Optimization methods for multimodal transport scheduling of emergency supplies that are sensitive to risks

Abstract

This invention belongs to the field of emergency material multimodal transport scheduling technology, specifically involving a risk-sensitive emergency material multimodal transport scheduling optimization method, including: A. Constructing an emergency material multimodal transport network that coordinates trunk and terminal transport, mapping trunk and terminal transport to a multimodal transport network diagram, and clarifying nodes, feasible route sets, and their parameters; B. Establishing a risk-sensitive emergency material multimodal transport scheduling optimization model, based on the emergency material multimodal transport network and the acquired parameters; C. Generating model solution results and scheduling schemes, transforming the scheduling optimization model into a solvable form, generating executable scheduling schemes, and verifying the results to form a complete technical closed loop. The advantages of this invention are: significantly improving the adaptability and robustness of the emergency logistics system under extreme scenarios such as route disruptions, providing important decision support for scientific disaster relief.

Description

Technical Field

[0001] This invention belongs to the field of emergency material multimodal transport scheduling technology, specifically involving an optimization method for emergency material multimodal transport scheduling that takes into account risk sensitivity. Background Technology

[0002] In recent years, frequent natural disasters and accidents have occurred globally, placing extremely high demands on emergency material dispatch due to their suddenness and destructiveness. Currently, the demand for emergency materials exhibits complex characteristics of being "multi-category and cross-regional," making it difficult for a single mode of transportation to meet the dispatch needs in complex scenarios. For example, road transport capacity is limited, and air transport is costly. Against this backdrop, drones, as an effective supplementary means, can quickly and accurately deliver materials to remote or inaccessible areas, providing a significant capability extension to traditional transportation methods. Drones can also serve as mobile platforms for disaster awareness, providing real-time situational awareness to support delivery operations by performing disaster assessment tasks (including damage assessment and identification of critical infrastructure damage). Therefore, combining drones with traditional multimodal transport (integrating multiple modes of transportation such as air, rail, and road) has become an important path to improve the efficiency of cross-regional emergency dispatch.

[0003] However, uncertainty is prevalent in emergency dispatch scenarios, and the limitations of traditional emergency response systems become increasingly apparent as the complexity of logistics systems increases in disaster environments. Particularly in post-disaster conditions, highway networks are often severely damaged, significantly limiting ground transportation capacity. Traditional point-to-point transportation models and single-level network structures have limited dispatch efficiency when dealing with cross-regional, multi-stage transportation tasks. Especially in the connection between trunk transportation and last-mile delivery, existing research has insufficiently explored the advantages and synergistic mechanisms between different transportation modes, failing to fully leverage the system efficiency of multimodal transport. Furthermore, regarding existing optimization methods, many existing optimization models rely on precise probability distribution assumptions or overly conservative robust optimization strategies, mostly based on the assumption of risk neutrality. Risk quantification is not precise enough, ignoring the inherent risk-averse tendencies of decision-makers in emergency rescue environments. This can lead to delays in material dispatch, or even serious consequences, due to underestimating extreme risks in actual implementation of theoretically optimal solutions.

[0004] Existing Chinese patent CN202411086330.7 discloses a multimodal transport method for emergency supplies under sudden scenarios. It constructs a joint transport network for emergency supplies based on key node information, solving the problem that optimization methods for multimodal transport schemes fail to highlight the uncertainty of sudden situations and fully consider the urgency of emergency supply needs, leading to insufficient transport efficiency and excessive costs. Chinese patent CN202311798767.9 discloses a cross-regional dispatch method for emergency supplies, balancing the geographical dispersion of cross-regional emergency supply allocation with the cost of capital investment. It also incorporates key elements such as vulnerable groups in disaster-stricken areas (e.g., the number of elderly people), constructing a dynamic emergency supply dispatch model that covers the soft and hard time windows of disaster-stricken groups under multiple cycles, thereby achieving scientific and rational planning of emergency supply dispatch schemes. Chinese patent CN201910642142.0 discloses a dispatch method and system for emergency auxiliary decision-making in cross-regional integrated transportation networks. Based on emergency alarm information, it generates dispatch schemes including single and multiple transport modes and provides the optimal ranking of schemes according to dispatch objectives. Chinese patent CN202211550295.0 discloses a material delivery service scheduling and management system based on drone and vehicle collaboration. It uses algorithms to process and analyze logistics data and plans drone and vehicle collaborative transportation routes from warehouse to customer in combination with the actual traffic topology.

[0005] While existing research on multimodal transport of emergency supplies focuses on optimizing transport schemes and constructing scheduling models, it fails to consider the advantages of different modes of transport in the cross-regional emergency supply dispatch process. It also fails to refine the matching logic of transport mode advantages based on the different stages of cross-regional transport from "trunk line to terminal." Existing research on emergency logistics networks mostly focuses on "point-to-point" scheduling or single-level optimization, without constructing a hierarchical network system deeply integrated with multimodal transport, resulting in low efficiency in transport mode connection. Existing research is insufficient in addressing special scenarios at the end of cross-regional transport (such as road damage or remote disaster areas), and has not developed targeted transport mode combination solutions. In the field of drone-assisted logistics distribution, most research is limited to single-layer networks or uses "synchronous multi-mode" in multi-layer networks, while research integrating "asynchronous multi-mode" into multi-layer distribution systems and incorporating multiple transport modes such as highways, railways, large cargo planes, and drones is extremely scarce.

[0006] Existing research (DOI: https: / / doi.org / 10.1016 / j.ejor.2024.10.041, "A risk-averselatency location-routing problem with stochastic travel times") studies the risk-averselatency location-routing problem with stochastic travel times. It relies on the explicit assumption that transport times follow a known distribution (e.g., uniform, log-normal) and generates scenarios or samples based on this to optimize the route. (DOI: https: / / doi.org / 10.1016 / j.engappai.2022.105530, "Multi-period dynamic multi-objective emergencymaterial distribution model under uncertain demand") addresses the rational allocation of emergency materials under multi-period dynamic and uncertain conditions during sudden large-scale disasters. It comprehensively considers realistic factors such as the severity of the disaster, the types of materials, the diversity of transportation tools, and the risk of road availability, establishing a mixed-integer linear programming multi-objective model with fuzzy chance constraints. Triangular fuzzy numbers represent the fuzzy transport times of different modes of transport. Chinese patent CN202111325141.7 discloses a dynamic allocation method for emergency relief supplies based on the split-Brow bar optimization technique, which uses the split-Brow bar optimization method to solve the problem of uncertain emergency supply demand during the rescue cycle.

[0007] Existing research is largely limited to optimizing a single transportation mode, failing to achieve differentiated task allocation based on material characteristics and lacking sufficient characterization of the dynamic changes in post-disaster road capacity. When addressing the uncertainty of transportation time distribution, it often relies on the strong distribution assumptions of stochastic programming or conservative strategies of classical robust optimization, ignoring the risk-averse tendencies commonly found in decision-makers. Traditional risk-neutral models may not only lead to suboptimal scheduling results in real-world disaster scenarios but could even cause adverse consequences such as material delays. Therefore, it is necessary to incorporate decision-makers' risk attitudes into the modeling framework and adopt appropriate risk measurement methods to systematically reduce the uncertainty risks in the emergency dispatch process.

[0008] Therefore, in order to break through the bottlenecks of existing research, there is an urgent need for a new method for emergency material dispatch that can not only design multimodal transport networks in a refined manner, but also scientifically characterize distribution uncertainties and incorporate decision-makers' risk preferences. Summary of the Invention

[0009] In view of the above problems, the purpose of this invention is to provide a risk-sensitive emergency material multimodal transport scheduling optimization method, which is used to establish a risk-sensitive sub-bar optimization model and a collaborative transport network, as well as a corresponding intelligent scheduling method, to overcome the shortcomings of the prior art.

[0010] This invention provides a method for optimizing the multimodal transport scheduling of emergency supplies, taking into account risk sensitivity, comprising the following steps:

[0011] A. Construct an emergency material multimodal transport network that coordinates trunk lines and terminal stations, integrate multi-source data on disaster assessment, supply and demand inventory, transportation network and vehicle status, and clean and merge the data. Then, map trunk line transport and terminal transport into a multimodal transport network diagram, clarify the nodes, feasible route sets and their parameters, and provide structured input for step B.

[0012] B. Establish a risk-sensitive emergency material multimodal transport scheduling optimization model. Based on the emergency material multimodal transport network with trunk and terminal coordination constructed in step A and the obtained parameters, establish a risk-sensitive scheduling optimization model.

[0013] C. Generate model solution results and scheduling schemes. Transform the scheduling optimization model established in step B into a solvable form, generate an executable scheduling scheme, and verify the results to form a complete technical closed loop.

