A multi-objective robust optimization method and system for multimodal logistics resource scheduling and path planning

Abstract

The application belongs to the technical field of logistics resource scheduling, and specifically discloses a multi-target robust optimization multimodal transport logistics resource scheduling and path planning method and system. The application solves the defects of single-point of the existing distributed scheduling scheme, solves the problems of disconnection between the scheduling scheme and actual transport capacity demand and response delay, solves the problems of low efficiency of transport capacity matching and single optimization target, solves the problem of multi-target imbalance of only paying attention to cost and ignoring time efficiency and reliability, and adapts to the delivery demand of high-priority goods. Through non-robust verification of the scheduling scheme, the problem of long time consumption of global recalculation and repair under sudden conditions is solved, and the continuity of the supply chain is ensured.

Description

Technical Field

[0001] This invention belongs to the field of logistics resource scheduling technology, specifically relating to a multi-objective robust optimization method and system for multimodal transport logistics resource scheduling and route planning. Background Technology

[0002] With the increasing complexity of global supply chain networks and the accelerated low-carbon transformation, multimodal transport, as a core logistics mode connecting road, rail, water, and air transport, has become a key path to improve transportation efficiency and reduce carbon emissions. Current logistics systems generally rely on centralized scheduling platforms for resource allocation and route planning. Their underlying architecture, built on static contracts and manual coordination, struggles to cope with dynamic challenges such as sudden capacity fluctuations, cross-border compliance differences, and a lack of carbon emission transparency. Especially in scenarios involving multi-entity collaboration, data silos lead to delays in capacity matching, and the disconnect between cost and sustainability goals causes decision-making imbalances, severely restricting the resilience and evolutionary capabilities of multimodal transport systems.

[0003] The existing technologies for scheduling and route optimization of multimodal transport logistics resources can basically meet the needs of use, but there are still some shortcomings: (1) The capacity scheduling still relies on manual bidding or fixed rule matching, and cannot generate Pareto optimal solutions based on real-time supply and demand, which results in high empty running rate of transport units and difficulty in compressing carbon emission redundancy.

[0004] (2) Route planning and resource matching often focus on a single objective and ignore key requirements such as transportation time and route reliability, resulting in high-priority goods often being affected by timeliness or route failure.

[0005] (3) There is a lack of robust verification mechanism for scheduling schemes. After the scheme is generated, the feasibility is judged only by static calculation without considering the disturbance risk in actual operation. Once an emergency occurs, the global calculation needs to be re-executed, which takes a long time and is prone to causing cargo backlog, making it difficult to ensure the continuity of the supply chain. Summary of the Invention

[0006] In view of this, in order to solve the problems mentioned in the background technology, a multi-objective robust optimization method and system for multimodal transport logistics resource scheduling and route planning is proposed.

[0007] The objective of this invention can be achieved through the following technical solution: The first aspect of this invention provides a multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning, including: S1, resource and cargo data acquisition: real-time acquisition of resource status data of each transport node in multimodal transport logistics, and synchronous acquisition of cargo attribute data to be scheduled.

[0008] S2. Construction of Multi-Dimensional Matching Matrix: Based on resource status data and cargo attribute data, construct a multi-dimensional resource-demand matching matrix with transportation units as rows and cargo batches as columns.

[0009] S3. Hierarchical constraint filtering: Perform constraint filtering on the multi-dimensional resource-demand matching matrix to output a feasible matching subset of transport units and cargo batches in multimodal logistics.

[0010] S4. Static candidate scheduling scheme generation: An improved discrete multi-objective particle swarm optimization algorithm is used to perform global optimization on the feasible matching subset to generate a candidate matching subset of transport unit-cargo batch in multimodal transport logistics.

[0011] S5. Candidate scheduling scheme correction: Construct a dynamic scheduling decision model to monitor disturbance signals in the multimodal transport logistics process in real time, and initiate local re-optimization operation when the disturbance signal triggering condition is met.

[0012] S6. Robustness verification of the scheme: Import the modified candidate matching subset into the digital twin simulation platform, calculate the robustness score based on the failure probability and average delay time calculated during the Monte Carlo simulation, and select the scheme with the highest score as the optimal scheduling scheme.

[0013] The second aspect of the present invention provides a multimodal transport logistics resource scheduling and route optimization system, comprising: a resource and cargo data acquisition module, a multi-dimensional matching matrix construction module, a hierarchical constraint filtering module, a static candidate scheduling scheme generation module, a candidate scheduling scheme correction module, and a scheme robustness verification module.

