Resource dynamic scheduling method and system for road post-disaster emergency repair

Abstract

The invention relates to a resource dynamic scheduling method and system for road post-disaster emergency repair. The method comprises the following steps: acquiring emergency repair information; the emergency repair information is input into a double-layer optimization model to obtain an optimal global scheduling strategy, the double-layer optimization model comprises an upper-layer global strategy optimization sub-model and a lower-layer loading and unloading plan dynamic optimization sub-model, and the upper-layer global strategy optimization sub-model is used for generating candidate global scheduling strategies according to the emergency repair information; performing iterative evolution on the population according to the fitness of the candidate global scheduling strategy to obtain an optimal global scheduling strategy; the lower-layer loading and unloading plan dynamic optimization sub-model executes a specific driving path plan according to the candidate global scheduling strategy and performs simulation optimization and evaluation on the loading and unloading plan to obtain the vehicle total driving time and global scheduling performance penalty of the candidate global scheduling strategy, and the vehicle total driving time and the global scheduling performance penalty serve as fitness values to be returned to the upper-layer global strategy optimization sub-model. Compared with the prior art, the scheduling efficiency and adaptability are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of emergency management and operations optimization technology, and in particular to a method and system for dynamic resource scheduling for emergency repair of highways after disasters. Background Technology

[0002] In the context of climate change, proactively preventing and mitigating disaster risks has become a global consensus. As the lifeline for emergency rescue during disasters, improving the inherent safety level and enhancing emergency response capabilities of the highway network is urgently needed. Disaster events disrupt transportation, severely impacting disaster relief operations, material transport, and daily traffic needs. Ensuring smooth traffic flow within the critical "golden 72 hours" after a disaster is paramount. Coordinating safety and development, rapidly and efficiently organizing repair resources, and repairing damaged road sections to restore the road network's functionality as quickly as possible are of paramount importance.

[0003] The highway emergency repair resource scheduling problem is an extremely complex combinatorial optimization problem, essentially a variant of the Vehicle Routing Problem (VRP). Existing research often employs traditional operations research methods or static intelligent algorithms (such as standard genetic algorithms and ant colony optimization) to solve this problem. These methods typically assume that the road network state and resource requirements remain constant throughout the scheduling cycle, i.e., performing a one-time static optimization to generate a complete scheduling plan from start to finish.

[0004] However, in real-world emergency repair scenarios, the situation is far more complex and dynamic than this:

[0005] 1) Information ambiguity: In the early stages of a disaster, information such as the specific location of the disaster site, the extent of damage, and the amount of resources required may not be entirely accurate and will be dynamically updated as the repair work progresses.

[0006] 2) Environmental dynamics: New disaster sites may appear during the emergency repair process, or a repaired section of road may be interrupted again due to secondary disasters.

[0007] 3) State Interactivity: The execution of a scheduling plan will change the system state. After a road is repaired, the topology of the entire road network and the shortest paths for vehicles will change accordingly, which will directly affect subsequent scheduling decisions.

[0008] Traditional static scheduling methods cannot effectively address the aforementioned dynamics and uncertainties. The scheduling schemes they generate are often very fragile; once the actual situation deviates from the preset model, the entire scheme may become suboptimal or even completely infeasible, leading to resource waste, missed rescue opportunities, and significantly restricting the feasibility, timeliness, and global optimality of scheduling decisions. For example, patent application CN116913090A discloses a method and system for post-disaster recovery of urban road networks considering traffic resilience. This method abstracts the urban road network into a weighted undirected graph, establishes an upper-level optimization decision model and a lower-level traffic assignment model, and uses the discrete particle swarm optimization algorithm and the Frank-Wolfe algorithm to iteratively solve the problem, generating a static repair sequence for damaged road sections and a scheduling scheme for engineering teams. This method can only perform a one-time global static optimization based on complete information in the early post-disaster period. It cannot perform online rescheduling or dynamic feedback adjustments based on uncertainties such as fuzzy post-disaster information, dynamic environmental changes, and the emergence of new disaster sites. This makes the scheduling scheme fragile and prone to failure in real emergency scenarios, unable to effectively cope with the complex and ever-changing repair needs within the "golden 72 hours."