[0014] As a preferred embodiment of the present invention, step A further includes the following step:

[0015] A1. Data Acquisition: Information on disaster situation, material supply and demand, transportation network, available transportation vehicles and their parameters, accessibility assessment data for affected areas, and damage status of critical infrastructure are acquired through disaster site monitoring, feedback from emergency management departments, and drone aerial surveys. A disaster parameter database is established, using historical data or expert experience values ​​when information is missing. Confidential intervals / upper and lower bounds are set for missing data and constrained in the scheduling optimization model using support sets. Dynamic data updates are also implemented to dynamically update the database.

[0016] A2. Identify key transportation nodes and determine material supply points, material forwarding points, and material demand points. Material forwarding points are selected from existing emergency warehouses, assembly points around airports / train stations, or a set of alternative candidate points.

[0017] A3. Design a trunk and terminal collaborative transportation mode based on asynchronous multi-mode. Based on the geographical distribution of nodes and disaster situation, the transportation stage is divided into trunk transportation and terminal transportation. Trunk transportation includes the material supply point to the material forwarding point, by integrating high-capacity transportation methods such as railway, highway and air. Terminal transportation includes the material forwarding point to the material demand point, by designing a road plus drone delivery method.

[0018] A4. Construct a three-tiered network topology: a three-tiered emergency material multimodal transport network topology based on material supply points, material forwarding points, and material demand points.

[0019] Abstract it into a network graph when constructing a network. , Network diagram The set of nodes includes supply points, forward deployment points, and demand points for supplies. , To gather at the supply point For the collection of supplies at the forward deployment point For the aggregation of material demand points, there is a three-tiered emergency material multimodal transport network. individual supply points One material forwarding point and One point of demand for supplies; Network diagram The set of edges; A collection of modes of transportation This includes highways, railways, aviation, and drones. It is a collection of trunk transportation modes, namely, material supply points. to the material forwarding point A collection of transportation methods ,in For railway transportation, For air transport, This refers to a collection of last-mile transportation methods, i.e., a forward distribution point for goods. to the point of need for supplies A collection of transportation methods ,in For road transport, For drone transportation;

[0020] A5. Road feasibility assessment,

[0021] The roads between the material distribution point and the material demand point are located in the disaster area. Based on the traffic network information obtained from the drone assessment mission and the disaster assessment data obtained from the analysis, image recognition or data analysis methods are used to assess the degree of damage to the route and determine the post-disaster road transport capacity attenuation coefficient of the route. With traffic capacity attenuation threshold The relationship between these factors allows for the pre-determination of road transport feasibility parameters for each route. As input to the scheduling optimization model, it constrains the selection of transportation modes, is determined based on the preliminary disaster assessment, and is dynamically updated according to real-time disaster information in actual operation;

[0022] For each path Determine its traffic capacity attenuation coefficient ;

[0023] like :set up This means that the road is feasible;

[0024] At this time, the actual traffic capacity of the highway is ;

[0025] Otherwise: Settings This means the road is not feasible;

[0026] in, This is a binary parameter representing the feasibility of a post-disaster recovery path. , which are input as known parameters into the scheduling optimization model; A collection of road transport routes from material distribution points to material demand points; This is the road transport capacity attenuation coefficient; This is the threshold for traffic capacity attenuation. To advance the distribution of supplies to the point of need for supplies Initial road transport capacity before the disaster; completion of the physical architecture design and parameter preparation of the emergency multimodal transport network, based on this network framework and the acquired parameters.

[0027] As a preferred embodiment of the present invention, step B further includes the following step:

[0028] B1. Model Assumptions

[0029] (1) In response to localized and regional disaster emergency scenarios, the disaster-stricken areas and their surrounding last-mile delivery networks are affected by the disaster;

[0030] (2) The locations of supply points, material forwarding points and material demand points are known. Due to the need for organization, management and distribution of cross-regional emergency support, materials cannot be transported across levels. That is, emergency materials can only be transported from the material supply point to the material forwarding point, and then from the material forwarding point to the material demand point. The material supply point and the material forwarding point are close to the airport, railway station and highway.

[0031] (3) The transportation from the material supply point to the material forwarding point is a long distance, and the two transportation modes are railway and air. The transportation from the material forwarding point to the material demand point is a short distance, and the two transportation modes are road and drone.

[0032] (4) The material demand points are divided into those that only arrive by drone swarm and those that arrive by truck based on geographical and task attributes. The emergency material demand of each material demand point comes from the disaster assessment results. The core uncertainty of the scheduling optimization model comes from the transportation time of the last delivery.

[0033] (5) The effectiveness of emergency rescue is determined by the time of the latest material demand point to be rescued. Therefore, the maximum rescue completion time is taken as the core optimization indicator of emergency dispatch.

[0034] As a preferred embodiment of the present invention, step B further includes the following step:

[0035] B2. Defining uncertainty: the road transport time from the point of origin of the goods to the point of demand for the goods. Defined as a random variable whose probability distribution information is incomplete. ;

[0036] B3. Construct moment fuzzy sets, defining uncertainty using fuzzy sets based on first-order and second-order moment information and physical support sets; estimate the expected value of transportation time for each route based on UAV disaster sentiment perception results, historical data, and disaster assessment expert experience. Upper bound of variance and feasible time for transport determined in conjunction with road physical constraints As input to moment-based fuzzy sets; construct moment fuzzy sets. Includes all values ​​that satisfy the expected value equal to Variance not exceeding probability distribution .

[0037] As a preferred embodiment of the present invention, step B further includes the following step:

[0038] B4. Establish a basic scheduling optimization model.

[0039] Establish a risk-neutral sub-Bruker optimization model as a benchmark, in scenario Below, material demand points The rescue completion time consists of trunk line transportation time, transfer time, and terminal transportation time, defined as... ;

[0040] For supplies from the point of origin via material forwarding point to the point of need for supplies Using transportation methods Transporting emergency supplies Its transportation time for: ;

[0041] in:

[0042] ;

[0043] To ensure that the completion time of material demand points is determined solely by the actual transportation route and mode, the following constraints are introduced:

[0044] ;

[0045] Among them, parameters To meet the constraints of scheduling cycle and time window The maximum possible value and The difference between the minimum possible values; For transportation methods From the supply point to the material forwarding point Average transit time; It is a binary variable; In the context The following transportation methods are adopted. From the material forwarding point to the point of need for supplies For road transport, the transportation time is a random variable. Its distribution is uncertain; For emergency supplies at the material forwarding point Transportation methods are Convert to The average transit time and the maximum rescue completion time of the entire emergency dispatch plan are determined by the latest completion time among all material demand points, satisfying the following:

[0046] ;

[0047] In the formula, For the context The time required to complete the rescue operation at the point where supplies are needed; For the context Maximum rescue completion time;

[0048] Without considering the risk aversion preferences of decision-makers, a risk-neutral partial Bruker optimization model is established as a benchmark, with the objective of minimizing the value in the moment fuzzy set. Worst-case scenario and maximum rescue completion time:

[0049] .

[0050] As a preferred embodiment of the present invention, step B further includes the following step:

[0051] B5. Introducing risk measurement and constructing a risk sensitivity model: To reflect decision-makers' risk aversion attitude in emergency dispatch, conditional risk value is introduced as a risk measurement tool, and a confidence level is set. and risk weight parameters Considering the probability distribution of post-disaster transportation time It is unknown and difficult to obtain. By adopting the Brussels bar optimization method, the value is minimized in the pre-constructed moment fuzzy set. The worst-case scheduling performance metrics are considered; therefore, a risk-sensitive sub-bar optimization model is constructed, with its objective function being the worst-case mean. The statement is as follows:

[0052] ;

[0053] in, , As an auxiliary variable;

[0054] In the formula, for confidence level, ; For risk weight parameters, ; This is the maximum time required to complete the rescue operation.

[0055] As a preferred embodiment of the present invention, step B further includes the following step:

[0056] B6. Define constraints.

[0057] (1) Constraints on the integrity of the transportation process;

[0058] ;

[0059] This means that for each material forwarding point, there must be a means of transportation to deliver materials from the material supply point to that material forwarding point. In other words, at least one material supply point serves the material forwarding point, which is used to ensure the integrity of the material supply point and material forwarding point link in the emergency material dispatching process.

[0060] ;

[0061] This means that for each material demand point, there is a means of transportation to deliver the materials from the material forwarding point to the corresponding material demand point. In other words, there is at least one material forwarding point serving the material demand point, ensuring the integrity of this key link between the material forwarding point and the material demand point in the emergency material dispatch process, and ensuring that the material demand point can receive emergency materials.

[0062] In the formula, It is a binary variable, if emergency supplies By transportation From the supply point Transport to the material forwarding point The value is 1 if it is 1, otherwise it is 0. It is a binary variable, if emergency supplies By transportation From the material forward deployment point Transport to the point of demand for supplies The value is 1 if it is 1, otherwise it is 0.

[0063] (2) Uniqueness constraint of mode of transport,

[0064] ;

[0065] ;

[0066] This indicates that when transporting a certain type of emergency supplies between two fixed nodes, only one mode of transportation can be selected.