[0014] The resource and cargo data acquisition module is connected to the multi-dimensional matching matrix construction module, the multi-dimensional matching matrix construction module is connected to the hierarchical constraint filtering module, the hierarchical constraint filtering module is connected to the static candidate scheduling scheme generation module, the static candidate scheduling scheme generation module is connected to the candidate scheduling scheme correction module, and the candidate scheduling scheme correction module is connected to the scheme robustness verification module.

[0015] The resource and cargo data acquisition module acquires real-time resource status data of each transportation node in multimodal logistics and simultaneously acquires the attribute data of cargo to be dispatched.

[0016] The multi-dimensional matching matrix construction module constructs a multi-dimensional resource-demand matching matrix based on resource status data and cargo attribute data, with transportation units as rows and cargo batches as columns.

[0017] The hierarchical constraint filtering module performs constraint filtering processing on the multi-dimensional resource-demand matching matrix and outputs a feasible matching subset of transport units and cargo batches in multimodal logistics.

[0018] The static candidate scheduling scheme generation module uses an improved discrete multi-objective particle swarm optimization algorithm to perform global optimization on the feasible matching subset, generating a candidate matching subset of transport unit-cargo batch in multimodal transport logistics.

[0019] The candidate scheduling scheme correction module constructs a dynamic scheduling decision model, monitors disturbance signals in the multimodal transport logistics process in real time, and initiates local re-optimization operations when the disturbance signal triggering conditions are met.

[0020] The robustness verification module imports the modified candidate matching subset into the digital twin simulation platform, calculates the robustness score based on the failure probability and average delay time calculated during the Monte Carlo simulation, and selects the scheme with the highest score as the optimal scheduling scheme.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention solves the single-point defects of the existing distributed scheduling scheme by constructing a systematic architecture from dynamic aggregation of transportation capacity to multi-objective optimization of paths and then to dynamic fault repair, and solves the problems of scheduling scheme being out of touch with actual transportation capacity demand and response delay.

[0022] 2. This invention uses a multi-objective particle swarm optimization algorithm to globally optimize feasible matching subsets, which solves the problems of inefficient capacity matching and singular optimization objectives. It also solves the multi-objective imbalance problem of focusing only on cost and ignoring timeliness and reliability, and is suitable for the delivery needs of high-priority goods.

[0023] 3. This invention solves the problem of needing global recalculation and time-consuming repair in case of emergencies by verifying the non-robustness of the scheduling scheme, thus ensuring the continuity of the supply chain. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a schematic diagram illustrating the implementation steps of the method of the present invention.

[0026] Figure 2 This is a schematic diagram of the system module connections of the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and 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.

[0028] Example 1

[0029] Please see Figure 1 As shown, the present invention provides a multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning. The specific steps are as follows: S1, Resource and cargo data acquisition: real-time acquisition of resource status data of each transport node in multimodal transport logistics, and synchronous acquisition of cargo attribute data to be scheduled.

[0030] In one feasible embodiment of the present invention, the specific content of step S1 includes: collecting resource status data of each transportation node in multimodal transport logistics through Internet of Things (IoT) sensing terminals deployed on transportation vehicles and hub facilities. The sensing terminals include vehicle positioning and status monitoring modules, rail vehicle formation identification modules, port crane operation status sensors, air cargo tracking tags, and warehouse rack weight and temperature and humidity sensor arrays.

[0031] Data on the attributes of goods to be dispatched is obtained by connecting to enterprise resource planning platforms, order management platforms, and customer relationship management platforms.

[0032] In one specific example, the resource status data includes the available capacity of road transport units, the loading and unloading capacity of railway marshalling yards, the occupancy status of port berths, the flight schedules of air cargo hubs, and the storage capacity and operational efficiency of each transshipment hub.

[0033] The cargo attribute data to be dispatched includes cargo volume, weight, category, temperature control requirements, time constraints, priority level, and geographical coordinates.

[0034] S2. Construction of Multi-Dimensional Matching Matrix: Based on resource status data and cargo attribute data, construct a multi-dimensional resource-demand matching matrix with transportation units as rows and cargo batches as columns.

[0035] In one feasible embodiment of the present invention, the specific construction process of the multi-dimensional resource-demand matching matrix includes: defining the row dimension of the multi-dimensional resource-demand matching matrix as transportation unit and the column dimension as cargo batch, and denoting the expression of each transportation unit-cargo batch matching pair in the matrix as follows: And the quantitative evaluation value of each transport unit-cargo batch matching pair is recorded as , representing the The first transport unit and the first the quantitative evaluation value of the matching relationship of the cargo batches, that is, the matrix elements of the multi-dimensional resource-demand matching matrix are quantitative evaluation values.