[0009] In particular, when multiple consecutive disaster sites appear on the same road segment, the material needs of some disaster sites located in the middle of the segment will face the problem of unmet or delayed fulfillment. In this case, static models are basically unable to provide feasible solutions. Therefore, there is an urgent need for a dynamic scheduling method for highway emergency repair resources that can adapt to changes in the post-disaster highway transportation system, in order to improve the feasibility and effectiveness of scheduling schemes. Summary of the Invention

[0010] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a dynamic resource scheduling method and system for emergency repair of highways after disasters, which significantly improves scheduling efficiency and adaptability.

[0011] The objective of this invention can be achieved through the following technical solutions:

[0012] A method for dynamic resource scheduling for emergency repairs after highway disasters includes the following steps:

[0013] Obtain emergency repair information;

[0014] The emergency repair information is input into a two-layer optimization model to obtain the optimal global scheduling strategy. The two-layer optimization model includes an upper-layer global strategy optimization sub-model and a lower-layer loading and unloading plan dynamic optimization sub-model. The upper-layer global strategy optimization sub-model generates candidate global scheduling strategies based on the emergency repair information and iteratively evolves a population of multiple candidate global scheduling strategies based on their fitness to obtain the optimal global scheduling strategy. The lower-layer loading and unloading plan dynamic optimization sub-model executes specific driving path planning based on the candidate global scheduling strategies and performs simulation optimization and evaluation of the loading and unloading plans to obtain the total vehicle travel time and global scheduling performance penalty of the candidate global scheduling strategies. The total vehicle travel time and global scheduling performance penalty are then returned to the upper-layer global strategy optimization sub-model as the fitness value of the candidate global scheduling strategies.

[0015] Furthermore, the emergency repair information includes multiple aspects such as the initial road network topology, capacity distribution, and resource distribution.

[0016] Furthermore, the upper-level global policy optimization sub-model uses the NSGA-II algorithm to iteratively evolve a population composed of multiple candidate global scheduling policies based on their fitness, to obtain the optimal global scheduling policy. Specific steps include:

[0017] S101. After obtaining the fitness of all candidate global scheduling strategies in the entire population, non-dominated sorting and crowding calculation are used to select excellent individuals in the population to enter the next generation.

[0018] S102. Using a sequence-oriented crossover operator, randomly select a task segment delimited by a warehouse separator from each of the parent generation P1 and parent generation P2, and then swap these two task segments to generate child generation O1 and child generation O2.

[0019] S103. Perturb the offspring sequence, the perturbation including randomly swapping the positions of two disaster points in the sequence, moving one disaster point to another position in the sequence, and randomly adding or deleting one visit to a certain disaster point.

[0020] S104. Repeat steps S101 to S103 until the maximum number of iterations is reached or the solution set converges, and obtain the optimal global scheduling strategy.

[0021] Furthermore, the lower-level loading and unloading plan dynamic optimization sub-model adopts a rolling time-domain optimization framework. After receiving a candidate global scheduling strategy from the upper-level global strategy optimization module, it performs simulation optimization and evaluation on the candidate global scheduling strategy. The specific steps include:

[0022] S201. At the current decision time t, set the prediction time domain [t, t+Tp] and the execution time domain [t, t+Te], and establish a vehicle loading and unloading plan optimization sub-problem within the prediction time domain [t, t+Tp] to obtain the vehicle access sequence;

[0023] S202. For the vehicle delivery task within the predicted time domain [t, t+Tp], solve the vehicle loading and unloading plan optimization subproblem to determine the vehicle driving path and material loading and unloading plan in the future time.

[0024] S203. Execute the vehicle travel route and material loading / unloading plan within the corresponding execution time domain [t, t+Te].

[0025] S204. After the execution time domain [t, t+Te] ends, advance the system time to t+Te, and update the road network accessibility status and the resource holding status of each disaster point, each warehouse, and each vehicle according to the plan execution results;

[0026] S205. Take t+Te as the new current decision time, return and repeat sub-steps S201 to S204 until the preset simulation end condition is met.