[0067] (3) Material forwarding point balancing constraints,

[0068] ;

[0069] The total amount of materials at the material forwarding point includes the total amount of materials received from each material supply point as well as the original inventory of the material forwarding point itself.

[0070] In the formula, For trunk line transport volume, a continuous variable, emergency supplies By transportation From the supply point Transport to the material forwarding point The volume of transport; As a forward position for supplies China Emergency Supplies Inventory levels; As a forward position for supplies emergency supplies The total quantity;

[0071] (4) Constraints on material connection at material forwarding points.

[0072] ;

[0073] This means that the total amount of materials sent from the material forwarding point to each material demand point shall not exceed the sum of the total amount of materials received by the center from each material supply point and its initial inventory, thus ensuring the rationality of material flow at the material forwarding point.

[0074] In the formula, For end-point transportation volume, a continuous variable, emergency supplies By transportation From the material forward deployment point Transport to the point of demand for supplies The volume of transport; A collection of emergency supplies types;

[0075] (5) Constraints on the amount of materials received at material forwarding points.

[0076] ;

[0077] This indicates that the amount of materials received by the material forwarding point does not exceed the material inventory of the material supply point;

[0078] In the formula, supply point China Emergency Supplies Inventory levels;

[0079] (6) Limits on the amount of materials that can be received at the material demand point.

[0080] ;

[0081] This represents the total quantity of materials received by each material demand point from the material forwarding point.

[0082] ;

[0083] This means that the amount of materials received by the material demand point must not only meet the minimum demand fulfillment rate and ensure the supply of materials to the material demand point, but also ensure that the total amount of materials transported to the demand node does not exceed its total demand.

[0084] In the formula, For material demand points Emergency supplies were received from the forward distribution point. The total quantity; For material demand points emergency supplies The minimum demand satisfaction rate; For material demand points emergency supplies Demand;

[0085] (7) Consistency constraints between transshipment and transportation modes,

[0086] , ;

[0087] This indicates that the transshipment operation must match the transportation methods of the preceding and following stages;

[0088] In the formula, It is a binary variable, if emergency supplies From the supply point via material forwarding point to the point of need for supplies Along the route, the mode of transportation changed from Convert to The value is 1 if it is 1, otherwise it is 0.

[0089] (8) Calculation of transportation time,

[0090] Mainline transport section: ;

[0091] Last-mile transport segment: ;

[0092] In the formula, For transportation methods From the supply point to the material forwarding point The shortest path mileage; For transportation methods Next, from the material forward deployment point to the point of need for supplies The shortest path mileage; For transportation methods The average speed; This refers to the railway transport time for the main transport section; For air transport time in the trunk line segment; To supply points to the material forwarding point The road network coefficient reflects the impact of disasters on ground transportation on that section of road, and the mode of transport. Including railway transportation or air transport Transportation methods Including road transport or drone transport ; For the last-mile delivery segment, the drone delivery time is specified.

[0093] (9) Constraints on the number of vehicles,

[0094] Number of transport vehicles dispatched by the supply points ;

[0095] Number of transport vehicles dispatched from the material forward deployment point ;

[0096] In the formula, For transportation methods From the supply point to the material forwarding point Transporting emergency supplies The number of means of transport used; For transportation methods Next, from the material forward deployment point to the point of need for supplies Transporting emergency supplies The number of transportation vehicles used All values ​​are non-negative integers, ensuring that the loading capacity of the transport vehicle is greater than the planned cargo volume to be transported; For emergency supplies The unit weight; For transportation methods Below, the rated load-bearing capacity of a unit transportation vehicle;

[0097] (10) Capacity constraints of transportation modes,

[0098] Supply point transportation capacity constraints:

[0099] ;

[0100] Capacity constraints of material forwarding point transportation methods:

[0101] ;

[0102] The number of various modes of transportation vehicles dispatched from each material supply point or material forwarding point shall be limited to no more than the number of available modes of transportation.

[0103] In the formula, supply point Transportation methods used Number of tools; As a forward position for supplies Last-mile transportation method used Number of tools;

[0104] (11) Transportation mode constraints based on road feasibility,

[0105] Road transport selection constraint: Ensures that road transport can only be selected when the road is passable;

[0106] ;

[0107] Road transport vehicle quantity constraint: The total number of road transport vehicles carrying all goods on each route is limited to the actual traffic capacity that the road can handle.

[0108] ;

[0109] Autonomous selection of drone mode: When roads are impassable or have insufficient capacity after a disaster, the scheduling optimization model selects drones to undertake the last-mile delivery under the constraints of feasibility and transport capacity.

[0110] (12) Unmanned aerial vehicle (UAV) flight distance constraints,

[0111] For routes using drones for transportation, flight distance constraints must be met;

[0112] ;

[0113] In the formula, A collection of drone transportation routes from material distribution points to material demand points; Battery consumption per unit flight distance of the drone; This represents the maximum battery capacity of the drone.

[0114] If drone delivery is chosen, the battery consumption for the round trip distance must not exceed the maximum battery capacity.

[0115] (13) Logical relationship constraints between transport volume and transport mode selection

[0116] To ensure that the transport volume occurs only on the selected transport arc, establish a logical relationship between continuous variables and binary variables;

[0117] ;

[0118] ;

[0119] In the formula, This represents the upper limit of the transport volume for the main transport section. This represents the upper limit of the transport volume in the final transportation segment.

[0120] As a preferred embodiment of the present invention, step C further includes the following step:

[0121] C1. Model Solving: The scheduling optimization model is solved by using the CCG method to generate columns and constraints;

[0122] C1.1. Main problem initialization,

[0123] The main problem is constructed, which minimizes the approximate form of the risk-sensitive objective function under the given set of scenarios, while satisfying the constraints of step B. In the main problem, auxiliary variables and scenario slack variables are introduced to contextualize and linearize the conditional risk value term defined in step B, thereby transforming the risk-sensitive objective function into a mixed-integer linear programming form. In the initial stage, the scenario set only includes nominal scenarios or representative scenarios with limited quantities constructed based on expected transportation time, in order to obtain an initial feasible scheduling scheme that satisfies the constraints.

[0124] C1.2. Subproblem Construction and Worst-Case Scenario Generation

[0125] After obtaining the current decision for the main problem, subproblems are constructed to solve the moment fuzzy set given the decision of the main problem. The subproblem aims to maximize the objective function value under a given scheduling scheme, and solves for the worst-case transportation time that leads to the maximum rescue completion time or the maximum tail risk, while satisfying the moment information constraint and support set constraint. The subproblem is solved by dualization method to obtain one or more worst-case scenarios.

[0126] C1.3. Column and Constraint Generation

[0127] Add the worst-case scenarios obtained from the subproblems to the scenario set, and then add them to the main problem. Linearization and time constraints are applied to gradually tighten the main problem's ability to characterize the worst-case scenario; then, the updated main problem is solved again to obtain a new scheduling decision; by repeatedly executing the iterative process of solving the main problem and generating worst-case scenarios from subproblems and updating the main problem, the objective value of the main problem gradually approaches the optimal solution of the original risk-sensitive sub-Bruker optimization model;

[0128] C1.4. Termination Conditions and Convergence Criteria

[0129] When the difference between the objective function value corresponding to the worst-case scenario generated by the subproblem and the objective value of the current main problem does not exceed the preset tolerance or reaches the maximum number of iterations, the CCG iteration is terminated; at this time, the solution to the main problem is regarded as the optimal or near-optimal solution of the original risk-sensitive sub-Bruker optimization model.

[0130] As a preferred embodiment of the present invention, step C further includes the following step:

[0131] C2. Scheduling scheme generation,

[0132] Based on the solution to the principal problem after CCG convergence, a multi-dimensional scheduling scheme is generated, including the trunk and terminal transportation modes and route selection schemes, the material flow allocation schemes for each transportation stage, and the expected value of the maximum rescue completion time and its conditional risk value at the confidence level.

[0133] As a preferred embodiment of the present invention, step C further includes the following step:

[0134] C3. Scheme Validation and Sensitivity Analysis

[0135] C3.1. Robustness testing of the solution.

[0136] In the fuzzy set Within a certain range, multiple random scenarios are generated using the Monte Carlo method, a scheduling scheme is executed, the actual objective function value is calculated, and the performance of the scheme under different scenarios is compared.

[0137] C3.2. Parameter sensitivity analysis provides decision-makers with a quantitative basis for risk preference through systematic parameter adjustments:

[0138] Confidence level Analysis: Different tests The tail risk control effect under the value provides a reference for setting risk tolerance;

[0139] Risk weight parameters Analysis: Plot a risk-efficiency trade-off curve by varying the value from 0 to 1, clarifying the different... The corresponding scheme characteristics are used to present decision-makers with scheme options under different risk preferences, so as to weigh and select among multiple alternatives;

[0140] Supply and demand change test: Simulate demand fluctuations and supply disruption scenarios to verify the adaptability of the solution.