[0036] It should be noted that the transportation units as the row dimension only include currently available transportation units, that is, units whose equipment status is normal or idle and not fully occupied, excluding units that are fully loaded, under maintenance, or assigned with fixed tasks; each row uses the transportation unit ID as the unique identifier, and the row header needs to be attached with a key attribute summary, for example, Truck-Yu A12345, with an available load capacity of 20 tons, refrigerated, Beijing Shanghai, to facilitate quick identification.

[0037] If the status of a transportation unit changes, for example, from idle to occupied, it needs to be removed from the matrix in real time or marked as unavailable.

[0038] The cargo batches as the column dimension only include cargo batches to be scheduled, that is, cargo batches that have not been assigned a transportation unit and meet the earliest departure time condition; each column uses the cargo batch ID as the unique identifier, and the column header is attached with a key attribute summary, for example, B20231001-001, with an available load capacity of 15 tons, requiring refrigeration, Beijing Shanghai, to be delivered no later than October 5th.

[0039] If a cargo batch has been assigned a transportation unit or exceeds the latest departure time, it needs to be removed from the matrix in real time.

[0040] The evaluation value of the physical feasibility dimension, the evaluation value of the timeliness matching dimension, and the evaluation value of the cost adaptability dimension of each transportation unit-cargo batch matching pair are calculated respectively through the set corresponding calculation rules, and the weighted sum thereof is obtained to get the quantitative evaluation value of the matching relationship of each transportation unit-cargo batch matching pair.

[0041] It should be noted that the specific formula of the corresponding calculation rule set for the evaluation value of the physical feasibility dimension is: (1) Load capacity matching degree: , wherein if the available load capacity of the transportation unit ≥ the weight of the cargo, the value is 1, which means the transportation unit is completely matched with the cargo batch; if the available load capacity of the transportation unit < the weight of the cargo, the value is <1.0, which means the transportation unit is partially matched with the cargo batch, and at this time, it is necessary to consider whether to split the cargo batch for cargo splitting processing; if the available load capacity of the transportation unit = 0, the value is 0, which means the transportation unit is completely mismatched with the cargo batch.

[0042] (2) Volume matching degree: , and its specific implementation manner is the same as that of the load capacity matching degree.

[0043] When both the load capacity matching degree and the volume matching degree are 1, the evaluation value of the physical feasibility dimension is recorded as 1, otherwise the evaluation value of the physical feasibility dimension is recorded as 0.

[0044] The specific formula for calculating the corresponding evaluation value of the timeliness matching dimension is as follows: If A value of ≥1 indicates that the transport unit can deliver the goods on time. The larger the value, the more sufficient the time redundancy; if If the value is less than 1, it means that the transport unit is unable to deliver the goods on time.

[0045] The specific formula for calculating the corresponding value of the cost adaptability dimension is as follows: ,in The higher the value, the more obvious the cost advantage of the transportation unit.

[0046] The quantitative evaluation values ​​of the matching relationship are stored in matrix form to construct a multi-dimensional resource-demand matching matrix.

[0047] S3. Hierarchical constraint filtering: Perform constraint filtering on the multi-dimensional resource-demand matching matrix to output a feasible matching subset of transport units and cargo batches in multimodal logistics.

[0048] In one feasible embodiment of the present invention, the constraint filtering process includes a physical feasibility constraint filtering layer, a time-sensitive hard constraint filtering layer, and a cost-flexible constraint filtering layer.

[0049] The specific process of outputting a feasible matching subset of transport units and cargo batches in multimodal transport logistics includes: extracting the transport unit-cargo batch matching pairs to be filtered from the multi-dimensional resource-demand matching matrix, passing them through a physical feasibility constraint filtering layer, and eliminating matching pairs caused by cargo size exceeding the transport unit vehicle limit, temperature control requirements not matching, cargo category belonging to the transport unit's prohibited list, or insufficient processing capacity of the transshipment node.

[0050] The physical feasibility constraint filter layer outputs each transport unit-cargo batch matching pair, which is then passed through the timeliness hard constraint filter layer to eliminate matching pairs whose estimated transport time exceeds the cargo delivery time window requirement.

[0051] The timeliness-constraint filter layer outputs each transport unit-cargo batch matching pair and passes it through the cost flexibility filter layer to remove matching items whose cost dimension evaluation value exceeds the preset threshold, thus outputting a feasible matching subset of transport units-cargo batches in multimodal logistics.

