact of the time-varying demand at the last mile on existing vehicle routing plans. Furthermore, traditional route planning algorithms, such as genetic algorithms and ant colony algorithms, suffer from high computational complexity and unstable results, making them difficult to adapt to complex three-tiered distribution networks. In practice, the application of collaborative distribution in three-tiered urban-rural distribution networks still faces challenges such as complex route planning systems, optimization of delivery stages with varying demands, and multi-objective optimization requirements related to vehicles and time windows. Therefore, it is urgent to propose a method for route planning of a three-level urban-rural joint distribution network based on an adaptive hybrid algorithm, and to use an improved heuristic algorithm of adaptive fireworks-quantum genetic hybrid algorithm to solve the problem, so as to improve the efficiency and optimization effect of route planning. Summary of the Invention

[0003] This invention addresses the shortcomings of existing research in this field by proposing a method for route planning in a three-tiered urban-rural shared delivery network based on an adaptive hybrid algorithm. This invention considers constraints such as delivery routes, service time windows, and maximum vehicle load capacity, establishing a vehicle delivery route determination model with the optimization objective of minimizing the total delivery cost, including fixed transportation costs, variable transportation costs, and time penalty costs. An improved heuristic algorithm combining an adaptive fireworks-quantum genetic hybrid algorithm is employed to solve the model, thereby improving both route planning efficiency and optimization effectiveness.

[0004] This invention is characterized by a method for route planning in a three-tiered urban-rural joint distribution network based on an adaptive hybrid algorithm.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A three-tiered urban-rural distribution network system based on collaborative distribution includes:

[0007] The system comprises a three-tiered distribution network. First, to address issues such as unbalanced resource allocation, high logistics costs, and low transportation efficiency, it integrates the urban and rural distribution tasks and other resources of multiple logistics companies, sharing vehicle, customer, and warehousing resources to build a joint distribution platform. Second, the urban and rural distribution network includes a first-tier urban joint distribution center responsible for integrating urban goods resources and distributing them to county and township transit stations; several second-tier county and township transit stations serving as intermediate hubs, responsible for receiving goods from the urban distribution center and further distributing them to rural last-mile nodes; and several third-tier rural last-mile nodes providing self-pickup services to solve the last-mile delivery problem in rural areas. Customers' required goods are transported by large vehicles from the urban joint distribution center to the various county and township transit stations, and then by smaller vehicles from the county and township transit stations to the various rural last-mile self-pickup points. In the three-tiered urban and rural distribution network, township and county transfer stations serve as the link between the upper-level urban joint distribution centers and the lower-level farm self-pickup points. In the upper-level distribution network, goods are transported from the urban joint distribution centers to various township and county transfer stations by large vehicles, while in the lower-level distribution network, goods are transported from the township and county transfer stations to various rural self-pickup points by small vehicles. Since the demand at the township and county transfer stations in the upper-level distribution network may exceed the maximum capacity of the delivery vehicles, the upper-level distribution network is a demand-splitable distribution network, while the lower-level distribution network is a demand-indivisible distribution network.

[0008] A method for route planning in a three-tiered urban-rural joint distribution network based on an adaptive hybrid algorithm; comprising:

[0009] S1. Obtain vehicle resources, status, and delivery task information from the shared delivery platform;

[0010] S2. Based on the vehicle resources, status, and delivery task information, construct a mathematical model with the objective of minimizing the total delivery cost;

[0011] S3. An improved heuristic algorithm based on an adaptive fireworks-quantum genetic hybrid algorithm is used to solve the mathematical model, obtaining the optimal vehicle routes for the three-tier urban-rural distribution network model, including:

[0012] In the first stage, an initial solution representing the delivery path is generated using qubit encoding. The solution is represented by qubits, with the position of each delivery node initialized in a quantum state. Multiple potential paths are generated through the superposition of qubit states, and the initial solution is selected through quantum measurement.

[0013] In the second stage, to improve the quality of the initial solution, an insertion algorithm is used to optimize it. The insertion algorithm optimizes the initially generated solution by selecting the insertion position with the smallest increment, that is, when inserting an unvisited node into an existing path, choosing the position that minimizes the total cost of the path.

[0014] In the third stage, the Fireworks Algorithm is used for optimization. The Fireworks Algorithm comes into play after the initial solution is generated, exploring better solutions in the solution space through sparks. Each spark represents a solution, and the number of sparks and the explosion radius are adaptive, gradually adjusted during the iteration process. During the optimization process of the Fireworks Algorithm, each spark performs a local search on the current solution. By randomly perturbing the spark positions, the algorithm can find a better solution in the local solution space.

[0015] In the fourth stage, a quantum genetic algorithm is used to continue updating the optimal solution. In the quantum genetic algorithm, qubit encoding is used to represent each individual, i.e., each solution. Each individual's path is represented by the state of its qubits, and qubit encoding can efficiently represent the decision space. In each iteration, new solutions are continuously generated through qubit encoding and quantum genetic operations, optimizing the objective function and improving path planning.

[0016] In the fifth stage, non-dominated sorting and selection are performed to determine whether the termination condition is met and output the optimal solution.

[0017] According to a preferred embodiment of the present invention, step S2, which involves calculating the criterion parameters for vehicle delivery based on basic information and establishing a mathematical model with the objective of minimizing total delivery costs, includes:

[0018] The criteria parameters include the transportation costs of upper-level delivery vehicles, the transportation costs of lower-level delivery vehicles, and the time penalty cost.

[0019] The formula for calculating the transportation cost C1 of the upper-level delivery vehicle is as follows:

[0020]

[0021] Where c1 represents the unit distance transportation cost of the upper-layer network delivery vehicles, d ij r represents the distance from node i to node j. iju Let be a 0-1 variable, representing whether vehicle u passes through nodes i and j, D represents the set of urban joint distribution centers, S represents the set of rural and county transfer stations, and U represents the set of upper-level delivery vehicles.