[0141] The beneficial effects of this invention are as follows:

[0142] 1. This invention constructs a multimodal transport network that coordinates "trunk line-last mile" transportation, integrating the advantages of railway and large freight aircraft trunk line transport with "truck + drone" last-mile delivery, achieving precise and efficient connection of emergency supplies from supply point to demand point. Based on this, it innovatively establishes a risk-sensitive sub-Bruker optimization model to effectively characterize the uncertainty of post-disaster transportation time and introduces conditional risk values ​​to quantify decision-makers' risk aversion attitude. Thus, in complex disaster environments with scarce data, it generates material scheduling schemes that combine high timeliness, high reliability, and high feasibility. This method overcomes the limitations of traditional scheduling models, which are often simplistic and lack sufficient risk consideration, significantly improving the adaptability and robustness of emergency logistics systems under extreme scenarios such as route disruptions, providing crucial decision support for scientific disaster relief.

[0143] 2. This invention is based on a asynchronous multi-mode "trunk line-terminal" collaborative emergency material cross-regional multimodal transport model, constructing a three-level emergency logistics network of supply point-forward point-demand point, integrating multiple transportation modes such as railway, highway, air (large cargo plane) and drone, breaking through the limitations of traditional single transportation mode. In the first phase (from supply point to forward location), due to the distance from the disaster area and the need for cross-regional dispatch of large volumes and concentrated batches of emergency supplies, railways, highways, and large cargo planes were used for cross-regional transportation, giving full play to their advantages of stable capacity and timeliness. The transportation goal was "bulk replenishment." In the second phase (from forward location to demand point), in response to problems such as road damage and poor accessibility in remote areas, an asynchronous multi-mode architecture was adopted, proposing a "truck + drone" dual-mode delivery solution: based on disaster assessment data and road feasibility assessment results, dynamic selection was made between truck transportation and drone transportation, allowing different transportation modes to independently perform last-mile delivery tasks at different demand points and at different time periods; when the road route is feasible and has sufficient traffic capacity, truck transportation is given priority; when the road is not feasible or the traffic capacity is insufficient, point-to-point delivery by drone swarms is switched to achieve accurate and timely delivery of the "last mile." This model effectively overcomes the safety hazards of traditional single-truck transport requiring deep penetration into dangerous disaster areas by differentiating task allocation to transport vehicles based on the characteristics of the supplies. It also specifically addresses the pain points of traditional models, such as transport disruptions caused by road damage and the inability or slow delivery of supplies to remote areas. Through the complementary advantages of the two modes of transport, even when the capacity of traditional road networks is reduced, timely and effective delivery of relief supplies can be ensured. While guaranteeing transport safety, the solution significantly improves emergency response speed and supply delivery coverage, thereby enhancing the robustness and overall response efficiency of the emergency logistics system.

[0144] 3. This invention is based on a three-layer network of emergency material supply points, forward deployment points, and disaster-stricken areas. Considering the high uncertainty in transport vehicle travel time due to severe ground traffic conditions near disaster areas (such as road damage and traffic congestion), and the scarcity and incomplete distribution information of relevant historical data in disaster environments, a fuzzy set is constructed based on easily estimable moment information to characterize the distribution uncertainty of transport time. To avoid low-probability, high-loss events, the scheduling optimization model introduces Conditional Value-at-Risk (CVaR) as a risk measurement tool to quantify the extreme delay risk caused by transport time uncertainty due to factors such as road damage, thereby reflecting the risk aversion preferences of decision-makers. Furthermore, a biblical optimization model based on the worst-case mean minus the conditional risk value is proposed to minimize the expected latest rescue completion time in the worst-case scenario. Simultaneously, considering the potential fluctuations in road capacity caused by post-disaster road damage, the model further incorporates a highway transport capacity attenuation coefficient to characterize this influencing factor. In addition, the model incorporates the operational limitations of drones into the constraint system, including the maximum flight distance constraint based on round-trip range and battery capacity, as well as the maximum load constraint limited by the carrier structure, thereby ensuring the feasibility and safety of drone transportation. Attached Figure Description

[0145] Other objects and results of the invention will become more apparent and readily understood with reference to the following description taken in conjunction with the accompanying drawings. In the drawings:

[0146] Figure 1 This is a technical roadmap for an embodiment of the present invention.

[0147] Figure 2 This is a topology diagram of a three-level emergency material multimodal transport network according to an embodiment of the present invention.

[0148] Figure 3 This is a system architecture diagram for optimizing the multimodal transport scheduling of emergency supplies, taking into account risk sensitivity, according to an embodiment of the present invention. Detailed Implementation

[0149] See Figure 1-3 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0150] This invention provides a method for optimizing the multimodal transport scheduling of emergency supplies that takes into account risk sensitivity, comprising the following steps:

[0151] A. Construct an emergency material multimodal transport network that coordinates trunk lines and terminal stations. Integrate multi-source data on disaster assessment, supply and demand inventory, transportation network and vehicle status, and clean and merge the data. Then, map trunk line transport and terminal transport into a multimodal transport network diagram, and clarify the nodes, feasible route sets and their parameters (distance / time, traffic capacity and feasibility, transport capacity limit) to provide structured input for step B.

[0152] A1. Data Acquisition: Information on disaster situation, material supply and demand (including inventory at each supply point and demand at each demand point), transportation network information (including node geographic coordinates, path distances, and road conditions), available transportation vehicles and their parameters (speed, load limits, drone battery capacity), accessibility assessment data for disaster-stricken areas, and damage to critical infrastructure are acquired through disaster site monitoring, feedback from emergency management departments, and drone aerial surveys. A disaster parameter database is established, using historical data or expert experience values ​​when information is missing. Confidential intervals / upper and lower bounds are set for missing data and constrained in the scheduling optimization model using support sets. A dynamic data update mechanism is also implemented, dynamically updating the database as reconnaissance information is continuously incorporated.

[0153] A2. Identify key transportation nodes and determine material supply points, material forwarding points, and material demand points. Material forwarding points are selected from existing emergency warehouses, assembly points around airports / train stations, or a set of alternative candidate points.

[0154] A3. Design a trunk and terminal collaborative transportation mode based on asynchronous multi-mode. Based on the geographical distribution of nodes and disaster situation, the transportation stage is divided into trunk transportation and terminal transportation. Trunk transportation includes the material supply point to the material forwarding point, by integrating high-capacity transportation methods such as railway, highway, and air (large cargo plane); terminal transportation includes the material forwarding point to the material demand point, by designing a road plus drone delivery method.

[0155] A4. Construct a three-tiered network topology: a three-tiered emergency material multimodal transport network topology based on material supply points, material forwarding points, and material demand points.

[0156] Abstract it into a network graph when constructing a network. , Network diagram The set of nodes includes supply points, forward deployment points, and demand points for supplies. , To gather at the supply point For the collection of supplies at the forward deployment point For the aggregation of material demand points, there is a three-tiered emergency material multimodal transport network. individual supply points One material forwarding point and One point of demand for supplies; Network diagram The set of edges; A collection of modes of transportation This includes road, rail, and air transport (large cargo planes) and drones. It is a collection of trunk transportation modes, namely, material supply points. to the material forwarding point A collection of transportation methods ,in For railway transportation, For air transport (large cargo planes). This refers to a collection of last-mile transportation methods, i.e., a forward distribution point for goods. to the point of need for supplies A collection of transportation methods ,in For road transport (trucks). For drone transportation;

[0157] A5. Road feasibility assessment,

[0158] The roads between the material distribution point and the material demand point are located in the disaster area. Based on the traffic network information obtained from the drone assessment mission and the disaster assessment data obtained from the analysis, image recognition or data analysis methods are used to assess the degree of damage to the route and determine the post-disaster road transport capacity attenuation coefficient of the route. With traffic capacity attenuation threshold The relationship between these factors allows for the pre-determination of road transport feasibility parameters for each route. As input to the scheduling optimization model, it constrains the selection of transportation modes, is determined based on the preliminary disaster assessment, and is dynamically updated according to real-time disaster information in actual operation;

[0159] For each path Determine its traffic capacity attenuation coefficient ;

[0160] like :set up This means that the road is feasible;

[0161] At this time, the actual traffic capacity of the highway is ;

[0162] Otherwise: Settings This means the road is not feasible;

[0163] in, This is a binary parameter representing the feasibility of a post-disaster recovery path. , which are input as known parameters into the scheduling optimization model; A collection of road transport routes from material distribution points to material demand points; This is the road transport capacity attenuation coefficient; This is the threshold for traffic capacity attenuation. To advance the distribution of supplies to the point of need for supplies Pre-disaster initial road transport capacity (number of vehicles); complete the physical architecture design and parameter preparation of the emergency multimodal transport network, based on this network framework and the acquired parameters.

[0164] B. Establish a risk-sensitive emergency material multimodal transport scheduling optimization model. Based on the emergency material multimodal transport network with trunk and terminal coordination constructed in step A and the obtained parameters, establish a risk-sensitive scheduling optimization model.