[0027] Furthermore, the decision variables of the vehicle loading and unloading plan optimization subproblem include the quantity of emergency repair resources loaded by vehicles at the material storage center and the quantity of emergency repair resources unloaded by vehicles at the disaster site. The constraints are the current vehicle load, the real-time inventory of the warehouse, and the real-time remaining demand at the disaster site. The objective function of the vehicle loading and unloading plan optimization subproblem is to minimize the global scheduling performance penalty.

[0028] Furthermore, the minimization of the global scheduling performance penalty is as follows:

[0029]

[0030] In the formula, Z2 is the global scheduling performance penalty, and C S The preset unit material shortage penalty factor, To determine the material shortage at disaster point f at time t, C H The pre-set penalty factor for stockpiling and leaving idle unit materials. Let F be the material redundancy at disaster point f at time t, where F is the set of disaster points and T is the time set.

[0031] Further, in step S202, based on the solution results of the vehicle loading and unloading plan optimization subproblem and the road network topology state at the corresponding time, the shortest path from the current position of the vehicle to the next node in the access sequence is calculated to obtain the vehicle's travel path in the future time.

[0032] Furthermore, in step S204, the road network accessibility status and resource holding status are updated according to the plan execution results. The specific steps include: updating the resource inventory, cumulative consumption, shortage and idle quantity of each disaster point, as well as the vehicle's movement and location, according to the executed plan execution results; determining whether any disaster point's cumulative consumption has reached its maximum demand; if so, recording the time of its reopening and updating the road network topology status accordingly.

[0033] Furthermore, the preset simulation termination condition is that all disaster-damaged points have been repaired and the road network has been fully restored.

[0034] According to another aspect of the present invention, a dynamic resource scheduling system for emergency repair of highways after disasters is provided, comprising:

[0035] The emergency repair information acquisition module is used to acquire emergency repair information.

[0036] The global scheduling strategy optimization module is used to input the emergency repair information into a two-layer optimization model to obtain the optimal global scheduling strategy. The two-layer optimization model includes an upper-layer global strategy optimization sub-model and a lower-layer loading and unloading plan dynamic optimization sub-model. The upper-layer global strategy optimization sub-model generates candidate global scheduling strategies based on the emergency repair information and iteratively evolves a population of multiple candidate global scheduling strategies based on the fitness of the candidate global scheduling strategies to obtain the optimal global scheduling strategy. The lower-layer loading and unloading plan dynamic optimization sub-model executes specific driving path planning based on the candidate global scheduling strategies and performs simulation optimization and evaluation of the loading and unloading plans to obtain the total vehicle driving time and global scheduling performance penalty of the candidate global scheduling strategies. The total vehicle driving time and global scheduling performance penalty are then returned to the upper-layer global strategy optimization sub-model as the fitness value of the candidate global scheduling strategies.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. This invention constructs a two-layer optimization model. The upper-layer global strategy optimization sub-model generates candidate global scheduling strategies based on emergency repair information. It then iteratively evolves a population of multiple candidate global scheduling strategies based on their fitness to obtain the optimal global scheduling strategy. The lower-layer loading and unloading plan dynamic optimization sub-model executes specific driving path planning based on the candidate global scheduling strategies and performs simulation optimization and evaluation of the loading and unloading plans. This yields the total vehicle travel time and global scheduling performance penalty for each candidate global scheduling strategy. The total vehicle travel time and global scheduling performance penalty are then returned to the upper-layer global strategy optimization sub-model as the fitness values ​​of the candidate global scheduling strategies. This avoids searching in a large and invalid solution space, significantly reduces the solution space, improves the algorithm's convergence speed and solution quality, and significantly improves scheduling efficiency.

[0039] 2. The lower-level loading and unloading plan dynamic optimization sub-model in this invention adopts the Rolling Time-Domain Optimization (RHO) framework. After receiving a candidate global scheduling strategy from the upper-level global strategy optimization module, it performs simulation optimization and evaluation on the candidate global scheduling strategy. Based on the real-time updated road network status and resource information during the emergency repair process, it dynamically adjusts the scheduling plan, effectively responds to emergencies, and improves the robustness of the generated scheduling scheme.