[0022] The formula for calculating the transportation cost C2 of the lower-level delivery vehicles is as follows:

[0023]

[0024] Where c2 represents the unit distance transportation cost of the lower-level network delivery vehicles, d ij x represents the distance from node i to node j. ijk Let S be a 0-1 variable, representing whether vehicle k passes through nodes i and j. Let S represent the set of county and township transfer stations, P represent the set of rural last-mile self-pickup points, and K represent the set of lower-level delivery vehicles.

[0025] The formula for calculating the time penalty cost c3 is as follows:

[0026]

[0027] Where c3 represents the unit cost coefficient of the delay penalty, T i t represents the latest arrival time expected by customer i (i.e., the upper limit of the time window). i Let P represent the time point at which the vehicle arrives at rural terminal pickup point i, and let P represent the set of rural terminal pickup points.

[0028] Based on the above-mentioned criteria and parameters, an objective function for minimizing total delivery cost is established, expressed as:

[0029]

[0030] A further preferred embodiment uses the delivery time of the last node, the continuity of the delivery route, vehicle capacity, and the satisfaction of node demand as constraints for the objective function of minimizing the total delivery cost, to construct a three-tier urban-rural delivery network model based on collaborative delivery, including:

[0031]

[0032] In the formula, z su Q1 represents the transport volume of the upper-level delivery vehicle u when it visits the transfer station s, and Q1 is the maximum capacity constraint of the upper-level delivery network vehicles.

[0033]

[0034] In the formula, q i The demand for rural self-pickup points at the end of the chain;

[0035]

[0036] In the formula,

[0037]

[0038] In the formula,

[0039]

[0040] In the formula, e s The demand for rural and county transit stations;

[0041]

[0042] In the formula, M is a sufficiently large real number;

[0043] T iu+d is / v 1is +M×(1-r isu )≥T su i∈D∪S,s∈S,u∈U (15)

[0044] T iu +d is / v 1is -M×(1-r isu )≤T su i∈D∪S,s∈S,u∈U (16)

[0045] In the formula, T iu Let d be the time when the upper-level delivery vehicle u arrives at node i. is v is the distance between node i and node s. 1is T represents the speed of the upper-level delivery vehicle u from node i to node s. su The time it takes for the upper-level delivery vehicle u to arrive at the township / county transfer station s;

[0046] T ik +d ij / v 2ij +max(T su )+M×(1-x ijk )≥T jk i∈S∪P, s∈S, j∈P, k∈K (17)

[0047] T ik +d ij / v 2ij +max(T su )-M×(1-x ijk )≤T jk i∈S∪P, s∈S, j∈P, k∈K (18)

[0048] In the formula, T ik Let d be the time it takes for the lower-level delivery vehicle k to arrive at node i. ij v is the distance between node i and node j. 2ij For the speed of the lower-level delivery vehicle k from node i to node j, T jk The time it takes for the lower-level delivery vehicle k to reach the rural end-point self-pickup point j;

[0049] t i =max(t) ik (19)

[0050] In the formula, t i The time it takes for the lower-level delivery vehicle k to reach the rural end-point self-pickup point i;

[0051] in,

[0052] Equation (5) represents the maximum load constraint of vehicles in the upper-level delivery network;

[0053] Equation (6) represents the maximum load constraint of vehicles in the lower-level delivery network, and Q2 is the maximum capacity constraint of vehicles in the lower-level delivery network.

[0054] Equations (7) and (8) indicate that in the upper-level delivery network, each node j must have one and only one vehicle arriving from other nodes, and each node i must have one and only one vehicle departing from it, ensuring the integrity and continuity of the path.

[0055] Equations (9) and (10) indicate that each rural self-pickup point must be accessed by a vehicle and can only be accessed once, and only one vehicle can depart from it;

[0056] Equation (11) indicates that the demand for distribution at the township / county transfer station has been completed;

[0057] Equation (12) indicates that each vehicle in the upper-level distribution network departs from the urban joint distribution center, passes through several township and county transfer stations, and returns to the urban joint distribution center;

[0058] Equation (13) indicates that each vehicle in the lower-level distribution network departs from the township / county transfer station, passes through several rural self-pickup points, and returns to the township / county transfer station from which it departs;

[0059] Equation (14) represents the demand for goods at township and county transit stations;

[0060] Equations (15)-(19) represent the time it takes for goods to arrive at the point of demand;

[0061] According to a preferred embodiment of the present invention, step S3, which uses an improved heuristic algorithm based on an adaptive fireworks-quantum genetic hybrid algorithm to solve the optimal path optimization model and determine the optimal vehicle delivery route, includes:

[0062] a. Encode each township / county transfer station and terminal rural node with qubits, using one qubit to represent each node or terminal demand point:

[0063] |ψ i >=α i |0>+β i |1> (20)

[0064] Where, |α i | 2 and |β i | 2 These represent the probabilities of a node being selected, where |0> indicates that the node is not selected and |1> indicates that the node is selected. A path is represented by a combination of multiple qubits.

[0065] b. Initialize the population by randomly setting the initial amplitude α for each qubit. i and β i , guarantee |α i | 2 +|β i | 2 =1, and then the quantum state |ψ is determined through quantum measurement. i >Decode the specific selection value 0 or 1, combine the selected nodes into a path, and assign the unselected nodes to other vehicles. Based on the superposition state of the qubits, randomly generate N candidate initial solutions. After decoding, perform a legality check on each path and repair solutions that do not meet the constraints.

[0066] c. Optimize the initial solution. Use an insertion algorithm to optimize the initial solution, reduce path cost, and improve the feasibility of the solution. For each vehicle's path in the initial solution, traverse the unvisited nodes in turn and insert the unvisited node i into all possible positions in the path. Calculate the path cost increment for each insertion scheme:

[0067] ΔC=C new -C original (twenty one)

[0068] Among them, C new C represents the path cost after inserting a node. original Let represent the path cost before node insertion. To increase the diversity of the initial population, this invention selects the position with the smallest cost increment to insert node i and updates the path. After inserting all nodes, the optimized path is obtained. The optimized path is checked to see if it meets the constraints of capacity and time window. If not, the node order is adjusted or the node is reassigned to other vehicle paths to ensure the feasibility of the optimized initial solution.