[0165] B1. Model Assumptions

[0166] (1) In response to localized and regional disaster emergency scenarios, the disaster-stricken area (material demand point) and its surrounding last-mile delivery network are affected by the disaster;

[0167] (2) The locations of supply points, material forwarding points and material demand points are known. Due to the need for organization, management and distribution of cross-regional emergency support, materials cannot be transported across levels. That is, emergency materials can only be transported from the material supply point to the material forwarding point, and then from the material forwarding point to the material demand point. The material supply point and the material forwarding point are close to the airport, railway station and highway.

[0168] (3) The transportation from the material supply point to the material forwarding point is a long distance, so the two transportation methods are railway and air (large cargo plane). The transportation from the material forwarding point to the material demand point is a short distance, so the two transportation methods are road (truck) and drone.

[0169] (4) Material demand points are divided into those only accessible by drone swarm and those accessible by truck based on geographical and mission attributes. The emergency material demand at each material demand point is derived from the disaster assessment results. The core uncertainty of the scheduling optimization model comes from the transportation time of the last delivery (for last-mile drone transportation, the focus is on its scheduling under normal weather conditions. In practical applications, this parameter can be dynamically adjusted based on real-time weather data. The impact of the uncertainty of truck travel time is analyzed in detail).

[0170] (5) The effectiveness of emergency rescue is determined by the time of the latest material demand point to be rescued. Therefore, the maximum rescue completion time is taken as the core optimization indicator of emergency dispatch.

[0171] B2. Defining uncertainty: the road transport time from the point of origin of the goods to the point of demand for the goods. Defined as a random variable whose probability distribution information is incomplete. ;

[0172] B3. Constructing moment fuzzy sets: Addressing the incomplete nature of road transport time distribution information in disaster environments, uncertainty is defined using fuzzy sets based on first- and second-order moment information and physical support sets. Based on UAV disaster perception results, historical data, and disaster assessment expert experience, the expected transport time for each route is estimated. Upper bound of variance and feasible time for transport determined in conjunction with road physical constraints As input to moment-based fuzzy sets; construct moment fuzzy sets. Includes all values ​​that satisfy the expected value equal to Variance not exceeding probability distribution .

[0173] ;

[0174] In the formula Let it be the probability distribution space;

[0175] Moment fuzzy set Without assuming a specific shape of the distribution, while utilizing some of the obtained statistical information, and avoiding excessive conservatism through support set constraints, it achieves the optimal balance between risk control and scheduling efficiency under limited information conditions.

[0176] B4. Establish a basic scheduling optimization model (risk neutral).

[0177] Establish a risk-neutral sub-Bruker optimization model as a benchmark, in scenario Below, material demand points The rescue completion time consists of trunk line transportation time, transfer time, and terminal transportation time, defined as... ;

[0178] For supplies from the point of origin via material forwarding point to the point of need for supplies Using transportation methods Transporting emergency supplies Its transportation time for: ;

[0179] in:

[0180] ;

[0181] To ensure that the completion time of material demand points is determined solely by the actual transportation route and mode, the following constraints are introduced:

[0182] ;

[0183] Among them, parameters To meet the constraints of scheduling cycle and time window The maximum possible value and The difference between the minimum possible values; For transportation methods From the supply point to the material forwarding point Average transit time; It is a binary variable; In the context The following transportation methods are adopted. From the material forwarding point to the point of need for supplies For road transport, the transportation time is a random variable. Its distribution is uncertain; For emergency supplies at the material forwarding point Transportation methods are Convert to The average transit time and the maximum rescue completion time of the entire emergency dispatch plan are determined by the latest completion time among all material demand points, satisfying the following:

[0184] ;

[0185] In the formula, For the context The time required to complete the rescue operation at the point where supplies are needed; For the context Maximum rescue completion time;

[0186] Without considering the risk aversion preferences of decision-makers, a risk-neutral partial Bruker optimization model is established as a benchmark, with the objective of minimizing the value in the moment fuzzy set. Worst-case scenario and maximum rescue completion time:

[0187] .

[0188] B5. Introducing Risk Measurement and Constructing a Risk Sensitivity Model: To reflect decision-makers' risk aversion attitude in emergency response, Conditional Value at Risk (CVaR) is introduced as a risk measurement tool, and a confidence level is set. and risk weight parameters Considering the probability distribution of post-disaster transportation time It is unknown and difficult to obtain. By adopting the Brussels bar optimization method, the value is minimized in the pre-constructed moment fuzzy set. The worst-case scheduling performance metrics are considered; therefore, a risk-sensitive sub-bar optimization model is constructed, with its objective function being the worst-case mean. The statement is as follows:

[0189] ;

[0190] in, , As an auxiliary variable;

[0191] In the formula, for confidence level, ; For risk weight parameters, ; This represents the maximum time required to complete the rescue operation. Confidence level. and weight parameters The size represents the relative importance of the risk item and reflects the decision-maker's risk attitude;

[0192] The larger the tail risk, the more extreme the tail risk becomes for policymakers. The smaller the value, the more attention is paid to relatively mild risk situations; when At that time, it degenerates into the traditional expected value model of the split bar. At that time, it focused entirely on the tail end of the risk. It involves weighing expected value against the tail of risk. and When used in combination, it can precisely characterize decision-makers' sensitivity to risks of varying severity, providing a flexible risk management tool for emergency decision-making in different disaster scenarios;

[0193] B6. Define constraints.

[0194] (1) Constraints on the integrity of the transportation process;

[0195] ;

[0196] This means that for each material forwarding point, there must be a means of transportation to deliver materials from the material supply point to that material forwarding point. In other words, at least one material supply point serves the material forwarding point. This is used to ensure the integrity of the material supply point and material forwarding point link in the emergency material dispatching process and to avoid situations where the material forwarding point cannot receive materials.

[0197] ;

[0198] This means that for each material demand point, there is a means of transportation to deliver the materials from the material forwarding point to the corresponding material demand point. In other words, there is at least one material forwarding point serving the material demand point, ensuring the integrity of this key link between the material forwarding point and the material demand point in the emergency material dispatch process, and ensuring that the material demand point can receive emergency materials.

[0199] In the formula, It is a binary variable, if emergency supplies By transportation From the supply point Transport to the material forwarding point The value is 1 if it is 1, otherwise it is 0. It is a binary variable, if emergency supplies By transportation From the material forward deployment point Transport to the point of demand for supplies The value is 1 if it is 1, otherwise it is 0.

[0200] (2) Uniqueness constraint of mode of transport,

[0201] ;

[0202] ;

[0203] This indicates that when transporting a certain type of emergency supplies between two fixed nodes, only one mode of transportation can be selected.

[0204] (3) Material forwarding point balancing constraints,

[0205] ;

[0206] The total amount of materials at the material forwarding point includes the total amount of materials received from each material supply point as well as the original inventory of the material forwarding point itself.

[0207] In the formula, For trunk line transport volume, a continuous variable, emergency supplies By transportation From the supply point Transport to the material forwarding point The volume of transport; As a forward position for supplies China Emergency Supplies Inventory levels; As a forward position for supplies emergency supplies The total quantity;

[0208] (4) Constraints on material connection at material forwarding points.

[0209] ;

[0210] This means that the total amount of materials sent from the material forwarding point to each material demand point shall not exceed the sum of the total amount of materials received by the center from each material supply point and its initial inventory, thus ensuring the rationality of material flow at the material forwarding point.

[0211] In the formula, For end-point transportation volume, a continuous variable, emergency supplies By transportation From the material forward deployment point Transport to the point of demand for supplies The volume of transport; This is a collection of emergency supplies (including daily necessities such as food, drinking water, tents, etc., and medical supplies such as first aid medicines and blood plasma).

[0212] (5) Constraints on the amount of materials received at material forwarding points.

[0213] ;

[0214] This indicates that the amount of materials received by the material forwarding point does not exceed the material inventory of the material supply point;

[0215] In the formula, supply point China Emergency Supplies Inventory levels;

[0216] (6) Limits on the amount of materials that can be received at the material demand point.

[0217] ;

[0218] This represents the total quantity of materials received by each material demand point from the material forwarding point.

[0219] ;

[0220] This means that the amount of supplies received by the material demand point must not only meet the minimum demand fulfillment rate to ensure the supply of materials to the material demand point and avoid the failure of rescue due to insufficient supply, but also ensure that the total amount of materials transported to the demand point does not exceed its total demand.

[0221] In the formula, For material demand points Emergency supplies were received from the forward distribution point. The total quantity; For material demand points emergency supplies The minimum demand satisfaction rate (emergency protection baseline); For material demand points emergency supplies Demand;

[0222] (7) Consistency constraints between transshipment and transportation modes,

[0223] , ;

[0224] This indicates that the transshipment operation must match the transportation methods of the preceding and following stages;

[0225] In the formula, It is a binary variable, if emergency supplies From the supply point via material forwarding point to the point of need for supplies Along the route, the mode of transportation changed from Convert to The value is 1 if it is 1, otherwise it is 0.