[0040] 3. The two-layer optimization model in this invention adopts a multi-objective optimization algorithm and strictly adheres to real-time constraints such as vehicle capacity, inventory limits, and road network connectivity. This ensures that the final generated scheduling scheme is always executable and effective throughout the entire dynamic execution process. It can still provide high-quality feasible solutions in complex scenarios where traditional static scheduling methods cannot solve the problem (multiple consecutive disaster points in multiple road segments), thus improving the adaptability of the algorithm. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating a dynamic resource scheduling method for emergency repair of highways after disasters, as proposed in this invention.

[0042] Figure 2 The algorithm flowchart for integrating rolling temporal optimization and NSGA-II algorithm;

[0043] Figure 3 This is a schematic diagram of the chromosome crossover operator in the embodiment. Detailed Implementation

[0044] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0045] The following English abbreviations are involved:

[0046] Non-dominated Sorting Genetic Algorithm II (NSGA-II)

[0047] Horizontal Optimization (RHO)

[0048] Discrete Event Simulation (DES)

[0049] Example 1

[0050] This embodiment provides a method for dynamic resource scheduling for emergency repairs after highway disasters, such as... Figure 1 As shown, it includes the following steps:

[0051] S1. Obtain emergency repair information.

[0052] Emergency repair information includes various aspects such as the initial road network topology, capacity distribution, and resource distribution.

[0053] S2. Input the emergency repair information into the two-layer optimization model to obtain the optimal global scheduling strategy.

[0054] The two-layer optimization model consists of an upper-layer global strategy optimization sub-model and a lower-layer loading and unloading plan dynamic optimization sub-model. The upper-layer global strategy optimization sub-model generates candidate global scheduling strategies based on emergency repair information and iteratively evolves a population of multiple candidate global scheduling strategies based on their fitness to obtain the optimal global scheduling strategy. The lower-layer loading and unloading plan dynamic optimization sub-model executes specific driving path planning based on the candidate global scheduling strategies and performs simulation optimization and evaluation of the loading and unloading plans. It obtains the total vehicle travel time and global scheduling performance penalty of the candidate global scheduling strategies and returns the total vehicle travel time and global scheduling performance penalty as the fitness value of the candidate global scheduling strategies to the upper-layer global strategy optimization sub-model.

[0055] This invention employs multi-objective optimization, aiming to achieve the optimal balance between efficiency and effectiveness simultaneously. Specifically, it includes the following two objectives:

[0056] Minimize total travel time (Z1): This objective focuses on the scheduling efficiency of repair resources, striving to complete the task with the lowest possible transportation time. Its mathematical expression is as follows:

[0057]

[0058] In the formula, Z1 is the total travel time, and K, N, and T are the sets of vehicles, nodes, and times, respectively. For decision variables, ω is 1 if vehicle k travels from node i to node j at time t, and 0 otherwise; ij Let be the travel time from node i to j.

[0059] Minimize global scheduling performance penalty (Z2): This objective comprehensively evaluates the effectiveness of scheduling schemes in meeting demand, aiming to avoid resource shortages and idle waste. Its mathematical expression is as follows:

[0060]

[0061] In the formula, Z2 is the global scheduling performance penalty, and C S The preset unit material shortage penalty factor, To determine the material shortage at disaster point f at time t, C H The pre-set penalty factor for stockpiling and leaving idle unit materials. Let F be the material redundancy at disaster point f at time t, where F is the set of disaster points and T is the time set.

[0062] The core decision variables of this invention include two categories: vehicle travel route and loading / unloading plan, as shown in the table below.

[0063] Table 1 Core Decision Variables

[0064]

[0065] The two-level optimization model includes the following key constraints:

[0066] 1) Constraints on updating the resource status of disaster-damaged sites

[0067] Actual material consumption at disaster site f at time t Depends on its current stockpile of materials Compared with the current vehicle unloading volume The sum of the expected consumption Expected maximum material consumption D fmax The minimum of the differences between the total amount consumed and the cumulative amount consumed. The remaining amount at disaster point f at time t+1. Then based on the holding quantity at time t Received supply amount Consumption at time t It is dynamically updated. Its mathematical expression is as follows:

[0068]

[0069] In the formula, This represents the actual amount of resources consumed at disaster point f at time t. The amount of supplies at time t. Let K be the amount of unloaded by vehicle K at time t at disaster point f. For the projected consumption, D fmax This represents the estimated maximum consumption of materials.