[0069] d. Using the fireworks algorithm for dynamic global search, firstly calculate the number of sparks N from the explosion at the center of each firework using equation (22). s , where N max N min Let F(x) represent the maximum and minimum number of sparks, respectively. Let F(x) represent the fitness value of the current solution x, reflecting the quality of the solution. Then, for each individual vehicle, an explosion operation is used to generate sparks to determine the spark location. The spark generation location is determined by the perturbation range R, which is determined by equation (23), where R represents the explosion radius, and Δx represents the random perturbation factor, which is randomly generated within the range [-1, 1]. Calculate each spark solution x. s fitness value F(x) s For fireworks centers with good adaptability, a smaller explosion radius is set to generate a finer spark solution; for fireworks centers with poor adaptability, a larger explosion radius is set to expand the search range.

[0070]

[0071] x s =x + R·Δx, Δx∈[-1,1] (23)

[0072] Preferably, the nodes or paths of the current path solution x are perturbed to generate N. s For each spark solution, the following three perturbation operators are designed:

[0073] Node swapping operator: Randomly select two nodes in the path, swap their positions, and generate a new path solution;

[0074] Node insertion operator: Using equation (24), a node i in the path is moved to another position j;

[0075] x′=Insert(x,i,j) (24)

[0076] Path splitting operator: splits a path into two parts and redistributes them to different vehicles.

[0077] Preferably, if the spark solution does not satisfy the constraints, its fitness value is reduced through a penalty function:

[0078] F penalized =F(x) s )+λ·violation_penalty

[0079] Preferably, among the newly generated spark solutions, the spark solution with the highest fitness is selected as the new firework center, and the fitness F(x) of the current firework center is compared. new ) and the global optimal solution F best If the current solution is better, then update the global optimal solution:

[0080] F best =min(F best ,F(x new (25)

[0081] Preferably, N optimal fireworks centers are retained as the initial solutions for the next iteration;

[0082] e. Local optimization of the solution of the fireworks algorithm is performed using the quantum genetic algorithm. First, the population is updated by dynamically adjusting the rotation angle θ according to the fitness using equation (26):

[0083]

[0084] Preferably, the rotation amplitude is dynamically adjusted according to the fitness value of the solution, so as to locally enhance high-quality solutions and increase randomness for poor solutions, thereby avoiding getting trapped in local optima;

[0085]

[0086] Preferably, the following two operators are designed:

[0087] Crossover operator: Randomly selects two path solutions from the population, swaps some of their nodes, generates a new path solution, and inherits the superior characteristics of the parent.

[0088] x new =Crossover(x1,x2) (29)

[0089] Mutation operators: randomly adjust the order of nodes in a path, randomly select two nodes in the path to swap their positions, and randomly insert or delete a node;

[0090] Preferably, the fitness values ​​of the solutions after crossover and mutation are calculated, the solutions with higher fitness values ​​are retained, the solutions with lower fitness values ​​are eliminated, the newly generated solutions are merged with the current population, the N solutions with the highest fitness values ​​are selected as the new population and the global optimal solution is updated;

[0091] f. If the maximum number of iterations is reached or the fitness value of the optimal solution in the population changes less than a threshold over several consecutive generations, then output the optimal delivery path, node access order, arrival time, total cost of each path, and hierarchical delivery network cost. If the above requirements are not met, return to step d.

[0092] Compared with the prior art, the present invention has the following beneficial technical effects:

[0093] 1. By adopting an optimization strategy that combines adaptive fireworks algorithm and quantum genetic algorithm, the parameters of spark generation and qubit rotation gate are dynamically adjusted, which significantly improves the algorithm's solution efficiency and global search capability in complex optimization problems, thus enabling it to cope with dynamic environments and complex constraints more efficiently.

[0094] 2. The introduction of an insertion algorithm to optimize the initial solution can quickly reduce invalid paths in path planning, optimize the quality of the initial population, effectively shorten the algorithm's convergence time, and improve the overall solution quality;

[0095] 3. By combining an adaptive parameter adjustment mechanism, the algorithm can dynamically adjust the number of sparks and the rotation angle of qubits according to the fitness, effectively balancing the algorithm's global search and local search capabilities, and avoiding getting trapped in local optima;

[0096] 4. By integrating mutation and crossover operations from quantum genetic algorithms, the diversity of the population is effectively maintained, enhancing the scalability and robustness of the solution, making it suitable for solving large-scale optimization problems and multi-objective scenarios.

[0097] 5. The algorithm structure has strong modularity, and the weight ratio of the fireworks and quantum genetic parts can be flexibly adjusted according to the specific problem requirements, so as to adapt to the diverse needs of system scalability and application scenarios in future dynamic optimization problems. Attached Figure Description

[0098] To more clearly illustrate the technical solution of the present invention, the specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, wherein:

[0099] Figure 1 This is a block diagram of the method of the present invention;

[0100] Figure 2 This is a schematic diagram of the three-tiered urban and rural distribution network system structure based on collaborative distribution according to the present invention;

[0101] Figure 3 This is a schematic diagram of an improved heuristic algorithm based on an adaptive fireworks-quantum genetic hybrid algorithm provided in an embodiment of the present invention;

[0102] Figure 4 This is a schematic diagram of a perturbation operator in a fireworks algorithm provided in an embodiment of the present invention. Detailed Implementation

[0103] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0104] This application provides a method for path planning in a three-tiered urban-rural joint distribution network based on an adaptive hybrid algorithm, thus solving the path planning problem of a three-tiered urban-rural joint distribution network.