[0052] This invention addresses the single-point deficiency of existing distributed scheduling schemes by constructing a systematic architecture that integrates dynamic capacity aggregation, multi-objective path optimization, and dynamic fault repair. It also compensates for the shortcomings of technologies such as blockchain that only automate a single link, and solves the problems of scheduling schemes being out of sync with actual capacity demand and response delays.

[0053] S4. Static candidate scheduling scheme generation: An improved discrete multi-objective particle swarm optimization algorithm is used to perform global optimization on the feasible matching subset to generate a candidate matching subset of transport unit-cargo batch in multimodal transport logistics.

[0054] In one feasible embodiment of the present invention, the multiple objectives include minimizing total transportation cost, minimizing total transportation time, and maximizing transportation route reliability.

[0055] In the process of generating candidate matching subsets of transport units and cargo batches in multimodal logistics, a multi-objective optimization model needs to be constructed. Specifically, the minimization of total transport cost, the minimization of total transport time, and the maximization of transport route reliability are transformed into computable mathematical functions, which serve as the evaluation basis for particle fitness.

[0056] It should be noted that the specific operation of converting the data into a computable mathematical function is: minimizing the total transportation cost. ,in .

[0057] The trunk line transportation fee = transportation unit price × distance × cargo weight, the transshipment fee = node loading and unloading fee + warehousing fee, and the resource scheduling fee, such as the additional cost of temporary vehicle dispatching.

[0058] Minimize total transit time in, .

[0059] The trunk line transportation time is the sum of the time spent on each segment of the route, the transfer time is the node waiting time plus the loading and unloading time, and the connection time is the waiting time for transferring between different modes of transportation.

[0060] Maximize the reliability of transportation routes ,in, .

[0061] The reliability of the path nodes is like the port's failure-free probability; the reliability of the transportation unit is like the historical on-time rate; and the environmental adaptability is like the anti-interference coefficient under heavy rain.

[0062] In one feasible embodiment of the present invention, the specific process of generating a candidate matching subset of transport unit-cargo batch in multimodal transport logistics includes: A1, extracting each transport unit-cargo batch matching pair from the feasible matching subset, randomly generating a feasible path for it, and recording the combination of the matching pair and the feasible path as an algorithm particle, thereby generating a particle swarm.

[0063] A2. Create an empty external archive to store Pareto optimal solutions generated during the iteration process.

[0064] It should be noted that the Pareto optimal solution is the solution that cannot improve all objectives without worsening at least one objective, thus providing a basis for the preservation of subsequent high-quality solutions and the selection of the global optimal solution.

[0065] A3. Calculate the multi-objective function value for each particle in the particle swarm. The non-dominated solutions selected based on Pareto dominance are then used as the high-quality solutions for the current iteration.

[0066] It should be noted that the specific process of screening non-dominated solutions based on Pareto dominance is as follows: If at least one of the three objective function values ​​of a certain particle is strictly greater than the corresponding objective function value of another particle, it means that the particle dominates the other particle, and the other particle is eliminated, while all non-dominated solutions are retained as the high-quality solutions for the current iteration.

[0067] For example, if particle A's ≤particle B And particle A's ≤particle B And particle A's ≥ Particle B If A dominates B, then B is eliminated; all non-dominated solutions are retained as high-quality solutions for the current iteration.

[0068] A4. Adaptive inertial weights based on non-linear decreasing iteration counts are used to balance the algorithm's global exploration and local development capabilities.

[0069] It should be noted that the specific formula for the adaptive inertia weight based on the nonlinear decrease of the number of iterations is as follows: ,in This represents the current iteration number. The maximum number of iterations, In other words, use high weights in the early stages to enhance global exploration capabilities. In other words, use smaller weights in later stages to enhance local development capabilities. It is a nonlinear coefficient, usually taken as 2, to control the deceleration rate.

[0070] A5. Based on the updated discrete velocity, i.e., the operation probability, adjust the particle position through discrete operators, i.e., the scheduling scheme.

[0071] It should be noted that after the scheduling scheme is adjusted, it is necessary to verify whether the new scheme meets the capacity constraints and path connectivity requirements. If the requirements are not met, the rollback operation is performed to ensure that the new particle is still a feasible solution.

[0072] A6. Add all non-dominated solutions of the current iteration to the external archive. If the size of the external archive exceeds the preset limit, such as 200 solutions, delete redundant solutions and randomly select a solution from the external archive as the global optimal solution to avoid the algorithm from converging to a local optimum too early.