[0070] 2) Constraints on updating the repair status of disaster sites

[0071] When the cumulative repair resource consumption of disaster point f reaches its demand D fmax When the damaged point is considered to have been repaired, it is considered to have been reopened. This constraint provides a precise basis for determining whether the road network topology has changed. Its mathematical expression is as follows:

[0072]

[0073] In the formula, D represents the actual material consumption at disaster point f at time t. fmax The estimated total consumption of emergency repair materials needed at disaster site f. The variable is used to mark the repair status of the disaster point. If the disaster point f has been restored to passability at or before time t, it is 1; otherwise, it is 0.

[0074] 3) Dynamic road network connectivity constraints

[0075] This constraint prohibits any vehicle from passing through a disaster point f as a transit point until the disaster point f has been repaired. The logic can be expressed as: if vehicle k visits an unrepaired disaster point at time t... Therefore, a path cannot travel from that point to any other node besides the previously mentioned path point. This constraint ensures that all path planning is based on the current, real-time, and truly traversable road network state, directly reflecting the dynamic adaptability of this invention. Its mathematical expression is as follows:

[0076]

[0077] In the formula, Is vehicle k at time t-ω? if The flag indicating when to start traveling from node i to node f is set to 1 if yes, and 0 otherwise. ω is a flag indicating whether vehicle k starts traveling from node f to node j at time t; it is 1 if yes, and 0 otherwise. if Let be the time required for a vehicle to travel from neighboring node i to node f. The variable is used to mark the repair status of the disaster point. If the disaster point f has been restored to passability at or before time t, it is 1; otherwise, it is 0.

[0078] In addition to the core constraints mentioned above, to ensure the scheduling scheme is fully feasible both physically and logically, the model also includes a series of conventional technical constraints. Those skilled in the art, based on the technical concept of this invention, can easily construct the complete mathematical form of these constraints. These conventional constraints mainly include:

[0079] 1) Vehicle route constraints

[0080] This includes stipulating that all vehicles must depart from the warehouse and eventually return to the warehouse; ensuring that the in-degree and out-degree of vehicles are consistent at any intermediate node; and guaranteeing that at any given time, a vehicle can only be in a unique state of either driving or stationary.

[0081] 2) Vehicle movement logic constraints

[0082] This includes restricting vehicles to loading supplies only at pre-designated warehouses and unloading supplies only at disaster sites; and ensuring that the transition between vehicle driving and stationary states is consistent with common sense and logic.

[0083] 3) Physical and resource constraints

[0084] This includes ensuring that the cargo capacity of a vehicle does not exceed its maximum capacity; that the total travel time or mileage of a single trip by a vehicle does not exceed the upper limit; that the total unloading volume of all vehicles does not exceed their total loading volume; and that the total shipment volume of a warehousing center does not exceed its initial inventory.

[0085] By organically combining the above objective function and constraints, this invention constructs an optimization model that can accurately reflect the dynamics of real emergency repair projects and guide the generation of high-quality scheduling schemes, providing a solid foundation for subsequent system module solutions.

[0086] like Figure 2 As shown, the execution flow of the upper-level global policy optimization sub-model, which iteratively evolves a population composed of multiple candidate global scheduling policies using the NSGA-II algorithm, is as follows:

[0087] The upper-level global strategy optimization sub-model uses a heuristic greedy strategy to generate a set of high-quality initial solutions. For each vehicle, the "proximity principle" is adopted, inducing it to prioritize visiting the nearest disaster point with resource needs; when the vehicle is empty, it goes to the nearest warehouse for resupply.

[0088] The NSGA-II algorithm employs a hierarchical, decoupled chromosome coding scheme. Each chromosome encodes only one high-level access sequence, which consists of the labels of the disaster point and warehouse, and separators used to distinguish different vehicles, defining the order in which vehicles access disaster points and the route division. The chromosomes in the population are encoded as follows: Figure 3 As shown, a chromosome is a one-dimensional integer array that encodes only the access sequence.

