A method, system, and storage medium for optimizing multimodal transport delivery routes.

Abstract

This invention relates to the field of logistics management technology, specifically to a method, system, and storage medium for optimizing multimodal transport delivery routes. The method includes: generating individual solutions based on basic cargo transportation data to form an initial population; updating the initial population using a hybrid optimization algorithm based on a multi-objective constraint model; and obtaining the optimal individual solution from the initial population as the Pareto optimal solution set. The hybrid optimization algorithm employs an adaptive PSO-CS operator to update the initial population. The PSO operator updates the speed and position of individual solutions and, based on congestion, obtains the globally optimal individual solution as the offspring population. The CS operator, through adaptive scheduling decisions, performs global or local CS perturbations to fine-tune the obtained individual solutions, generating the offspring population. This scheme effectively handles the heterogeneous search space and complex constraints of MTOP, dynamically and adaptively refining the shrinkage speed, thus improving search efficiency and quality.

Description

Technical Field

[0001] This invention relates to the field of logistics management technology, specifically to a method, system, and storage medium for optimizing multimodal transport delivery routes. Background Technology

[0002] With the development of global trade, multimodal transport has become a core organizational form of modern logistics transportation due to its ability to integrate multiple modes of transport to achieve efficient and low-cost logistics services. The core of multimodal transport lies in designing a comprehensive optimal transport route from origin to destination, which is essentially a complex multi-objective, multi-constraint combinatorial optimization problem. Researchers often use heuristic algorithms to solve such problems, such as NSGA-II (Non-Dominated Sorting Genetic Algorithm) and PSO (Particle Swarm Optimization). While heuristic algorithms have advantages in solving these problems, they also have performance limitations. Specifically, NSGA-II, as a classic algorithm for solving multi-objective problems, can handle the priority relationships between multiple objectives and maintain the distribution of the population on the Pareto front, possessing advantages such as strong global search capability, clear process, and ease of operation. However, its ability to handle the Pareto front is limited, and its convergence speed is slow. PSO has the advantage of fast convergence speed when solving multi-objective problems, but it is prone to getting trapped in local optima, leading to insufficient solution diversity.

[0003] To improve performance, existing technologies attempt to combine different algorithms, such as: Previous studies have combined NSGA-II with PSO and introduced dynamic entropy weights to balance global optimization capability and local solution accuracy. However, this method is highly dependent on solution distribution and is prone to getting trapped in local optimum loops. Alternatively, to address the issue of PSO easily getting trapped in local optima, researchers proposed combining the Cuckoo Search (CS) algorithm with the PSO algorithm, allowing the algorithms to run alternately and iteratively. This solves the problems of premature convergence and weak local search ability of the PSO algorithm. However, it is difficult to balance the efficiency and stability of the algorithm. The iterative switching between PSO and CS is not flexible enough and can easily lead to a waste of search resources. Moreover, the optimization effect is too dependent on the sample quality.

[0004] In summary, the problems with existing technologies include: general multi-objective algorithms have difficulty effectively handling the heterogeneous search space and complex constraints of MTOP (multi-task optimization problem), resulting in low search efficiency and poor solution quality; and most hybrid algorithms are statically parallel, failing to adapt to the dynamic needs of the optimization process by adjusting search strategies in an adaptive and refined manner.

[0005] Therefore, there is an urgent need for a multimodal transport delivery route optimization method, system, and storage medium that can effectively handle the heterogeneous search space and complex constraints of MTOP, dynamically and adaptively refine the shrinkage speed, and improve search efficiency and quality. This would solve the problems of slow convergence speed, easy getting trapped in local optima, and uneven distribution of Pareto front in existing technologies when solving multi-objective optimization problems, and quickly obtain high-quality multimodal transport optimized route solutions. Summary of the Invention

[0006] One of the objectives of this invention is to provide a multimodal transport delivery route optimization method that can effectively handle the heterogeneous search space and complex constraints of MTOP, dynamically and adaptively refine the shrinkage speed, and improve search efficiency and quality. This solves the problems of slow convergence speed, easy getting trapped in local optima, and uneven distribution of Pareto front in the existing technology when solving multi-objective optimization problems, and quickly obtains high-quality multimodal transport optimized route solutions.