[0226] (8) Calculation of transportation time,

[0227] Mainline transport section: ;

[0228] Last-mile transport segment: ;

[0229] In the formula, For transportation methods From the supply point to the material forwarding point The shortest path mileage; For transportation methods Next, from the material forward deployment point to the point of need for supplies The shortest path mileage; For transportation methods The average speed; This refers to the railway transport time for the main transport section; For air transport time in the trunk line segment; To supply points to the material forwarding point The road network coefficient reflects the impact of disasters on ground transportation on that section of road, and the mode of transport. Including railway transportation Or air transport (large cargo planes) Transportation methods Including road transport (trucks) or drone transport ; For the last-mile delivery segment, the drone delivery time is specified.

[0230] (9) Constraints on the number of vehicles,

[0231] Number of transport vehicles dispatched by the supply points ;

[0232] Number of transport vehicles dispatched from the material forward deployment point ;

[0233] In the formula, For transportation methods From the supply point to the material forwarding point Transporting emergency supplies The number of means of transport used; For transportation methods Next, from the material forward deployment point to the point of need for supplies Transporting emergency supplies The number of transportation vehicles used All values ​​are non-negative integers, ensuring that the loading capacity of the transport vehicle is greater than the planned cargo volume to be transported; For emergency supplies The unit weight; For transportation methods Below is the rated load capacity (weight) of a unit transportation vehicle.

[0234] (10) Capacity constraints of transportation modes,

[0235] Supply point transportation capacity constraints:

[0236] ;

[0237] Capacity constraints of material forwarding point transportation methods:

[0238] ;

[0239] The number of various modes of transportation vehicles dispatched from each material supply point or material forwarding point shall be limited to no more than the number of available modes of transportation.

[0240] In the formula, supply point Transportation methods used Number of tools; As a forward position for supplies Last-mile transportation method used Number of tools;

[0241] (11) Transportation mode constraints based on road feasibility,

[0242] Road transport selection constraint: Ensures that road transport can only be selected when the road is passable;

[0243] ;

[0244] Road transport vehicle quantity constraint: The total number of road transport vehicles carrying all goods on each route is limited to the actual traffic capacity that the road can handle.

[0245] ;

[0246] Autonomous selection of drone mode: When roads are impassable or have insufficient capacity after a disaster, the scheduling optimization model selects drones to undertake the last-mile delivery under the constraints of feasibility and transport capacity.

[0247] (12) Unmanned aerial vehicle (UAV) flight distance constraints,

[0248] For routes using drones for transportation, flight distance constraints must be met;

[0249] ;

[0250] In the formula, A collection of drone transportation routes from material distribution points to material demand points; Battery consumption per unit flight distance of the drone; This represents the maximum battery capacity of the drone.

[0251] If drone delivery is chosen, the constraint ensures that the battery consumption for the round trip does not exceed the maximum battery capacity.

[0252] (13) Logical relationship constraints between transport volume and transport mode selection

[0253] To ensure that the transport volume occurs only on the selected transport arc, establish a logical relationship between continuous variables and binary variables;

[0254] ;

[0255] ;

[0256] In the formula, This represents the upper limit of the transport volume for the main transport section. This represents the upper limit of the transport volume in the final transportation segment.

[0257] C. Generate model solution results and scheduling schemes. Transform the scheduling optimization model established in step B into a solvable form, generate an executable scheduling scheme, and verify the results to form a complete technical closed loop.

[0258] C1. Model Solving: Since the original scheduling optimization model based on split-bar optimization is a min-max problem, which includes an inner dual problem and an outer optimization problem, it is difficult to directly transform it into a deterministic model for solving in one go. Therefore, the column-and-constraint generation (CCG) method is used to solve the scheduling optimization model.

[0259] C1.1. Main problem initialization,

[0260] The main problem is constructed, which minimizes the approximate form of the risk-sensitive objective function under the given set of scenarios, while satisfying the constraints of step B. In the main problem, auxiliary variables and scenario slack variables are introduced to contextualize and linearize the conditional risk value (CVaR) term defined in step B, thereby transforming the risk-sensitive objective function into a mixed-integer linear programming form. In the initial stage, the scenario set only includes nominal scenarios or representative scenarios with limited quantities constructed based on expected transportation time, in order to obtain an initial feasible scheduling scheme that satisfies the constraints.

[0261] C1.2. Subproblem Construction and Worst-Case Scenario Generation

[0262] After obtaining the current decision for the main problem, subproblems are constructed to solve the moment fuzzy set given the decision of the main problem. The subproblem aims to maximize the objective function value under a given scheduling scheme, and solves for the worst-case transportation time that leads to the maximum rescue completion time or the maximum tail risk, while satisfying the moment information constraint and support set constraint. The subproblem is solved by dualization method to obtain one or more worst-case scenarios.

[0263] C1.3. Column and Constraint Generation

[0264] Add the worst-case scenarios obtained from the subproblems to the scenario set, and then add them to the main problem. Linearization and time constraints are applied to gradually tighten the main problem's ability to characterize the worst-case scenario; then, the updated main problem is solved again to obtain a new scheduling decision; by repeatedly executing the iterative process of solving the main problem and generating worst-case scenarios from subproblems and updating the main problem, the objective value of the main problem gradually approaches the optimal solution of the original risk-sensitive sub-Bruker optimization model;

[0265] C1.4. Termination Conditions and Convergence Criteria

[0266] When the difference between the objective function value corresponding to the worst-case scenario generated by the subproblem and the objective value of the current main problem does not exceed the preset tolerance or reaches the maximum number of iterations, the CCG iteration is terminated; at this time, the solution to the main problem is regarded as the optimal or near-optimal solution of the original risk-sensitive sub-Bruker optimization model.

[0267] C2. Scheduling scheme generation,

[0268] Based on the solution to the principal problem after CCG convergence, a multi-dimensional scheduling scheme is generated, including the trunk and terminal transportation modes and route selection schemes, the material flow allocation schemes for each transportation stage, and the expected value of the maximum rescue completion time and its conditional risk value at the confidence level.

[0269] The generated scheduling scheme takes into account both transportation efficiency and robustness, considering uncertainty and risk aversion preferences.

[0270] C3. Scheme Validation and Sensitivity Analysis

[0271] C3.1. Robustness testing of the solution.

[0272] In the fuzzy set Within a certain range, multiple random scenarios are generated using the Monte Carlo method, a scheduling scheme is executed, the actual objective function value is calculated, and the performance of the scheme under different scenarios is compared.

[0273] C3.2. Parameter sensitivity analysis provides decision-makers with a quantitative basis for risk preference through systematic parameter adjustments:

[0274] Confidence level Analysis: Different tests The tail risk control effect under the value provides a reference for setting risk tolerance;

[0275] Risk weight parameters Analysis: Plot a risk-efficiency trade-off curve by varying the value from 0 to 1, clarifying the different... The corresponding scheme characteristics are used to present decision-makers with scheme options under different risk preferences, so as to weigh and select among multiple alternatives;

[0276] Supply and demand change test: Simulate demand fluctuations and supply disruption scenarios to verify the adaptability of the solution.

[0277] This embodiment also relates to a system for implementing the above method. The system mainly includes: a data fusion and preprocessing module, a multimodal transport network modeling module, a risk parameter configuration and scheduling optimization module, and a scheduling scheme generation and verification module. It also realizes data interaction and result feedback through a system database, forming a closed-loop emergency dispatch system.

[0278] The data fusion and preprocessing module receives multi-source data, with its core function being to receive multi-source disaster-related data. The multi-source data includes at least disaster monitoring data, emergency material supply and demand data, transportation network data, and UAV reconnaissance data. The data fusion and preprocessing module is used to fuse and process the multi-source data. The multimodal transport network modeling module is used to construct a three-tiered emergency material multimodal transport network, including material supply points, material forwarding points, and material demand points, based on the parameters output by the data fusion and preprocessing module. It also determines the optional transport modes and corresponding network structure parameters for different transport stages, corresponding to step A. The risk parameter configuration and scheduling optimization module, corresponding to step B, introduces uncertainty and risk measurement indicators into the multimodal transport network. Based on the risk parameter configuration results, it optimizes the emergency material scheduling problem and outputs scheduling decision results, which at least include transport mode selection and material flow allocation results. The scheduling scheme generation and verification module, corresponding to step C, analyzes and integrates the scheduling decision results output by the optimization solution module, generates an executable emergency material scheduling scheme, and verifies the feasibility and performance of the scheme. The scheduling scheme can be further output in a visual form and generate scheduling instructions. The system database stores the multi-source data, network parameters, and scheduling decision results, and receives feedback information from the scheduling scheme generation and verification module to achieve continuous updating and optimization of the scheduling scheme.