[0105] The technical solution in this application is to solve the above-mentioned technical problems, and the general idea is as follows:

[0106] In typical urban-rural joint delivery scenarios, goods are uniformly dispatched by various courier companies to a large urban joint distribution center. After sorting, they are then transported in batches to several rural transfer stations. These rural transfer stations temporarily store the transferred goods before unified delivery to rural self-pickup points. The vehicle routing problem in this delivery scenario can be abstracted as a two-layer vehicle routing problem, but current research on solving both layers uniformly is limited. This invention considers the large demand in urban-rural joint delivery scenarios, especially in the upper-layer delivery network where multiple vehicles are needed for delivery, which violates the access uniqueness constraint of general routing problems. Therefore, this embodiment relaxes this constraint, allowing upper-layer delivery network vehicles to access satellite stations multiple times, meaning the demand can be split for delivery.

[0107] Furthermore, since express delivery products have certain delivery time requirements, this embodiment of the invention also addresses a path planning problem with a soft time window. Therefore, this invention has not yet received much attention from scholars, but it meets the needs of some real-world delivery scenarios.

[0108] In summary, this invention can obtain a three-tier urban and rural distribution network route planning scheme based on joint distribution within an acceptable time, thereby significantly reducing the total cost of joint distribution of urban and rural logistics.

[0109] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0110] like Figure 1 As shown, this embodiment of the invention provides a method for route planning in a three-tiered urban-rural shared delivery network based on an adaptive hybrid algorithm, including:

[0111] S1. Obtain vehicle resources, status, and delivery task information from the shared delivery platform;

[0112] S2. Based on the vehicle resources and status and delivery task information, construct a mathematical model with the goal of minimizing the total delivery cost;

[0113] S3. An improved heuristic algorithm based on an adaptive fireworks-quantum genetic hybrid algorithm is used to solve the mathematical model, obtaining the optimal vehicle routes for the three-tier urban-rural distribution network model, including:

[0114] S31. In the first stage, an initial solution representing the delivery path is generated through qubit encoding. The solution is represented by qubits, and the position of each delivery node is initialized in quantum state form. Multiple potential paths are generated through the superposition of qubits, and the initial solution is selected through quantum measurement.

[0115] S32. In the second stage, to improve the quality of the initial solution, an insertion algorithm is used to optimize it. The insertion algorithm optimizes the initially generated solution by selecting the insertion position with the smallest increment, that is, when inserting an unvisited node into an existing path, choosing the position that minimizes the total cost of the path.

[0116] S33. In the third stage, the Fireworks Algorithm is used for optimization. The Fireworks Algorithm comes into play after the initial solution is generated, exploring better solutions in the solution space through sparks. Each spark represents a solution, and the number of sparks and the explosion radius are adaptive, gradually adjusted during the iteration process. During the optimization process of the Fireworks Algorithm, each spark performs a local search on the current solution. By randomly perturbing the spark positions, the algorithm can find a better solution in the local solution space.

[0117] S34. In the fourth stage, a quantum genetic algorithm is used to continue updating the optimal solution. In the quantum genetic algorithm, qubit encoding is used to represent each individual, i.e., each solution. Each individual's path is represented by the state of its qubits, and qubit encoding can effectively represent the decision space. In each iteration, new solutions are continuously generated through qubit encoding and quantum genetic operations, optimizing the objective function and improving path planning.

[0118] S35. In the fifth stage, non-dominated sorting and selection are performed to determine whether the termination condition is met and to output the optimal solution.

[0119] In the above scheme, considering the large demand in the urban-rural joint delivery scenario, especially in the upper-level delivery network where multiple vehicles are needed for delivery, which does not meet the access uniqueness constraint of the general route planning problem, the embodiments of the present invention relax this constraint, allowing vehicles in the upper-level delivery network to access satellite stations multiple times, that is, the demand can be split for delivery.

[0120] The following will detail each step of the above plan:

[0121] In step S1, the vehicle resources and status of the shared delivery platform and the delivery task information are obtained.

[0122] like Figure 3 As shown, this embodiment of the invention essentially solves a path planning problem for a three-tiered urban-rural joint distribution network based on an adaptive hybrid algorithm. In this problem, all information is deterministic. In the upper-level distribution network, vehicles depart from the urban joint distribution center, complete delivery tasks at township / county transfer stations, and return to the urban joint distribution center. In the lower-level distribution network, delivery vehicles depart from township / county transfer stations, complete delivery tasks at rural last-mile self-pickup points, and return to their originating township / county transfer stations. Each township / county transfer station is allowed to be visited multiple times, while each rural last-mile self-pickup point can only be visited once.

[0123] Based on this, and for the needs of subsequent modeling, it is known that this step requires at least the following information regarding vehicle resources and status, and delivery tasks from the same delivery platform:

[0124] D represents the set of urban joint distribution centers, S represents the set of township and county transfer stations, and P represents the set of rural last-mile self-pickup points.

[0125] U represents the upper-level delivery vehicle set, c1 represents the unit distance transportation cost of the upper-level network delivery vehicles, K represents the lower-level delivery vehicle set, c2 represents the unit distance transportation cost of the lower-level network delivery vehicles, and v 1is v represents the speed of the upper-level delivery vehicle u from node i to node s. 2ij Let c3 represent the speed of the lower-level delivery vehicle k from node i to node j, and T represent the unit cost coefficient of the delay penalty. i This represents the latest arrival time expected by customer i (i.e., the upper limit of the time window).

[0126] z su Let T be the transport volume of the upper-level delivery vehicle u when it visits the transfer station s, Q1 be the maximum capacity constraint of the upper-level delivery network vehicles, Q2 be the maximum capacity constraint of the lower-level delivery network vehicles, and T be the transport volume of the upper-level delivery vehicle u when it visits the transfer station s. iu Let d be the time when the upper-level delivery vehicle u arrives at node i. is Let T be the distance between node i and node s. jk The time it takes for the lower-level delivery vehicle k to reach the rural end-point self-pickup point j.

[0127] In step S2, a mathematical model is constructed based on the vehicle resources and status and delivery task information, with the goal of minimizing the total delivery cost.