[0073] It should be noted that the deletion of redundant solutions is specifically achieved through crowding distance filtering, that is, retaining solutions with high crowding and more even distribution.

[0074] A7. When the iteration stopping condition is met, the algorithm stops and selects several representative Pareto optimal solutions from the external archive as several static candidate scheduling schemes, which are denoted as candidate matching subsets.

[0075] It should be noted that the iteration stopping conditions specifically include: reaching the maximum number of iterations, such as 200; and the Pareto optimal solution in the external archive not being updated for a number of consecutive iterations, such as 50.

[0076] It should also be noted that each of the static candidate scheduling schemes must include: a matching table of transportation units and cargo batches; a detailed path node sequence for each transportation unit; specific values ​​of the three objective functions; and key parameters such as transportation unit utilization rate and transit node load rate.

[0077] This invention uses a multi-objective particle swarm optimization algorithm to globally optimize feasible matching subsets, solving the problems of inefficient capacity matching and singular optimization objectives. It also addresses the imbalance of multiple objectives that focus only on cost while neglecting timeliness and reliability, thus adapting to the delivery needs of high-priority goods.

[0078] S5. Candidate scheduling scheme correction: Construct a dynamic scheduling decision model to monitor disturbance signals in the multimodal transport logistics process in real time, and initiate local re-optimization operation when the disturbance signal triggering condition is met.

[0079] In one feasible embodiment of the present invention, the specific content of the disturbance signal triggering condition includes: transportation interruption causing the transportation unit to be completely unable to perform its task.

[0080] It should be explained that the transportation interruption refers to the loss of all functions of transportation units such as trucks, ships, and aircraft due to irreversible events such as mechanical failure, accident damage, or sudden policy control; the complete inability to perform the task refers to the inability to transport all batches of goods allocated to the transportation unit as originally planned, excluding scenarios where the fault can be temporarily repaired or some functions are damaged.

[0081] Severe delays cause high-priority goods to exceed the latest delivery window. The severe delays refer to time losses caused by disturbances that exceed the original plan's buffer redundancy and cannot be compensated for by conventional adjustments such as accelerating travel or reducing transit time.

[0082] It should be explained that the above content excludes situations where the original plan itself already has a timeout risk, and only addresses timeout risks newly added by disturbances.

[0083] For example, the high-priority goods include, but are not limited to, time-sensitive goods such as pharmaceutical logistics, emergency relief supplies, and fresh and perishable products.

[0084] The disturbance affects a number of shipments exceeding a preset threshold or involves key transit points.

[0085] For example, the preset quantity threshold can be 5 batches.

[0086] It should be explained that the affected cargo batches include the total number of batches directly and indirectly related to the disturbance; the key transit node refers to the core connecting hub of multimodal transport. Disturbance to such nodes will trigger systemic link risks, which can be triggered without counting the number of batches.

[0087] In one feasible embodiment of the present invention, the specific content of the local re-optimization operation includes: locating the affected transportation segment and the upstream and downstream segments that are spatiotemporally related to it based on the disturbance signal, and forming a sub-network to be re-optimized.

[0088] It should be noted that the location of the affected transport segment is as follows: the directly affected transport segment is determined based on the matching of the disturbance location with the real-time trajectory of the transport unit.

[0089] For example, (1) Truck A in Beijing If a fault occurs in transport segment ID: R101 in the Tianjin section, then the directly affected segment is R101.

[0090] (2) Shanghai Port node ID: N5 is closed due to typhoon. All transport segments that originate from, terminate at, or transfer at N5, such as R202 Ningbo. Shanghai, R203 Shanghai All of Qingdao's sections are directly affected.

[0091] The specific location of the upstream and downstream segments with spatiotemporal correlation is as follows: that is, through spatiotemporal correlation analysis: identify the indirectly affected upstream and downstream segments through time connection and spatial dependence. (1) Time correlation: if the start time of transportation segment B depends on the end time of transportation segment A, such as if the goods of A need to be transferred to B, then the delay of A will cause B to be affected; (2) Spatial correlation: transportation segments that share key resources, such as the same transportation unit and the same transfer node, are correlated.

[0092] For example, (1) transport segment A, such as Beijing Goods arriving in Tianjin at 10:00 AM require transfer to transport segment B, such as Tianjin. If train A departs from Shanghai at 10:30 and arrives at 11:00 due to disruptions, then train B will be affected.

[0093] (2) Transport unit C is also responsible for transport section R301, such as Jinan. Zhengzhou and R302 Zhengzhou If R301 is disrupted due to disturbances in Wuhan, R302 will also be affected, meaning there will be no goods to transport.