[0007] The basic solution provided by this invention is a multimodal transport delivery route optimization method, comprising: S1. Construct a multi-objective constraint model for multimodal transport delivery routes; S2. Obtain basic data on cargo transportation, generate individual solutions, form an initial population, and update the initial population using a hybrid optimization algorithm based on a multi-objective constraint model. Obtain the optimal individual solution in the initial population as the Pareto optimal solution set. The hybrid optimization algorithm uses the adaptive PSO-CS operator to mutate the individual schemes in the parent-child population, generate the offspring population, merge the offspring population and the parent-child population, and update the initial population. The PSO operator in the adaptive PSO-CS operator updates the velocity and position of individual solutions and obtains the globally optimal individual solution based on the crowding degree, which is then used as the offspring population. In the adaptive PSO-CS operator, the CS operator makes adaptive scheduling decisions, performs global or local CS perturbations, fine-tunes the acquired individual schemes, and generates offspring populations.

[0008] Furthermore, the multi-objective constraint model includes: a multi-objective function; The multi-objective functions include: transportation cost objective function. and transportation time objective function : , ; in A set of nodes; For the set of transport arc segments; For transportation modes; , , For nodes, , , ∈ ; , These are the starting and ending points for cargo transportation. , ∈N; For nodes To the node arc segment, , ∈ ; , For the mode of transport designation, , ∈M; , For nodes The set of preceding and following nodes, , ⊆ ; For connecting nodes A collection of transportation methods ⊆ ; For arc The collection of transportation methods on the Internet ⊆ .

[0009] Furthermore, the multi-objective constraint model also includes: constraint conditions; The constraints include: The first constraint is used to restrict the cargo relationship between nodes: ; The second constraint is used to restrict the transit nodes. arrive Choose at most one mode of transport: ; The third constraint is used to restrict transit nodes. arrive Choose at most one transit method: ; The fourth constraint restricts transit nodes from receiving and transferring goods using the same mode of transport: ; The fifth constraint is used to restrict the inflow and outflow of goods at transit nodes to be the same: .

[0010] Furthermore, S2 includes: S201. Initialize the number of iterations t=1, and use heuristic random walk to generate an initial population containing several initial individual schemes; S202. Based on the multi-objective function, evaluate the fitness of individual schemes in the initial population, perform non-dominated sorting based on fitness, divide the population into levels, and calculate the crowding of individual schemes. S303. Using a tournament selection method, based on hierarchy and crowding, select the first preset number of parent individuals from the initial population to form a parent-child group; S204. Using the adaptive PSO-CS operator, the individual schemes in the parent-child population are mutated to generate the offspring population. The offspring population and the parent-child population are merged to update the initial population. S205, determine whether the current iteration meets the termination condition. If yes, output the individual solutions of the first level F1 of the initial population as the Pareto optimal solution set. If no, increase the number of iterations and execute S202 again.

[0011] Furthermore, S202 includes: configuring each individual scheme of the initial population. Substitute them into the multi-objective function respectively: Assess the fitness of individual programs; Based on fitness, perform non-dominated sorting and divide into levels; according to Calculate the crowding degree of individual options; where , The target of this layer The next fitness value and the previous fitness value, , These represent the maximum and minimum fitness values ​​in this layer, respectively.

[0012] Furthermore, S203 includes: randomly selecting a second preset number of individuals from the initial population; Compare the levels of the extracted individual solutions, and retain the individual solution with the better level as the parent individual. If the levels of the individual solutions are the same, compare the crowding of the individual solutions, and retain the individual solution with the greater crowding as the parent individual. Among them, retaining the individual solution with the smaller level means retaining the individual solution with the level lower than the preset level; retaining the individual solution with the greater crowding means retaining the individual solution with the greater crowding than the preset crowding. Determine whether the number of parent individuals is greater than or equal to the first preset number. If yes, execute S204; otherwise, execute the scheme of randomly selecting a second preset number of individuals from the initial population again.

[0013] Furthermore, the operation of the PSO operator includes: Each individual scheme updates its velocity and position according to the PSO operator's velocity and position vector formulas. Based on the first level of the Pareto front, the globally optimal individual scheme is selected using a roulette wheel selection method based on congestion, and the new velocity vector is obtained by substituting it into the velocity update formula. Substituting into the position update formula yields the new position vector formula. The new position vector formula is used as the current candidate solution.