[0128] The mathematical model includes:

[0129] According to a preferred embodiment of the present invention, based on the aforementioned basic information, criterion parameters for vehicle delivery are calculated to establish a total delivery cost. This minimum objective function includes:

[0130] Transportation costs for upper-level delivery vehicles, transportation costs for lower-level delivery vehicles, and time penalty costs;

[0131] The formula for calculating the transportation cost C1 of the upper-level delivery vehicle is as follows:

[0132]

[0133] Where c1 represents the unit distance transportation cost of the upper-layer network delivery vehicles, d ij r represents the distance from node i to node j. iju Let be a 0-1 variable, representing whether vehicle u passes through nodes i and j, D represents the set of urban joint distribution centers, S represents the set of rural and county transfer stations, and U represents the set of upper-level delivery vehicles.

[0134] The formula for calculating the transportation cost C2 of the lower-level delivery vehicles is as follows:

[0135]

[0136] Where c2 represents the unit distance transportation cost of the lower-level network delivery vehicles, d ij x represents the distance from node i to node j. ijk Let S be a 0-1 variable, representing whether vehicle k passes through nodes i and j. Let S represent the set of county and township transfer stations, P represent the set of rural last-mile self-pickup points, and K represent the set of lower-level delivery vehicles.

[0137] The formula for calculating the time penalty cost c3 is as follows:

[0138]

[0139] Where c3 represents the unit cost coefficient of the delay penalty, T i t represents the latest arrival time expected by customer i (i.e., the upper limit of the time window). i Let P represent the time point at which the vehicle arrives at rural terminal pickup point i, and let P represent the set of rural terminal pickup points.

[0140] Based on the above-mentioned criteria and parameters, an objective function for minimizing total delivery cost is established, expressed as:

[0141]

[0142] Using delivery time at the last mile, delivery route connectivity, vehicle capacity, and node demand satisfaction as constraints for the objective function of minimizing total delivery cost, a three-tiered urban-rural delivery network model based on collaborative delivery is constructed, including:

[0143]

[0144] In the formula, z su Q1 represents the transport volume of the upper-level delivery vehicle u when it visits the transfer station s, and Q1 is the maximum capacity constraint of the upper-level delivery network vehicles.

[0145]

[0146] In the formula, q i Q1 represents the demand at the rural self-pickup point i at the end point, and Q2 represents the maximum capacity constraint of the vehicles in the lower-level delivery network.

[0147]

[0148] In the formula,

[0149]

[0150] In the formula,

[0151]

[0152] In the formula, e s The demand for rural and county transit stations;

[0153]

[0154] In the formula, M is a sufficiently large real number;

[0155] T iu +d is / v 1is +M×(1-r isu )≥T su i∈D∪S,s∈S,u∈U (15)

[0156] T iu +d is / v 1is -M×(1-r isu )≤T su i∈D∪S,s∈S,u∈U (16)

[0157] In the formula, T iu Let d be the time when the upper-level delivery vehicle u arrives at node i. is v is the distance between node i and node s. 1is T represents the speed of the upper-level delivery vehicle u from node i to node s. su The time it takes for the upper-level delivery vehicle u to arrive at the township / county transfer station s;

[0158] T ik +d ij / v 2ij +max(T su )+M×(1-x ijk )≥T jk i∈S∪P, s∈S, j∈P, k∈K (17)

[0159] T ik +d ij / v 2ij +max(T su )-M×(1-x ijk )≤T jk i∈S∪P, s∈S, j∈P, k∈K (18)

[0160] In the formula, T ik Let d be the time it takes for the lower-level delivery vehicle k to arrive at node i. ij v is the distance between node i and node j.2ij For the speed of the lower-level delivery vehicle k from node i to node j, T jk The time it takes for the lower-level delivery vehicle k to reach the rural end-point self-pickup point j;

[0161] t i =max(t) ik (19)

[0162] In the formula, t i The time it takes for the lower-level delivery vehicle k to reach the rural end-point self-pickup point i;

[0163] in,

[0164] Equation (5) represents the maximum load constraint of vehicles in the upper-level delivery network;

[0165] Equation (6) represents the maximum load constraint of vehicles in the lower-level delivery network, and Q2 is the maximum capacity constraint of vehicles in the lower-level delivery network.

[0166] Equations (7) and (8) indicate that in the upper-level delivery network, each node j must have one and only one vehicle arriving from other nodes, and each node i must have one and only one vehicle departing from it, ensuring the integrity and continuity of the path.

[0167] Equations (9) and (10) indicate that each rural self-pickup point must be accessed by a vehicle and can only be accessed once, and only one vehicle can depart from it;

[0168] Equation (11) indicates that the demand for distribution at the township / county transfer station has been completed;

[0169] Equation (12) indicates that each vehicle in the upper-level distribution network departs from the urban joint distribution center, passes through several township and county transfer stations, and returns to the urban joint distribution center;

[0170] Equation (13) indicates that each vehicle in the lower-level distribution network departs from the township / county transfer station, passes through several rural self-pickup points, and returns to the township / county transfer station from which it departs;

[0171] Equation (14) represents the demand for goods at township and county transit stations;

[0172] Equations (15)-(19) represent the time it takes for goods to arrive at the point of demand;

[0173] In step S3, an improved heuristic algorithm based on an adaptive fireworks-quantum genetic hybrid algorithm is used to solve the optimal path optimization model and determine the optimal vehicle delivery route.

[0174] This embodiment proposes a three-tiered urban-rural distribution network route planning framework based on a shared distribution model to solve the vehicle routing problem. It can obtain a three-tiered urban-rural distribution network route planning scheme based on shared distribution within an acceptable timeframe, significantly reducing the total cost of shared urban-rural logistics distribution. This includes:

[0175] like Figure 3 As shown, in each stage:

[0176] S31. Encode each township / county transfer station and terminal rural node using qubits, with each node or terminal demand point represented by one qubit:

[0177] |ψ i >=α i |0>+β i |1> (20)

[0178] Where, |α i | 2 and |β i | 2 These represent the probabilities of a node being selected, where |0> indicates that the node is not selected and |1> indicates that the node is selected. A path is represented by a combination of multiple qubits.