[0279] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for optimizing the scheduling of multimodal transport of emergency supplies, considering risk sensitivity, characterized in that, Includes the following steps: A. Construct an emergency material multimodal transport network that coordinates trunk lines and terminal stations, integrate multi-source data on disaster assessment, supply and demand inventory, transportation network and vehicle status, and clean and merge the data. Then, map trunk line transport and terminal transport into a multimodal transport network diagram, clarify the nodes, feasible route sets and their parameters, and provide structured input for step B. B. Establish a risk-sensitive emergency material multimodal transport scheduling optimization model. Based on the emergency material multimodal transport network with trunk and terminal coordination constructed in step A and the obtained parameters, establish a risk-sensitive scheduling optimization model. Establish a basic scheduling optimization model. Establish a risk-neutral sub-Bruker optimization model as a benchmark, in scenario Below, material demand points The rescue completion time consists of trunk line transportation time, transfer time, and terminal transportation time, defined as... ; For supplies from the point of origin via material forwarding point to the point of need for supplies Using transportation methods Transporting emergency supplies Its transportation time for: ; in: ； For road transport, For drone transportation; To ensure that the completion time of material demand points is determined solely by the actual transportation route and mode, the following constraints are introduced: ； Among them, parameters To meet the constraints of scheduling cycle and time window The maximum possible value and The difference between the minimum possible values; For transportation methods From the supply point to the material forwarding point Average transit time; It is a binary variable; In the context The following transportation methods are adopted. From the material forwarding point to the point of need for supplies For road transport, the transportation time is a random variable. Its distribution is uncertain; For emergency supplies at the material forwarding point Transportation methods are Convert to The average transit time and the maximum rescue completion time of the entire emergency dispatch plan are determined by the latest completion time among all material demand points, satisfying the following: ； In the formula, For the context Lower material demand points The time it takes for the rescue to be completed; For the context Maximum rescue completion time; Without considering the risk aversion preferences of decision-makers, a risk-neutral split-bar optimization model is established as a benchmark. Its objective is to minimize the worst-case expected maximum rescue completion time on the moment fuzzy set. ； To reflect decision-makers' risk aversion attitude in emergency response, a risk measurement tool is introduced, along with a risk sensitivity model. This tool incorporates conditional risk value and sets confidence levels. and risk weight parameters Considering the probability distribution of post-disaster transportation time It is unknown and difficult to obtain. By adopting the Brussels bar optimization method, the value is minimized in the pre-constructed moment fuzzy set. The worst-case scheduling performance metrics are considered; therefore, a risk-sensitive sub-bar optimization model is constructed, with its objective function being the worst-case mean. The statement is as follows: ； in, , As an auxiliary variable; In the formula, for confidence level, ; For risk weight parameters, ; To maximize the time needed to complete the rescue operation, It is a collection of trunk transportation modes; C. Generate model solution results and scheduling schemes. Transform the scheduling optimization model established in step B into a solvable form, generate an executable scheduling scheme, and verify the results to form a complete technical closed loop.

2. The method for optimizing the multimodal transport scheduling of emergency supplies considering risk sensitivity as described in claim 1, characterized in that, Step A also includes the following steps: A1. Data Acquisition: Information on disaster situation, material supply and demand, transportation network, available transportation vehicles and their parameters, accessibility assessment data for affected areas, and damage to critical infrastructure are acquired through disaster site monitoring, feedback from emergency management departments, and drone aerial surveys. A disaster parameter database is established, using historical data or expert experience values ​​when information is missing. Confidential intervals / upper and lower bounds are set for missing data and constrained in the scheduling optimization model using support sets. Dynamic data updates are also implemented to dynamically update the database. A2. Identify key transportation nodes and determine material supply points, material forwarding points, and material demand points. Material forwarding points are selected from existing emergency warehouses, assembly points around airports / train stations, or a set of alternative candidate points. A3. Design a trunk and terminal collaborative transportation mode based on asynchronous multi-mode. Based on the geographical distribution of nodes and disaster situation, the transportation stage is divided into trunk transportation and terminal transportation. Trunk transportation includes the material supply point to the material forwarding point, by integrating high-capacity transportation methods such as railway, highway and air. Terminal transportation includes the material forwarding point to the material demand point, by designing a road plus drone delivery method. A4. Construct a three-tiered network topology: a three-tiered emergency material multimodal transport network topology based on material supply points, material forwarding points, and material demand points. Abstract it into a network graph when constructing a network. , Network diagram The set of nodes includes supply points, forward deployment points, and demand points for supplies. , To gather at the supply point For the collection of supplies at the forward deployment point For the aggregation of material demand points, there is a three-tiered emergency material multimodal transport network. individual supply points One material forwarding point and One point of demand for supplies; Network diagram The set of edges; A collection of modes of transportation This includes highways, railways, aviation, and drones. It is a collection of trunk transportation modes, namely, material supply points. to the material forwarding point A collection of transportation methods ,in For railway transportation, For air transport, This refers to a collection of last-mile transportation methods, i.e., material forwarding points. to the point of need for supplies A collection of transportation methods ,in For road transport, For drone transportation; A5. Road feasibility assessment, The roads between the material distribution point and the material demand point are located in the disaster area. Based on the traffic network information obtained from the drone assessment mission and the disaster assessment data obtained from the analysis, image recognition or data analysis methods are used to assess the degree of damage to the route and determine the post-disaster road transport capacity attenuation coefficient of the route. With traffic capacity attenuation threshold The relationship between these factors allows for the pre-determination of road transport feasibility parameters for each route. As input to the scheduling optimization model, it constrains the selection of transportation modes, is determined based on the preliminary disaster assessment, and is dynamically updated according to real-time disaster information in actual operation; For each path Determine its traffic capacity attenuation coefficient. ; like :set up This means the road is feasible; At this time, the actual traffic capacity of the highway is ; Otherwise: Settings This means the road is not feasible; in, This is a binary parameter representing the feasibility of a post-disaster recovery path. , which are input as known parameters into the scheduling optimization model; A collection of road transport routes from material distribution points to material demand points; This is the road transport capacity attenuation coefficient; This is the threshold for traffic capacity attenuation. To advance the distribution of supplies to the point of need for supplies Initial road transport capacity before the disaster; completion of the physical architecture design and parameter preparation of the emergency multimodal transport network.

3. The method for optimizing the multimodal transport scheduling of emergency supplies considering risk sensitivity as described in claim 2, characterized in that, Step B also includes the following steps: B1. Model Assumptions (1) In response to localized and regional disaster emergency scenarios, the disaster-stricken areas and their surrounding last-mile delivery networks are affected by the disaster; (2) The locations of supply points, material forwarding points and material demand points are known. Due to the need for organization, management and distribution of cross-regional emergency support, materials cannot be transported across levels. That is, emergency materials can only be transported from the material supply point to the material forwarding point, and then from the material forwarding point to the material demand point. The material supply point and the material forwarding point are close to the airport, railway station and highway. (3) The transportation from the material supply point to the material forwarding point is a long distance, and the two transportation modes are railway and air. The transportation from the material forwarding point to the material demand point is a short distance, and the two transportation modes are road and drone. (4) The material demand points are divided into those that only arrive by drone swarm and those that arrive by truck based on geographical and task attributes. The emergency material demand of each material demand point comes from the disaster assessment results. The core uncertainty of the scheduling optimization model comes from the transportation time of the last delivery. (5) The effectiveness of emergency rescue is determined by the time of the latest material demand point to be rescued. Therefore, the maximum rescue completion time is taken as the core optimization indicator of emergency dispatch.

4. The method for optimizing the multimodal transport scheduling of emergency supplies considering risk sensitivity as described in claim 3, characterized in that, Step B also includes the following steps: B2. Defining uncertainty: the road transport time from the point of origin of the goods to the point of demand for the goods. Defined as a random variable whose probability distribution information is incomplete. ; B3. Construct moment fuzzy sets, defining uncertainty using fuzzy sets based on first-order and second-order moment information and physical support sets; estimate the expected value of transportation time for each route based on UAV disaster sentiment perception results, historical data, and disaster assessment expert experience. Upper bound of variance and feasible time for transport determined in conjunction with road physical constraints As input to moment-based fuzzy sets; construct moment fuzzy sets. , including all that satisfy the expected value equal to Variance not exceeding probability distribution .