[0014] Furthermore, the adaptive scheduling determination includes: Based on the current iteration number and preset iteration threshold Calculate the global disturbance trigger probability and the probability of triggering local disturbances : , ; Generate random numbers Combined with the global disturbance trigger probability and the probability of triggering local disturbances Perform adaptive scheduling judgment, if Less than Then perform CS global perturbation and Lévy flight global perturbation. ,like Less than Then, perform local perturbation of CS and randomly select two better individuals for fine-tuning. .

[0015] The second objective of this invention is to provide a multimodal transport delivery route optimization system that can effectively handle the heterogeneous search space and complex constraints of MTOP, dynamically and adaptively adjust the shrinkage speed, and improve search efficiency and quality. This solves the problems of slow convergence speed, easy getting trapped in local optima, and uneven distribution of Pareto front in the existing technology when solving multi-objective optimization problems, and quickly obtains high-quality multimodal transport optimization route solutions.

[0016] This invention provides a second basic solution: a multimodal transport delivery route optimization system for executing the above-mentioned multimodal transport delivery route optimization method, comprising: The data acquisition module is used to acquire basic data on cargo transportation. The model building module is used to construct multi-objective constrained models of multimodal transport delivery routes. The collaborative analysis module is used to generate individual solutions based on basic cargo transportation data, form an initial population, and update the initial population using a hybrid optimization algorithm based on a multi-objective constraint model, obtaining the optimal individual solution in the initial population as the Pareto optimal solution set.

[0017] The third objective of this invention is to provide a storage medium that can effectively handle the heterogeneous search space and complex constraints of MTOP, dynamically and adaptively adjust the shrinkage speed, and improve search efficiency and quality. This addresses the problems of slow convergence speed, easy getting trapped in local optima, and uneven distribution of Pareto fronts in existing technologies when solving multi-objective optimization problems, and enables the rapid acquisition of high-quality multimodal transport optimization path schemes.

[0018] The present invention provides a third basic solution: a storage medium including a content extraction program, wherein when the content extraction program is executed by a processor, it implements the steps of the above-mentioned multimodal transport delivery route optimization method.

[0019] Beneficial effects: This scheme uses the NSGA-II algorithm as the environment selection framework for multi-objective optimization. It constructs a multi-objective constraint model for iterative optimization to obtain the optimal solution, preserving its excellent distribution and convergence throughout the solution process. On this basis, this scheme introduces a hybrid mutation operator (adaptive PSO-CS operator) with PSO (Particle Swarm Optimization) and CS (Cuckoo Search) characteristics to replace the crossover and mutation operations in the traditional genetic algorithm. It mutates the individual solutions in the parent and child populations to generate the offspring population. The offspring population and the parent and child populations are merged to update the initial population. The adaptive PSO-CS operator updates the velocity and position of individual solutions and obtains the globally optimal individual solution based on congestion, serving as a progeny population. It leverages the guided search capability of the PSO operator for global search. The adaptive PSO-CS operator, through adaptive scheduling, performs global or local CS perturbations, fine-tuning the obtained individual solutions to generate a progeny population, enhancing the individual's directional exploration and global escape capabilities in the solution space. The PSO-CS operator, based on an adaptive triggering mechanism that considers iteration progress and population fitness changes, dynamically adjusts the strength of the search strategy according to the search stage and convergence state, achieving a dynamic balance between search strength and convergence speed. This effectively handles the heterogeneous search space and complex constraints of MTOP, improving search efficiency and quality. This scheme improves optimization efficiency, enhances adaptability to complex constraints, multi-objectives, and dynamic scenarios in multimodal transport, and can adapt to dynamic changes in logistics networks in actual business, possessing higher engineering application value.

[0020] In summary, this scheme can effectively handle the heterogeneous search space and complex constraints of MTOP, dynamically and adaptively refine the shrinkage speed, and improve search efficiency and quality. It solves the problems of slow convergence speed, easy getting trapped in local optima, and uneven distribution of Pareto front in existing technologies when solving multi-objective optimization problems, and quickly obtains high-quality multimodal transport optimization path schemes. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating an embodiment of a multimodal transport delivery route optimization method according to the present invention; Figure 2 This is a flowchart illustrating an embodiment of a multimodal transport delivery route optimization method according to the present invention; Figure 3 This is a network structure diagram in an embodiment of a multimodal transport delivery route optimization method of the present invention; Figure 4 This is a comparative experimental result diagram of an embodiment of the multimodal transport delivery route optimization method of the present invention. Detailed Implementation