[0089] Disaster point identifier: represented by positive integers, such as F1, F2, ...

[0090] Warehouse separator: Represented by a specific positive integer, such as C1, C2, ... It serves as a separator between different trips of the same vehicle.

[0091] Vehicle separator: Represented by a negative integer, such as -1. It serves as a marker separating different vehicles.

[0092] An example chromosome is: [C1, F2, F1, C1, F4, F3, F1, C2, -1, C1, F4, F2, C2].

[0093] The decoding work of this encoding is entirely delegated to the lower-level module. It only defines the order of tasks. For example, the access sequence of vehicle 1 is [C1, F2, F1, C1, F4, F3, F1, C2], starting from warehouse C1, passing through disaster points F2 and F1, returning to C1 for restocking, and continuing to disaster points F4, F3, and F1 before returning to C2; the access sequence of vehicle 2 is [C1, F4, F2, C2], starting from warehouse C1, passing through disaster points F4 and F3, and returning to C2. This design completely decouples the combination and sorting of access order from the numerical decision of loading and unloading volume, greatly improving the robustness and efficiency of the algorithm.

[0094] The upper-level global policy optimization sub-model uses the NSGA-II algorithm to iteratively evolve a population of multiple candidate global scheduling policies based on their fitness, thereby obtaining the optimal global scheduling policy. Specific steps include:

[0095] S101. After obtaining the fitness of all candidate global scheduling strategies in the entire population, non-dominated sorting and crowding calculation are used to select excellent individuals in the population to enter the next generation.

[0096] S102. Using a sequence-oriented crossover operator, randomly select a task segment delimited by a warehouse separator from each of the parent generation P1 and parent generation P2, and then swap these two task segments to generate child generation O1 and child generation O2.

[0097] S103. Perturb the offspring sequence, the perturbation including randomly swapping the positions of two disaster points in the sequence, moving one disaster point to another position in the sequence, and randomly adding or deleting one visit to a certain disaster point.

[0098] S104. Repeat steps S101 to S103 until the maximum number of iterations is reached or the solution set converges, and obtain the optimal global scheduling strategy.

[0099] The lower-level loading and unloading plan dynamic optimization sub-model adopts the Rolling Time-Domain Optimization (RHO) framework. After receiving a candidate global scheduling strategy (chromosome) from the upper-level global strategy optimization module, it performs simulation optimization and evaluation on the candidate global scheduling strategy. The specific steps include:

[0100] S201. At the current decision time t, set the prediction time domain [t, t+Tp] and the execution time domain [t, t+Te], and establish a vehicle loading and unloading plan optimization subproblem within the prediction time domain [t, t+Tp] to obtain the vehicle access sequence.

[0101] The simulation time is reset to t=0, the road network is restored to its initial post-disaster state, and the access sequence table of all vehicles is decoded based on the access sequence. The entire scheduling cycle is decomposed into a series of continuous prediction time domains Tp and shorter execution time domains Te, for example, Tp=10, Te=5. Based on the currently known road network state and the vehicle travel paths passed from the upper layer, the sequence of nodes that each vehicle can access is extracted from [t, t+Tp]. For example, in this prediction time domain, the access sequence of vehicle 1 is [C1, F2, F1, C1, F4], and the access sequence of vehicle 2 is [C1, F4, F2].

[0102] S202. For the vehicle delivery task within the predicted time domain [t, t+Tp], solve the vehicle loading and unloading plan optimization subproblem to determine the vehicle driving path and material loading and unloading plan in the future time.

[0103] The decision variables for the vehicle loading and unloading plan optimization subproblem include the quantity of emergency repair resources loaded by vehicles at the material storage center and the quantity of emergency repair resources unloaded by vehicles at disaster sites. The constraints are the current vehicle load, the real-time inventory of warehouse C1, and the real-time remaining demand at disaster sites F1, F2, and F4. The objective function of the vehicle loading and unloading plan optimization subproblem is minimizing the global scheduling performance penalty. Based on the solution results of the vehicle loading and unloading plan optimization subproblem and the corresponding road network topology, the shortest path from the current location to the next node in the access sequence is calculated using Dijkstra's algorithm or A*, yielding the vehicle's future travel path.