5. The method for optimizing the multimodal transport scheduling of emergency supplies considering risk sensitivity as described in claim 1, characterized in that, Step B also includes the following steps: B6. Define constraints. (1) Constraints on the integrity of the transportation process; ； This means that for each material forwarding point, there must be a means of transportation to deliver materials from the material supply point to that material forwarding point. In other words, at least one material supply point serves the material forwarding point, which is used to ensure the integrity of the material supply point and material forwarding point link in the emergency material dispatching process. ； This means that for each material demand point, there is a means of transportation to deliver the materials from the material forwarding point to the corresponding material demand point. In other words, there is at least one material forwarding point serving the material demand point, ensuring the integrity of this key link between the material forwarding point and the material demand point in the emergency material dispatch process, and ensuring that the material demand point can receive emergency materials. In the formula, It is a binary variable, if emergency supplies By transportation From the supply point Transport to the material forwarding point The value is 1 if it is 1, otherwise it is 0. It is a binary variable, if emergency supplies By transportation From the material forward deployment point Transport to the point of demand for supplies The value is 1 if it is 1, otherwise it is 0. (2) Uniqueness constraint of mode of transport, ； ； This indicates that when transporting a certain type of emergency supplies between two fixed nodes, only one mode of transportation can be selected. (3) Material forwarding point balance constraints, ； The total amount of materials at the material forwarding point includes the total amount of materials received from each material supply point as well as the original inventory of the material forwarding point itself. In the formula, For trunk line transport volume, a continuous variable, emergency supplies By transportation From the supply point Transport to the material forwarding point The volume of transport; As a forward position for supplies China Emergency Supplies Inventory levels; As a forward position for supplies emergency supplies The total quantity; (4) Constraints on material connection at material forwarding points. ； This means that the total amount of materials sent from the material forwarding point to each material demand point shall not exceed the sum of the total amount of materials received by the center from each material supply point and its initial inventory, thus ensuring the rationality of material flow at the material forwarding point. In the formula, For end-point transportation volume, a continuous variable, emergency supplies By transportation From the material forward deployment point Transport to the point of demand for supplies The volume of transport; A collection of emergency supplies types; (5) Constraints on the amount of materials received at material forwarding points. ； This indicates that the amount of materials received by the material forwarding point does not exceed the material inventory of the material supply point; In the formula, supply point China Emergency Supplies Inventory levels; (6) Limits on the amount of materials that can be received at the material demand point. ； This represents the total quantity of materials received by each material demand point from the material forwarding point. ； This means that the amount of materials received by the material demand point must not only meet the minimum demand fulfillment rate and ensure the supply of materials to the material demand point, but also ensure that the total amount of materials transported to the demand node does not exceed its total demand. In the formula, For material demand points Emergency supplies were received from the forward distribution point. The total quantity; For material demand points emergency supplies The minimum demand satisfaction rate; For material demand points emergency supplies Demand; (7) Consistency constraints between transshipment and transportation modes, ， ； This indicates that the transshipment operation must match the transportation methods of the preceding and following stages; In the formula, It is a binary variable, if emergency supplies From the supply point via material forwarding point to the point of need for supplies Along the route, the mode of transportation changed from Convert to The value is 1 if it is 1, otherwise it is 0. (8) Calculation of transportation time, Mainline transport section: ; Last-mile transport segment: ; In the formula, For transportation methods From the supply point to the material forwarding point The shortest path mileage; For transportation methods Next, from the material forward deployment point to the point of need for supplies The shortest path mileage; For transportation methods The average speed; This refers to the railway transport time for the main transport section; For air transport time in the trunk line segment; To supply points to the material forwarding point The road network coefficient reflects the impact of disasters on ground transportation on that section of road, and the mode of transport. Including railway transportation or air transport Transportation methods Including road transport or drone transport ; For the last-mile delivery segment of drone transport time; (9) Constraints on the number of vehicles, Number of transport vehicles dispatched by the supply points ; Number of transport vehicles dispatched from the material forward deployment point ; In the formula, For transportation methods From the supply point to the material forwarding point Transporting emergency supplies The number of means of transport used; For transportation methods Next, from the material forward deployment point to the point of need for supplies Transporting emergency supplies The number of transportation vehicles used All values ​​are non-negative integers, ensuring that the loading capacity of the transport vehicle is greater than the planned cargo volume to be transported; For emergency supplies The unit weight; For transportation methods Below, the rated load-bearing capacity of a unit transportation vehicle; (10) Capacity constraints of transportation modes, Supply point transportation capacity constraints: ； Capacity constraints of material forwarding point transportation methods: ； The number of various modes of transportation vehicles dispatched from each material supply point or material forwarding point shall be limited to no more than the number of available modes of transportation. In the formula, supply point Transportation methods used Number of tools; As a forward position for supplies Last-mile transportation method used Number of tools; (11) Transportation mode constraints based on road feasibility, Road transport selection constraint: Ensures that road transport can only be selected when the road is passable; ； Road transport vehicle quantity constraint: The total number of road transport vehicles carrying all goods on each route is limited to the actual traffic capacity that the road can handle. ； Autonomous selection of drone mode: When roads are impassable or have insufficient capacity after a disaster, the scheduling optimization model selects drones to undertake the last-mile delivery under the constraints of feasibility and transport capacity. (12) Unmanned aerial vehicle (UAV) flight distance constraints, For routes using drones for transportation, flight distance constraints must be met; ； In the formula, A collection of drone transportation routes from material distribution points to material demand points; Battery consumption per unit flight distance of the drone; This represents the maximum battery capacity of the drone. If drone delivery is chosen, the battery consumption for the round trip distance must not exceed the maximum battery capacity. (13) Logical relationship constraints between transport volume and transport mode selection To ensure that the transport volume occurs only on the selected transport arc, establish a logical relationship between continuous variables and binary variables; ； ； In the formula, This represents the upper limit of the transport volume for the main transport section. This represents the upper limit of the transport volume in the final transportation segment.

6. The method for optimizing the scheduling of multimodal transport of emergency supplies considering risk sensitivity as described in claim 1, characterized in that, Step C also includes the following steps: C1. Model Solving: The scheduling optimization model is solved by using the CCG method to generate columns and constraints; C1.

1. Main problem initialization, The main problem is constructed, which minimizes the approximate form of the risk-sensitive objective function under the given set of scenarios, while satisfying the constraints of step B. In the main problem, auxiliary variables and scenario slack variables are introduced to contextualize and linearize the conditional risk value term defined in step B, thereby transforming the risk-sensitive objective function into a mixed-integer linear programming form. In the initial stage, the scenario set only includes nominal scenarios or representative scenarios with limited quantities constructed based on expected transportation time, in order to obtain an initial feasible scheduling scheme that satisfies the constraints. C1.

2. Subproblem Construction and Worst-Case Scenario Generation After obtaining the current decision for the main problem, subproblems are constructed to solve the moment fuzzy set given the decision of the main problem. The subproblem aims to maximize the objective function value under a given scheduling scheme, and solves for the worst-case transportation time that leads to the maximum rescue completion time or the maximum tail risk, while satisfying the moment information constraint and support set constraint. The subproblem is solved by dualization method to obtain one or more worst-case scenarios. C1.

3. Column and Constraint Generation Add the worst-case scenarios obtained from the subproblems to the scenario set, and then add them to the main problem. Linearization and time constraints are applied to gradually tighten the main problem's ability to characterize the worst-case scenario; then, the updated main problem is solved again to obtain a new scheduling decision; through repeated iterations of solving the main problem, generating worst-case scenarios from subproblems, and updating the main problem, the main problem is made more robust. The target value gradually approaches the optimal solution of the original risk-sensitive sub-bar optimization model; C1.

4. Termination Conditions and Convergence Criteria When the difference between the objective function value corresponding to the worst-case scenario generated by the subproblem and the objective value of the current main problem does not exceed the preset tolerance or reaches the maximum number of iterations, the CCG iteration is terminated; at this time, the solution to the main problem is regarded as the optimal or near-optimal solution of the original risk-sensitive sub-Bruker optimization model.

7. The method for optimizing the multimodal transport scheduling of emergency supplies considering risk sensitivity as described in claim 6, characterized in that, Step C also includes the following steps: C2. Scheduling scheme generation, Based on the solution to the principal problem after CCG convergence, a multi-dimensional scheduling scheme is generated, including the trunk and terminal transportation modes and route selection schemes, the material flow allocation schemes for each transportation stage, and the expected value of the maximum rescue completion time and its conditional risk value at the confidence level.

8. The method for optimizing the scheduling of multimodal transport of emergency supplies considering risk sensitivity as described in claim 7, characterized in that, Step C also includes the following steps: C3. Scheme Validation and Sensitivity Analysis C3.

1. Robustness testing of the solution. In the fuzzy set Within a certain range, the Monte Carlo method is used to generate multiple random scenarios, execute scheduling schemes, calculate actual objective function values, and compare the performance of the schemes under different scenarios. C3.

2. Parameter sensitivity analysis, through systematic parameter adjustments, provides decision-makers with a quantitative basis for risk preference: Confidence level Analysis: Different tests The tail risk control effect under the value provides a reference for setting risk tolerance; Risk weight parameters Analysis: Plot a risk-efficiency trade-off curve by varying the value from 0 to 1, clearly defining the different... The corresponding scheme characteristics are used to present decision-makers with scheme options under different risk preferences, so as to weigh and select among multiple alternatives; Supply and demand change test: Simulate demand fluctuations and supply disruption scenarios to verify the adaptability of the solution.

Images