[0022] The following detailed description illustrates the specific implementation methods: Example 1 This embodiment is basically as shown in the appendix. Figure 1 As shown, a multimodal transport delivery route optimization method is provided, including: S1. Construct a multi-objective constraint model for multimodal transport delivery routes; Under the premise that the cargo transportation capacity of each node meets the transportation requirements, a multi-objective constraint model is constructed, including: multi-objective functions and constraint conditions; The multi-objective functions include: transportation cost objective function. and transportation time objective function : , ; in A set of nodes; For the set of transport arcs (distances); For transportation modes; , , For nodes, , , ∈ ; , These are the starting and ending points for cargo transportation. , ∈N; For nodes To the node arc segment, , ∈ ; , For the mode of transport designation, , ∈M; , For nodes The set of preceding and following nodes, , ⊆ ; For connecting nodes A collection of transportation methods ⊆ ; For arc The collection of transportation methods on the Internet ⊆ ; Constraints include: The first constraint is used to restrict the relationships between nodes and ensure the effective flow of goods: ; The second constraint is used to restrict the transit nodes. arrive Choose at most one mode of transport: ; The third constraint is used to restrict transit nodes. arrive Choose at most one transit method: ; The fourth constraint restricts transit nodes from receiving and transferring goods using the same mode of transport: ; The fifth constraint is used to restrict the inflow and outflow of goods at transit nodes to be the same: .

[0023] S2. Obtain basic data on cargo transportation, generate individual solutions, form an initial population, and update the initial population using a hybrid optimization algorithm based on a multi-objective constraint model. Obtain the optimal individual solution in the initial population as the Pareto optimal solution set. Specifically, in S201, the number of iterations is initialized to t=1, and a heuristic random walk is used to generate an initial population containing several initial individual schemes; The individual scheme is a scheme for encoding basic freight transportation data; the basic freight transportation data obtained includes: freight transportation node distance, unit freight transportation cost and carbon emissions, speed of each transportation mode, etc. This embodiment employs a node-transportation dual-layer encoding method. Each individual scheme contains two chromosome encoding methods. Example of individual scheme encoding: [Path sequence: [A,B,C,D,E,F], Mode sequence: [1,3,2,1,4]], where 1, 2, and 3 represent highway, railway, and waterway respectively. The path is interpreted as A—B (highway), B—F (waterway), F→I (railway), I→J (highway), and E→F ​​(waterway).

[0024] Specifically, in the transportation network topology, starting from the starting node of each demand, a path sequence is generated by selecting the next node from the set of adjacent nodes of the current node through a random walk strategy, leading to the destination node, forming an individual scheme. Individual schemes with unreasonable or duplicate paths are structurally corrected to form an initial population. In this embodiment, local repair is used to correct the structure of the initial individuals. The random walk strategy involves calculating the distances between the current node and its neighbors. Select from the set of adjacent nodes The largest adjacent node is the next node, and so on up to the destination node; where , These are empirical weighting coefficients. For the current node The set of adjacent nodes; Furthermore, a repair operator is set to detect duplicate nodes in the path sequence and delete the duplicate nodes; At the same time, adopt Calculate the node time and cost weighted values, and prioritize the selection. Structural modifications are performed on the individual solutions that minimize the violation of constraints.

[0025] S202. Based on the multi-objective function, evaluate the fitness of individual schemes in the initial population, perform non-dominated sorting based on fitness, divide the population into levels, and calculate the crowding degree of individual schemes. Specifically, the individual schemes of the initial population Substitute them into the multi-objective function respectively: The objective function value for calculating cost and time is the fitness of the individual solution; Based on fitness, non-dominated ranking is used to divide the hierarchy, such as F1, F2, F3, etc.; where F1 > F2 > F3, and ">" indicates that the dominant hierarchy (i.e., the hierarchy) is better. according to Calculate the crowding degree of individual options; where , The level of the individual solutions Fitness of adjacent individuals to a target; , These represent the levels of the individual solutions. The maximum and minimum fitness of each target are assumed in this embodiment to be infinitely crowded at the beginning and end. S203. Using a tournament selection method, based on hierarchy and crowding, select the first preset number of parent individuals from the initial population to form a parent-child group. Specifically, a second preset number of individuals are randomly selected from the initial population; Compare the levels of the extracted individual solutions, and retain the individual solution with the better level as the parent individual. If the levels of individual solutions are the same, compare the crowding degree of the individual solutions, and retain the individual solution with the larger crowding degree value as the parent individual. Among them, retaining the individual solution with the smaller level means retaining the individual solution with the level lower than the preset level; retaining the individual solution with the larger crowding degree means retaining the individual solution with the crowding degree greater than the preset crowding degree. Determine whether the number of parent individuals is greater than or equal to the first preset number. If yes, execute S204; otherwise, execute the scheme of randomly selecting a second preset number of individuals from the initial population again.