[0094] It should be noted that by constructing a correlation map of transportation segment-node-transportation unit, the correlation segment can be located within 10 seconds.

[0095] A fast repair algorithm based on arc consistency verification constraint propagation is adopted to replan the path and resource matching relationship of high-priority goods in the sub-network to be re-optimized.

[0096] It should be noted that the constraint propagation fast repair algorithm of the arc consistency verification prioritizes ensuring the timeliness of high-priority goods and minimizes the scope of solution adjustments as a secondary constraint. The arc consistency verification ensures that the replanning solution meets all constraints, such as the transportation unit adapting to the physical requirements of the goods, the transit node meeting the capacity limit, and the path meeting the timeliness requirements, thus avoiding infeasible solutions. Through the constraint propagation mechanism, the timeliness constraints of high-priority goods are passed layer by layer to the transportation unit matching and path selection stages, prioritizing the locking of suitable resources and paths, while reducing unnecessary adjustments to non-high-priority goods, thereby achieving fast and accurate solution repair.

[0097] For example, the constraint propagation can specifically be as follows: if the load of transport unit A is insufficient, then the matching of transport unit A with cargo batch B is excluded, and the list of optional units for cargo batch B is updated until there are no constraint conflicts.

[0098] S6. Robustness verification of the scheme: Import the modified candidate matching subset into the digital twin simulation platform, calculate the robustness score based on the failure probability and average delay time calculated during the Monte Carlo simulation, and select the scheme with the highest score as the optimal scheduling scheme.

[0099] In one feasible embodiment of the present invention, the specific calculation process of the failure probability, average delay duration and robustness score includes: the ratio of the number of trials in which the scheduling scheme obtained in the Monte Carlo simulation cannot meet the constraints to the total number of trials is recorded as the failure probability.

[0100] It should be noted that the scheduling scheme cannot meet the constraints if any of the following conditions occur: the goods exceed the latest delivery time window, the transportation unit is completely disconnected, or the total cost exceeds the preset upper limit.

[0101] The average delay time is defined as the ratio of the total delay time of the cargo to the total number of trials accumulated during the Monte Carlo simulation.

[0102] It should be noted that the total delay time of the goods refers to the sum of the differences between the actual delivery time and the planned delivery time of all goods, and it only includes trials in which the scheduling scheme did not fail.

[0103] The robustness score is calculated by importing the failure probability and average delay duration into a standard formula.

[0104] The standard formula is as follows: The score ranges from 0 to 100. The higher the score, the stronger the robustness. The robustness score is negatively correlated with the failure probability and the average delay time. The lower the failure probability, the less likely the scheduling scheme is to collapse, and the higher the score. The shorter the average delay time, the stronger the scheme's recovery ability after being disturbed, and the higher the score.

[0105] This invention solves the problems of needing global recalculation and time-consuming repair in case of emergencies by verifying the non-robustness of the scheduling scheme, thus ensuring the continuity of the supply chain.

[0106] Example 2

[0107] Please see Figure 2 As shown, the present invention provides a multimodal transport logistics resource scheduling and route optimization system, with the following specific module distribution: resource and cargo data acquisition module, multi-dimensional matching matrix construction module, hierarchical constraint filtering module, static candidate scheduling scheme generation module, candidate scheduling scheme correction module, and scheme robustness verification module.

[0108] The resource and cargo data acquisition module is connected to the multi-dimensional matching matrix construction module, the multi-dimensional matching matrix construction module is connected to the hierarchical constraint filtering module, the hierarchical constraint filtering module is connected to the static candidate scheduling scheme generation module, the static candidate scheduling scheme generation module is connected to the candidate scheduling scheme correction module, and the candidate scheduling scheme correction module is connected to the scheme robustness verification module.

[0109] The resource and cargo data acquisition module acquires real-time resource status data of each transportation node in multimodal logistics and simultaneously acquires the attribute data of cargo to be dispatched.

[0110] The multi-dimensional matching matrix construction module constructs a multi-dimensional resource-demand matching matrix based on resource status data and cargo attribute data, with transportation units as rows and cargo batches as columns.

[0111] The hierarchical constraint filtering module performs constraint filtering processing on the multi-dimensional resource-demand matching matrix and outputs a feasible matching subset of transport units and cargo batches in multimodal logistics.

[0112] The static candidate scheduling scheme generation module uses an improved discrete multi-objective particle swarm optimization algorithm to perform global optimization on the feasible matching subset, generating a candidate matching subset of transport unit-cargo batch in multimodal transport logistics.