[0179] S32. Initialize the population by randomly setting the initial amplitude α for each qubit. i and β i , guarantee |α i | 2 +|β i | 2 =1, and then the quantum state |Ψ is determined through quantum measurement. i >Decode the specific selection value 0 or 1, combine the selected nodes into a path, and assign the unselected nodes to other vehicles. Based on the superposition state of the qubits, randomly generate N candidate initial solutions. After decoding, perform a legality check on each path and repair solutions that do not meet the constraints.

[0180] Furthermore, the parameter settings for the following algorithm can also be provided as an example:

[0181] Define the population size, for example, set the population size to N=50, which means that 50 candidate solutions are generated in the initial stage, each individual is represented by a string of qubits, and the initial state can be uniformly distributed as follows: In the parameter initialization of the fireworks algorithm, the maximum number of sparks is M. max =20, in the initialization of the quantum genetic algorithm parameters, the initial value of the angular step size of the rotating door is set to Δθ = 0.05π, and the perturbation range of spark generation is set to The termination condition for the entire algorithm is set to a maximum number of iterations T. max=200, or the change in the objective function is less than the threshold ∈ =10 -6 .

[0182] S33. Optimize the initial solution. Use an insertion algorithm to optimize the initial solution, reduce path cost, and improve the feasibility of the solution. For each vehicle's path in the initial solution, traverse the unvisited nodes in turn and insert the unvisited node i into all possible positions in the path. Calculate the path cost increment for each insertion scheme:

[0183] ΔC=C new -C original (twenty one)

[0184] Among them, C new C represents the path cost after inserting a node. original Let represent the path cost before node insertion. To increase the diversity of the initial population, this invention selects the position with the smallest cost increment to insert node i and updates the path. After inserting all nodes, the optimized path is obtained. The optimized path is checked to see if it meets the constraints of capacity and time window. If not, the node order is adjusted or the node is reassigned to other vehicle paths to ensure the feasibility of the optimized initial solution.

[0185] S34. Using the fireworks algorithm for dynamic global search, first calculate the number of sparks N at the center of each firework using equation (22). s , where N max N min Let F(x) represent the maximum and minimum number of sparks, respectively. Let F(x) represent the fitness value of the current solution x, reflecting the quality of the solution. Then, for each individual vehicle, an explosion operation is used to generate sparks to determine the spark location. The spark generation location is determined by the perturbation range R, which is determined by equation (23), where R represents the explosion radius, and Δx represents the random perturbation factor, which is randomly generated within the range [-1, 1]. Calculate each spark solution x. s fitness value F(x) s For fireworks centers with good adaptability, a smaller explosion radius is set to generate a finer spark solution; for fireworks centers with poor adaptability, a larger explosion radius is set to expand the search range.

[0186]

[0187] x s =x + R·Δx, Δx∈[-1,1] (23)

[0188] like Figure 4 As shown, this invention perturbs the nodes or paths of the current path solution x, generating N. s For each spark solution, the following three perturbation operators are designed:

[0189] Node swapping operator: Randomly select two nodes in the path, swap their positions, and generate a new path solution;

[0190] Node insertion operator: Using equation (24), a node i in the path is moved to another position j;

[0191] x′=Insert(x,i,j) (24)

[0192] Path splitting operator: splits a path into two parts and redistributes them to different vehicles.

[0193] Preferably, if the spark solution does not satisfy the constraints, its fitness value is reduced through a penalty function:

[0194] F penalized =F(x) s )+λ·violation_penalty

[0195] Preferably, among the newly generated spark solutions, the spark solution with the highest fitness is selected as the new firework center, and the fitness F(x) of the current firework center is compared. new ) and the global optimal solution F best If the current solution is better, then update the global optimal solution:

[0196] F best =min(F best ,F(x new (25)

[0197] Preferably, N optimal fireworks centers are retained as the initial solutions for the next iteration;

[0198] S35. The solution of the fireworks algorithm is locally optimized using the quantum genetic algorithm. First, the population is updated by dynamically adjusting the rotation angle θ according to the fitness using equation (26):

[0199]

[0200] Preferably, the rotation amplitude is dynamically adjusted according to the fitness value of the solution, so as to locally enhance high-quality solutions and increase randomness for poor solutions, thereby avoiding getting trapped in local optima;

[0201]

[0202] Preferably, the following two operators are designed:

[0203] Crossover operator: Randomly selects two path solutions from the population, swaps some of their nodes, generates a new path solution, and inherits the superior characteristics of the parent.

[0204] x new=Crossover(x1,x2) (29)

[0205] Mutation operators: randomly adjust the order of nodes in a path, randomly select two nodes in the path to swap their positions, and randomly insert or delete a node;

[0206] Preferably, the fitness values ​​of the solutions after crossover and mutation are calculated, the solutions with higher fitness values ​​are retained, the solutions with lower fitness values ​​are eliminated, the newly generated solutions are merged with the current population, the N solutions with the highest fitness values ​​are selected as the new population and the global optimal solution is updated;

[0207] Preferably, if the maximum number of iterations is reached or the fitness value of the optimal solution in the population changes less than a threshold over several consecutive generations, then the optimal delivery path, node access order, arrival time, total cost of each path, and hierarchical delivery network cost are output. If the above requirements are not met, step de is repeated.

[0208] Compared with the prior art, the present invention has the following beneficial technical effects:

[0209] 1. By adopting an optimization strategy that combines adaptive fireworks algorithm and quantum genetic algorithm, the parameters of spark generation and qubit rotation gate are dynamically adjusted, which significantly improves the algorithm's solution efficiency and global search capability in complex optimization problems, thus enabling it to cope with dynamic environments and complex constraints more efficiently.

[0210] 2. The introduction of an insertion algorithm to optimize the initial solution can quickly reduce invalid paths in path planning, optimize the quality of the initial population, effectively shorten the algorithm's convergence time, and improve the overall solution quality;

[0211] 3. By combining an adaptive parameter adjustment mechanism, the algorithm can dynamically adjust the number of sparks and the rotation angle of qubits according to the fitness, effectively balancing the algorithm's global search and local search capabilities, and avoiding getting trapped in local optima;

[0212] 4. By integrating mutation and crossover operations from quantum genetic algorithms, the diversity of the population is effectively maintained, enhancing the scalability and robustness of the solution, making it suitable for solving large-scale optimization problems and multi-objective scenarios.