[0026] S204. Using the adaptive PSO-CS operator, the individual schemes in the parent-child population are mutated to generate the offspring population. The offspring population and the parent-child population are merged to update the initial population. Specifically, the PSO operator operation is performed: the PSO operator is used to update the velocity and position of individual schemes, and based on the crowding degree, the globally optimal individual scheme that meets the preset population size is obtained as the offspring population; In this embodiment, during the iteration process, each individual scheme updates its velocity and position according to the PSO operator's velocity vector and position vector formulas. Based on the first level of the Pareto front, the globally optimal individual scheme is selected through a roulette wheel selection method based on congestion, and the new velocity vector is obtained by substituting it into the velocity update formula. Substituting into the position update formula yields the new position vector formula. The new position vector formula serves as the current candidate solution and participates in subsequent fitness calculations and non-dominated sorting.

[0027] Adaptive scheduling decisions are made, the CS operator is controlled, global or local CS perturbations are performed, and fine-tuning is carried out on the obtained individual schemes to generate offspring populations. In this embodiment, based on the current iteration number and preset iteration threshold Calculate the global disturbance trigger probability and the probability of triggering local disturbances : , ; During the iteration process, corresponding perturbations are dynamically executed. In the early stage of the search, global perturbations dominate, while in the later stage, local perturbations are strengthened, so as to achieve a dynamic balance between global and local search capabilities. Generate random numbers Combined with the global disturbance trigger probability and the probability of triggering local disturbances Perform adaptive scheduling judgment, if Less than Then, a global CS perturbation is performed; in this embodiment, it is a global perturbation for Lévy flight. ,like Less than Then, perform local perturbation of CS and randomly select two better individuals for fine-tuning. ; Right now: ; Merge the parent generation and all child generation populations to update the initial population; S205, determine whether the current iteration meets the termination condition. If yes, output the individual solutions of the first level F1 of the initial population as the Pareto optimal solution set. If no, increase the number of iterations and execute S202 again. Compared to existing technologies, this scheme uses the NSGA-II algorithm as the environment selection framework for multi-objective optimization. Through a non-dominated sorting mechanism and a crowding-based diversity maintenance mechanism, it ensures that the distribution and convergence are always maintained during the solution process. On this basis, a hybrid mutation operator with the characteristics of PSO (Particle Swarm Optimization) and CS (Cuckoo Search) is introduced to replace the crossover and mutation operations in the traditional genetic algorithm. It combines the guided search capability of PSO and the Lévy flight and jump characteristics of CS, so that while inheriting the stable selection capability of NSGA-II, this scheme enhances the directional exploration and global jump capability of individuals in the solution space. To address the common problem of traditional optimization algorithms struggling to effectively balance global exploration and local exploitation, this solution further designs the PSO-CS operator. Based on an adaptive triggering mechanism that adapts to changes in iteration progress and population fitness, it can dynamically adjust the strength of the search strategy according to the search stage and convergence state, thereby achieving a dynamic balance between search strength and convergence speed. This solution improves the efficiency of the algorithm, enhances its adaptability to complex constraints, multiple objectives, and dynamic scenarios in multimodal transport, and can adapt to the dynamic changes in logistics networks in actual business, thus possessing higher engineering application value.

[0028] This solution is applied to a multimodal transport delivery route example and compared with existing technologies such as NSGA-II, MOPSO, and MOPSO-CS. The convergence, solution diversity, and solution distribution uniformity are compared. This embodiment compares this solution with NSGA-II. The effectiveness of the model and algorithm is verified by selecting existing multimodal transport cases. A shipment of 20 tons of goods is transported from point A to point J, with the specific network as follows: Figure 3 As shown in Table 1, the distances between nodes are shown in Table 2 and Table 3, and the data on transportation costs and carbon emissions are shown in Table 3.