[0113] The candidate scheduling scheme correction module constructs a dynamic scheduling decision model, monitors disturbance signals in the multimodal transport logistics process in real time, and initiates local re-optimization operations when the disturbance signal triggering conditions are met.

[0114] The robustness verification module imports the modified candidate matching subset into the digital twin simulation platform, calculates the robustness score based on the failure probability and average delay time calculated during the Monte Carlo simulation, and selects the scheme with the highest score as the optimal scheduling scheme.

[0115] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0116] Those skilled in the art will recognize that the algorithmic steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0117] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0118] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0119] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning, characterized in that: include: S1. Resource and Cargo Data Acquisition: Real-time acquisition of resource status data of each transportation node in multimodal logistics, and synchronous acquisition of cargo attribute data to be dispatched; S2. Construction of multi-dimensional matching matrix: Based on resource status data and cargo attribute data, construct a multi-dimensional resource-demand matching matrix with transportation units as rows and cargo batches as columns; S3. Hierarchical constraint filtering: Perform constraint filtering on the multi-dimensional resource-demand matching matrix to output a feasible matching subset of transport units-cargo batches in multimodal logistics. S4. Static candidate scheduling scheme generation: The improved discrete multi-objective particle swarm optimization algorithm is used to perform global optimization on the feasible matching subset to generate a candidate matching subset of transport unit-cargo batch in multimodal transport logistics. S5. Candidate scheduling scheme correction: Construct a dynamic scheduling decision model to monitor disturbance signals in the multimodal transport logistics process in real time, and start local re-optimization operation when the disturbance signal triggering condition is met; S6. Robustness verification of the scheme: Import the modified candidate matching subset into the digital twin simulation platform, calculate the robustness score based on the failure probability and average delay time calculated during the Monte Carlo simulation, and select the scheme with the highest score as the optimal scheduling scheme.

2. The multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning according to claim 1, characterized in that: The specific content of step S1 includes: Resource status data of each transportation node in multimodal transport logistics is collected by IoT sensing terminals deployed on transportation vehicles and hub facilities. The sensing terminals include vehicle positioning and status monitoring modules, rail vehicle formation identification modules, port crane operation status sensors, air cargo tracking tags, and warehouse rack weight and temperature and humidity sensor arrays. Data on the attributes of goods to be dispatched is obtained by connecting to enterprise resource planning platforms, order management platforms, and customer relationship management platforms.

3. The multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning according to claim 1, characterized in that: The specific construction process of the multi-dimensional resource-demand matching matrix includes: The multi-dimensional resource-demand matching matrix is ​​defined with rows representing transportation units and columns representing cargo batches. The expression for each transportation unit-cargo batch matching pair in the matrix is ​​denoted as... And the quantitative evaluation value of each transport unit-cargo batch matching pair is recorded as , representing the The first transport unit and the first The quantitative evaluation value of the matching relationship of each batch of goods, that is, the matrix elements of the multi-dimensional resource-demand matching matrix are the quantitative evaluation values; The physical feasibility dimension assessment value, timeliness matching dimension assessment value, and cost adaptability dimension assessment value of each transportation unit-cargo batch matching pair are calculated by setting corresponding calculation rules, and then weighted and summed to obtain the quantitative assessment value of the matching relationship of each transportation unit-cargo batch matching pair. The quantitative evaluation values ​​of the matching relationship are stored in matrix form to construct a multi-dimensional resource-demand matching matrix.

4. The multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning according to claim 1, characterized in that: The constraint filtering process includes a physical feasibility constraint filtering layer, a time-sensitive hard constraint filtering layer, and a cost-flexible constraint filtering layer. The specific process of outputting a feasible matching subset of transport units and cargo batches in multimodal transport logistics includes: extracting each transport unit-cargo batch matching pair to be filtered from the multi-dimensional resource-demand matching matrix, passing them through a physical feasibility constraint filtering layer, and eliminating each matching pair caused by cargo size exceeding the transport unit vehicle limit, temperature control requirements not matching, cargo category belonging to the transport unit's prohibited list, or insufficient processing capacity of the transshipment node. The physical feasibility constraint filter layer outputs each transport unit-cargo batch matching pair and passes it through the timeliness hard constraint filter layer to remove each matching pair whose expected transport time exceeds the cargo delivery time window requirement. The timeliness-constraint filter layer outputs each transport unit-cargo batch matching pair and passes it through the cost flexibility filter layer to remove matching items whose cost dimension evaluation value exceeds the preset threshold, thus outputting a feasible matching subset of transport units-cargo batches in multimodal logistics.