[0213] 5. The algorithm structure has strong modularity, and the weight ratio of the fireworks and quantum genetic parts can be flexibly adjusted according to the specific problem requirements, so as to adapt to the diverse needs of system scalability and application scenarios in future dynamic optimization problems.

[0214] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications, alterations, and equivalent transformations made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for route planning in a three-tiered urban-rural joint distribution network based on an adaptive hybrid algorithm, characterized in that... Includes the following steps: S1. Obtain vehicle resources, status, and delivery task information from the shared delivery platform; S2. Based on the vehicle resources, status, and delivery task information, construct a mathematical model with the objective of minimizing the total delivery cost; S3. An improved heuristic algorithm based on an adaptive fireworks-quantum genetic hybrid algorithm is used to solve the mathematical model, obtaining the optimal vehicle routes for the three-tier urban-rural distribution network model, including: In the first stage, an initial solution representing the delivery path is generated by encoding with qubits. The solution is represented by qubits, and the position of each delivery node is initialized in the form of a quantum state. Multiple potential paths are generated through the superposition state of qubits, and the initial solution is selected by quantum measurement. In the second stage, in order to improve the quality of the initial solution, an insertion algorithm is used to optimize the initial solution. The insertion algorithm optimizes the initially generated solution by selecting the insertion position with the smallest increment, that is, when inserting an unvisited node into an existing path, selecting the position that minimizes the total cost of the path. In the third stage, the fireworks algorithm is used for optimization. The fireworks algorithm comes into play after the initial solution is generated, and explores better solutions in the solution space through sparks. Each spark represents a solution, and the number of sparks and the explosion radius are adaptive and are gradually adjusted with the iteration process. During the optimization process of the fireworks algorithm, each spark performs a local search for the current solution. By randomly perturbing the position of the sparks, the algorithm can find a better solution in the local solution space. In the fourth stage, a quantum genetic algorithm is used to continue updating the better solution. In the quantum genetic algorithm, qubit encoding is used to represent each individual, i.e., each solution. Each individual represents its path through the state of qubits, and qubit encoding can effectively represent the decision space. In each iteration, new solutions are continuously generated through qubit encoding and quantum genetic operations to optimize the objective function and improve path planning. In the fifth stage, non-dominated sorting and selection are performed to determine whether the termination condition is met and output the optimal solution. The mathematical model described in step S2 includes: Transportation costs for upper-level delivery vehicles, transportation costs for lower-level delivery vehicles, and time penalty costs; The formula for calculating the transportation cost C1 of the upper-level delivery vehicle is as follows: Where c1 represents the unit distance transportation cost of the upper-layer network delivery vehicles, d ij r represents the distance from node i to node j. iju The variables are 0-1, representing whether vehicle u passes through nodes i and j, D represents the set of urban joint distribution centers, S represents the set of rural and county transfer stations, and U represents the set of upper-level delivery vehicles; The formula for calculating the transportation cost C2 of the lower-level delivery vehicles is as follows: Where c2 represents the unit distance transportation cost of the lower-level network delivery vehicles, d ij x represents the distance from node i to node j. ijk Let S be a 0-1 variable, representing whether vehicle k passes through nodes i and j; S represents the set of county and township transfer stations; P represents the set of rural last-mile self-pickup points; and K represents the set of lower-level delivery vehicles. The formula for calculating the time penalty cost c3 is as follows: Where c3 represents the unit cost coefficient of the delay penalty, T i t represents the latest arrival time expected by customer i, i.e., the upper limit of the time window. i Let P represent the time point at which the vehicle arrives at rural terminal pickup point i, and let P represent the set of rural terminal pickup points. The objective function for minimizing total delivery cost is defined as follows: minC=λ1C1+λ2C2+λ3C3 =λ1c1∑ i∈D∪S ∑ j∈D∪S ∑ u∈U d ij r iju +λ2c2∑ i∈S∪P ∑ j∈S∪P ∑ k∈K d ij x ijk +λ3c3∑ i∈P max(t i -T i ,0)(4)。 2. The method for route planning of a three-tiered urban-rural joint distribution network as described in claim 1, characterized in that, The mathematical model also includes: Constraints: Using delivery time at the last mile, delivery route connectivity, vehicle capacity, and node demand satisfaction as constraints for the objective function of minimizing total delivery cost, a three-tiered urban-rural delivery network model based on collaborative delivery is constructed, including: In the formula, z su Q1 represents the transport volume of the upper-level delivery vehicle u when it visits the transfer station s, and Q1 is the maximum capacity constraint of the upper-level delivery network vehicles. In the formula, q i The demand for rural self-pickup points at the end of the chain; In the formula, In the formula, In the formula, e s The demand for rural and county transit stations; In the formula, M is a sufficiently large real number; T iu +d is / v 1is +M×(1-r isu )≥T su i∈D∪S,s∈S,u∈U (15) T iu +d is / v 1is -M×(1-r isu )≤T su i∈D∪S,s∈S,u∈U (16) In the formula, T iu Let d be the time when the upper-level delivery vehicle u arrives at node i. is v is the distance between node i and node s. 1is T represents the speed of the upper-level delivery vehicle u from node i to node s. su The time it takes for the upper-level delivery vehicle u to arrive at the township / county transfer station s; T ik +d ij / v 2ij +max(T su )+M×(1-x ijk )≥T jk i∈S∪P,s∈S,j∈P,k∈K (17) T ik +d ij / v 2ij +max(T su )-M×(1-x ijk )≤T jk i∈S∪P,s∈S,j∈P,k∈K (18) In the formula, T ik Let d be the time it takes for the lower-level delivery vehicle k to arrive at node i. ij v is the distance between node i and node j. 2ij For the speed of the lower-level delivery vehicle k from node i to node j, T jk The time it takes for the lower-level delivery vehicle k to reach the rural end-point self-pickup point j; t i =max(t ik ) (19) In the formula, t i The time it takes for the lower-level delivery vehicle k to reach the rural end-point self-pickup point i; in, Equation (5) represents the maximum load constraint of vehicles in the upper-level delivery network; Equation (6) represents the maximum load constraint of vehicles in the lower-level delivery network, and Q2 is the maximum capacity constraint of vehicles in the lower-level delivery network. Equations (7) and (8) indicate that in the upper-level delivery network, each node j must have one and only one vehicle arriving from other nodes, and each node i must have one and only one vehicle departing from it, ensuring the integrity and continuity of the path. Equations (9) and (10) indicate that each rural self-pickup point must be accessed by a vehicle and can only be accessed once, and only one vehicle can depart from it; Equation (11) indicates that the demand for distribution at the township / county transfer station has been completed; Equation (12) indicates that each vehicle in the upper-level distribution network departs from the urban joint distribution center, passes through several township and county transfer stations, and returns to the urban joint distribution center; Equation (13) indicates that each vehicle in the lower-level distribution network departs from the township / county transfer station, passes through several rural self-pickup points, and returns to the township / county transfer station from which it departs; Equation (14) represents the demand for goods at township and county transit stations; Equations (15) to (19) represent the time it takes for goods to arrive at the point of demand.