[0029] Table 1: Distances between nodes Node pairs highway railway waterway A—B 394 303 376 A—C 352 669 421 A→D 432 628 557 B—E 1016 1109 1185 B—F 1107 1173 846 C—E 874 1194 — C—F 894 972 — D—E 822 795 1000 D—F 554 577 665 E-G 242 515 237 E—H 717 — 670 E—I 580 456 554 F—G 481 996 574 F—H 960 — 959 F—I 823 581 837 G—J 439 535 524 H—J 124 — 198 I—J 136 147 159 Table 2: Average transport speed, unit transport cost, and unit carbon emissions for each mode of transport Transportation methods Average transport speed (km / h) Unit transportation cost (yuan / t.km) Carbon emissions per unit (kg / t.km) highway 80 0.316 0.796 railway 60 0.270 0.028 waterway 40 0.112 0.049 Table 3: Unit Transit Costs, Transit Time, and Transit Carbon Emissions for Each Mode of Transportation Transportation mode conversion Unit transshipment cost (yuan / t) Unit transit time (h) Carbon emissions per unit of transit (kg / t) Public (railway) to railway (public) 3.090 0.350 1.560 Railway (waterway) to waterway (railway) 26.620 0.450 3.120 Waterway to highway 5.230 0.600 6.000 Specific results are as follows Figure 4As shown, the solid line represents this scheme, and the dashed line represents NSGA-II. This scheme converges faster than NSGA-II and can find a scheme with a lower total transportation cost. This scheme improves the efficiency of the algorithm, enhances its adaptability to complex constraints, multiple objectives, and dynamic scenarios in multimodal transport, and can adapt to the dynamic changes of logistics networks in actual business, thus having higher engineering application value.

[0030] This embodiment also provides a multimodal transport delivery route optimization system for executing the above-described multimodal transport delivery route optimization method, including: The data acquisition module is used to acquire basic data on cargo transportation. The model building module is used to construct multi-objective constrained models of multimodal transport delivery routes. The collaborative analysis module is used to generate individual solutions based on basic cargo transportation data, form an initial population, and update the initial population using a hybrid optimization algorithm based on a multi-objective constraint model, and obtain the optimal individual solution in the initial population as the Pareto optimal solution set. The results output module is used to output the Pareto optimal solution set; The control interaction module is used to control multimodal transport delivery based on the Pareto optimal solution set.

[0031] This embodiment also provides a storage medium, which includes a content extraction program. When the content extraction program is executed by a processor, it implements the steps of the multimodal transport delivery route optimization method as described in any one of claims 1 to 8.

[0032] Those skilled in the art will understand that implementing all or part of the processes in the above-described content extraction method can be accomplished by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the various embodiments of content extraction described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0033] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for optimizing multimodal transport delivery routes, characterized in that, include: S1. Construct a multi-objective constraint model for multimodal transport delivery routes; S2. Obtain basic data on cargo transportation, generate individual solutions, form an initial population, and update the initial population using a hybrid optimization algorithm based on a multi-objective constraint model. Obtain the optimal individual solution in the initial population as the Pareto optimal solution set. The hybrid optimization algorithm uses the adaptive PSO-CS operator to mutate the individual schemes in the parent-child population, generate the offspring population, merge the offspring population and the parent-child population, and update the initial population. The PSO operator in the adaptive PSO-CS operator updates the velocity and position of individual solutions and obtains the globally optimal individual solution based on the crowding degree, which is then used as the offspring population. In the adaptive PSO-CS operator, the CS operator makes adaptive scheduling decisions, performs global or local CS perturbations, fine-tunes the acquired individual schemes, and generates offspring populations.

2. The multimodal transport delivery route optimization method according to claim 1, characterized in that, The multi-objective functions in the multi-objective constraint model include: the transportation cost objective function. and transportation time objective function : ， ； in A set of nodes; For the set of transport arc segments; For transportation modes; , , For nodes, , , ∈ ; , These are the starting and ending points for cargo transportation. , ∈N; For nodes To the node arc segment, , ∈ ; , For the mode of transport designation, , ∈M; , For nodes The set of preceding and following nodes, , ⊆ ; For connecting nodes A collection of transportation methods ⊆ ; For arc The collection of transportation methods on the Internet ⊆ .