5. The multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning according to claim 1, characterized in that: The multiple objectives include minimizing total transportation cost, minimizing total transportation time, and maximizing transportation route reliability. In the process of generating candidate matching subsets of transport units and cargo batches in multimodal logistics, a multi-objective optimization model needs to be constructed. Specifically, the minimization of total transport cost, the minimization of total transport time, and the maximization of transport route reliability are transformed into computable mathematical functions, which serve as the evaluation basis for particle fitness.

6. The multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning according to claim 5, characterized in that: The specific process for generating a candidate matching subset of transport units and cargo batches in multimodal logistics includes: A1. Extract each transport unit-cargo batch matching pair from the feasible matching subset, randomly generate a feasible path for it, and record the combination of the matching pair and the feasible path as an algorithm particle to generate a particle swarm. A2. Create an empty external archive to store Pareto optimal solutions generated during the iteration process; A3. Calculate the multi-objective function value for each particle in the particle swarm, and take the non-dominated solution selected based on the Pareto dominance relationship as the high-quality solution for the current iteration; A4. Adaptive inertial weights based on non-linear decreasing iteration count are used to balance the algorithm's global exploration and local development capabilities. A5. Based on the updated discrete velocity, i.e., the operation probability, adjust the particle position through discrete operators, i.e., the scheduling scheme. A6. Add all non-dominated solutions of the current iteration to the external archive. If the size of the external archive exceeds the preset limit, delete redundant solutions and randomly select a solution from the external archive as the global optimal solution. A7. When the iteration stopping condition is met, the algorithm stops and selects several representative Pareto optimal solutions from the external archive as several static candidate scheduling schemes, which are denoted as candidate matching subsets.

7. The multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning according to claim 1, characterized in that: The specific details of the disturbance signal triggering conditions include: The disruption of transportation rendered the transportation unit completely unable to perform its tasks; Severe delays cause high-priority goods to exceed the latest delivery window. The severe delays refer to time losses caused by disturbances that exceed the original plan's buffer redundancy and cannot be compensated for by conventional adjustments such as accelerating travel or reducing transit time. The disturbance affects a number of shipments exceeding a preset threshold or involves key transit points.

8. The multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning according to claim 7, characterized in that: The specific content of the local re-optimization operation includes: Based on the disturbance signals, the affected transportation segments and their spatiotemporally related upstream and downstream segments are located, forming a sub-network to be re-optimized; A fast repair algorithm based on arc consistency verification is adopted to replan the path and resource matching relationship of high-priority goods in the sub-network to be re-optimized.

9. The multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning according to claim 1, characterized in that: The specific calculation process for the failure probability, average delay duration, and robustness score includes: The ratio of the number of trials in which the scheduling scheme obtained during the Monte Carlo simulation fails to meet the constraints to the total number of trials is denoted as the failure probability. The average delay time is recorded as the ratio of the total delay time of the cargo to the total number of trials accumulated during the Monte Carlo simulation. The robustness score is calculated by importing the failure probability and average delay duration into a standard formula.

10. A system for executing the multi-objective robust optimization method for multimodal transport logistics resource scheduling and route planning as described in claims 1-9, characterized in that: include: The resource and cargo data acquisition module acquires real-time resource status data of each transportation node in multimodal logistics and simultaneously acquires the attribute data of cargo to be dispatched. The multi-dimensional matching matrix construction module constructs a multi-dimensional resource-demand matching matrix based on resource status data and cargo attribute data, with transportation units as rows and cargo batches as columns. The hierarchical constraint filtering module performs constraint filtering processing on the multi-dimensional resource-demand matching matrix and outputs a feasible matching subset of transport units and cargo batches in multimodal transport logistics. The static candidate scheduling scheme generation module uses an improved discrete multi-objective particle swarm optimization algorithm to perform global optimization on the feasible matching subset, generating a candidate matching subset of transport unit-cargo batch in multimodal transport logistics; The candidate scheduling scheme correction module constructs a dynamic scheduling decision model, monitors disturbance signals in the multimodal transport logistics process in real time, and initiates local re-optimization operation when the disturbance signal triggering condition is met. The robustness verification module imports the modified candidate matching subset into the digital twin simulation platform, calculates the robustness score based on the failure probability and average delay time calculated during the Monte Carlo simulation, and selects the scheme with the highest score as the optimal scheduling scheme.

Images