3. The method for route planning of a three-tiered urban-rural joint distribution network as described in claim 2, characterized in that, Step S3 includes: S31. Encode each township / county transfer station and terminal rural node using qubits, with each node or terminal demand point represented by one qubit: |ψ i >=a i |0>+β i |1> (20) Where, |α i | 2 and |β i | 2 These represent the probabilities of a node being selected, where |0> indicates that the node is not selected and |1> indicates that the node is selected. A path is represented by a combination of multiple qubits. S32. Initialize the population by randomly setting the initial amplitude α for each qubit. i and β i , guarantee |α i | 2 +|β i | 2 =1, and then the quantum state |ψ is determined through quantum measurement. i >Decode the specific selection value 0 or 1, combine the selected nodes into a path, and assign the unselected nodes to other vehicles. Based on the superposition state of the qubits, randomly generate N candidate initial solutions. After decoding, perform a legality check on each path and repair solutions that do not meet the constraints. S33. Optimize the initial solution. Use an insertion algorithm to optimize the initial solution, reduce path cost, and improve the feasibility of the solution. For each vehicle's path in the initial solution, traverse the unvisited nodes in turn and insert the unvisited node i into all possible positions in the path. Calculate the path cost increment for each insertion scheme: ΔC=C new -C original (21) Among them, C new C represents the path cost after inserting a node. original Let i represent the path cost before the node is inserted. To increase the diversity of the initial population, node i is inserted at the position with the smallest cost increment, and the path is updated. After inserting all nodes, the optimized path is obtained. Check whether the optimized path meets the constraints of capacity and time window. If it does not meet the constraints, adjust the node order or reassign the node to other vehicle paths to ensure that the optimized initial solution is feasible. After the initial solution is optimized, calculate the fitness value of each path and update the population optimal solution. S34. Using the fireworks algorithm for dynamic global search, first calculate the number of sparks N at the center of each firework using equation (22). s , where N max N min Let F(x) represent the maximum and minimum number of sparks, respectively. Let F(x) represent the fitness value of the current solution x, reflecting the quality of the solution. Then, for each vehicle, an explosion operation is used to generate sparks to determine the spark location. The spark generation location is determined by the perturbation range R, which is determined by equation (23), where R represents the explosion radius and Δx represents the random perturbation factor, which is randomly generated within the range of [-1,1]. Calculate each spark solution x. s fitness value F(x) s For fireworks centers with good adaptability, a smaller explosion radius is set to generate a finer spark solution; for fireworks centers with poor adaptability, a larger explosion radius is set to expand the search range. x s =x+R·Δx,Δx∈[-1,1] (23) Perturb the nodes or paths of the current path solution x to generate N. s For each spark solution, the following three perturbation operators are designed: Node swapping operator: Randomly select two nodes in the path, swap their positions, and generate a new path solution; Node insertion operator: Using equation (24), a node i in the path is moved to another position j; x′=Insert(x,i,j) (24) Path splitting operator: splits a path into two parts and redistributes them to different vehicles; Preferably, if the spark solution does not satisfy the constraints, its fitness value is reduced through a penalty function: F penalized =F(x s )+λ·violation_penalty Preferably, among the newly generated spark solutions, the spark solution with the highest fitness is selected as the new firework center, and the fitness F(x) of the current firework center is compared. new ) and the global optimal solution F best If the current solution is better, then update the global optimal solution: F best =min(F best ,F(x new )) (25) Preferably, N optimal fireworks centers are retained as the initial solutions for the next iteration; S35. The solution of the fireworks algorithm is locally optimized using the quantum genetic algorithm. First, the population is updated by dynamically adjusting the rotation angle θ according to the fitness using equation (26): Preferably, the rotation amplitude is dynamically adjusted according to the fitness value of the solution, so as to locally enhance high-quality solutions and increase randomness for poor solutions, thereby avoiding getting trapped in local optima; Preferably, the following two operators are designed: Crossover operator: Randomly selects two path solutions from the population, swaps some of their nodes, generates a new path solution, and inherits the superior characteristics of the parent. x new =Crossover(x1,x2) (29) Mutation operators: randomly adjust the order of nodes in a path, randomly select two nodes in the path to swap their positions, and randomly insert or delete a node; After crossover and mutation, calculate the fitness value of the solutions, retain the solutions with higher fitness values, eliminate the solutions with lower fitness values, merge the newly generated solutions with the current population, select the N solutions with the highest fitness as the new population, and update the global optimal solution. S36. If the maximum number of iterations is reached or the fitness value of the optimal solution in the population changes less than the threshold within several consecutive generations, then output the optimal delivery path, node access order and arrival time, as well as the total cost and hierarchical delivery network cost corresponding to each path; if the above requirements are not met, return to step S34.

Images