3. The multimodal transport delivery route optimization method according to claim 2, characterized in that, The constraints in the multi-objective constraint model include: The first constraint is used to restrict the cargo relationship between nodes: ； The second constraint is used to restrict the transit nodes. arrive Choose at most one mode of transport: ； The third constraint is used to restrict transit nodes. arrive Choose at most one transit method: ； The fourth constraint restricts transit nodes from receiving and transferring goods using the same mode of transport: ； The fifth constraint is used to restrict the inflow and outflow of goods at transit nodes to be the same: 。 4. The multimodal transport delivery route optimization method according to claim 1, characterized in that, S2 includes: S201. Initialize the number of iterations t=1, and use heuristic random walk to generate an initial population containing several initial individual schemes; S202. Based on the multi-objective function, evaluate the fitness of individual schemes in the initial population, perform non-dominated sorting based on fitness, divide the population into levels, and calculate the crowding of individual schemes. S303. Using a tournament selection method, based on hierarchy and crowding, select the first preset number of parent individuals from the initial population to form a parent-child group; S204. Using the adaptive PSO-CS operator, the individual schemes in the parent-child population are mutated to generate the offspring population. The offspring population and the parent-child population are merged to update the initial population. S205, determine whether the current iteration meets the termination condition. If yes, output the individual solutions of the first level F1 of the initial population as the Pareto optimal solution set. If no, increase the iteration number t=t+1 and then execute S202.

5. The multimodal transport delivery route optimization method according to claim 4, characterized in that, S202 includes: according to Calculate the crowding degree of individual options; where , The level of the individual solutions Fitness of adjacent individuals to a target; , These represent the levels of the individual solutions. The maximum and minimum fitness of each objective.

6. The multimodal transport delivery route optimization method according to claim 5, characterized in that, S203 includes: a scheme to randomly select a second preset number of individuals from the initial population; Compare the levels of the extracted individual solutions, and retain the individual solution with the better level as the parent individual. If the levels of the individual solutions are the same, compare the crowding of the individual solutions, and retain the individual solution with the greater crowding as the parent individual. Among them, retaining the individual solution with the smaller level means retaining the individual solution with the level lower than the preset level; retaining the individual solution with the greater crowding means retaining the individual solution with the greater crowding than the preset crowding. Determine whether the number of parent individuals is greater than or equal to the first preset number. If yes, execute S204; otherwise, execute the scheme of randomly selecting a second preset number of individuals from the initial population again.

7. The multimodal transport delivery route optimization method according to claim 6, characterized in that, The operations of the PSO operator include: Each individual scheme updates its velocity and position according to the PSO operator's velocity and position vector formulas. Based on the first level of the Pareto front, the globally optimal individual scheme is selected using a roulette wheel selection method based on congestion, and the new velocity vector is obtained by substituting it into the velocity update formula. Substituting into the position update formula yields the new position vector formula. The new position vector formula is used as the current candidate solution.

8. The multimodal transport delivery route optimization method according to claim 7, characterized in that, The adaptive scheduling decision includes: Based on the current iteration number and preset iteration threshold Calculate the global disturbance trigger probability and the probability of triggering local disturbances : ， ； Generate random numbers Combined with the global disturbance trigger probability and the probability of triggering local disturbances Perform adaptive scheduling judgment, if Less than Then perform CS global perturbation and Lévy flight global perturbation. ,like Less than Then, perform local perturbation of CS and randomly select two better individuals for fine-tuning. .

9. A multimodal transport delivery route optimization system, characterized in that, For performing the multimodal transport delivery route optimization method as described in any one of claims 1-8, comprising: The data acquisition module is used to acquire basic data on cargo transportation. The model building module is used to construct multi-objective constrained models of multimodal transport delivery routes. The collaborative analysis module is used to generate individual solutions based on basic cargo transportation data, form an initial population, and update the initial population using a hybrid optimization algorithm based on a multi-objective constraint model, obtaining the optimal individual solution in the initial population as the Pareto optimal solution set.

10. A storage medium, characterized in that, The storage medium includes a content extraction program, which, when executed by a processor, implements the steps of a multimodal transport delivery route optimization method as described in any one of claims 1 to 8.

Images