[0104] S203. Execute the vehicle travel route and material loading and unloading plan within the corresponding execution time domain [t, t+Te].

[0105] The vehicle only executes the material scheduling plan within the execution time domain [t, t+Te] of the optimization subproblem solution, and the vehicle travels and loads / unloads materials according to this plan. For example, the vehicle can only execute a portion of the plan in the prediction time domain within [t, t+Te]: Vehicle 1 loads 60 tons of goods at C1, unloads 40 tons at disaster point F2, and travels to the next disaster point F1, and is en route to disaster point F1 at t+Te; Vehicle 2 loads 40 tons of goods at C2 and travels to disaster point F4, and is en route to disaster point F4 at t+Te.

[0106] S204. After the execution time domain [t, t+Te] ends, advance the system time to t+Te, and update the road network accessibility status and the resource holding status of each disaster point, warehouse, and vehicle according to the plan execution results.

[0107] The road network connectivity and resource availability status are updated based on the plan execution results. Specific steps include: updating the resource inventory, cumulative consumption, shortage, and idle resources at each disaster point, along with vehicle actions and locations, based on the executed plan results; determining if any disaster point's cumulative consumption has reached its maximum demand; if so, recording the time of reopening and updating the road network topology accordingly. This stage uses discrete event simulation to ensure the system state changes due to events (road segment repair, resource delivery, resource consumption, etc.). For example, when vehicle 1 arrives at F2, a state update is triggered, with the following steps:

[0108] Incident Handling: Vehicle 1 unloaded 40 tons of supplies at disaster site F2 at time 3.85;

[0109] Status variable update: The vehicle's location is updated to F2, the vehicle's load is reduced by 40 tons, the warehouse inventory remains unchanged, and the amount of received materials at disaster point F2 increases by 40 tons.

[0110] Road network topology update: Check if the accumulated supplies received by F2 have met its repair requirements. If so, announce globally that F2 has been repaired and immediately update the global road network map, setting any connected damaged road segments as passable. This update will immediately affect the route planning of all subsequent vehicles.

[0111] S205. Take t+Te as the new current decision time, return and repeat sub-steps S201 to S204 until the preset simulation end condition is met.

[0112] Rolling iteration: Based on the updated system state, the next rolling cycle begins until the preset simulation end condition is met. The preset simulation end condition is that all disaster-damaged points have been repaired and the road network has been fully restored.

[0113] Through iterative iterations of RHO, the optimal resource allocation scheme and dynamic process of road network repair under the vehicle scheduling scheme provided by the upper-level NSGA-II can be simulated and optimized. After the simulation optimization of the entire access sequence is completed, the cumulative total travel time Z1 and the total scheduling performance penalty Z2 are extracted from the simulation log, and this pair of {Z1, Z2} values ​​is returned to the upper-level global policy optimization sub-model as fitness values.

[0114] Through the above implementation methods, the present invention successfully combines the powerful global search capability of genetic algorithms with the fine dynamic decision-making capability of rolling time-domain optimization, thereby achieving an effective solution to the problem of highway emergency repair resource scheduling under complex dynamic environments.

[0115] Example 2

[0116] This embodiment provides a dynamic resource scheduling system for emergency repairs after highway disasters, including:

[0117] The emergency repair information acquisition module is used to acquire emergency repair information.

[0118] The global scheduling strategy optimization module is used to input emergency repair information into a two-layer optimization model to obtain the optimal global scheduling strategy. The two-layer optimization model includes an upper-layer global strategy optimization sub-model and a lower-layer loading and unloading plan dynamic optimization sub-model. The upper-layer global strategy optimization sub-model generates candidate global scheduling strategies based on the emergency repair information and iteratively evolves a population of multiple candidate global scheduling strategies based on their fitness to obtain the optimal global scheduling strategy. The lower-layer loading and unloading plan dynamic optimization sub-model executes specific driving path planning based on the candidate global scheduling strategies and performs simulation optimization and evaluation of the loading and unloading plans. It obtains the total vehicle travel time and global scheduling performance penalty of the candidate global scheduling strategies and returns the total vehicle travel time and global scheduling performance penalty as the fitness value of the candidate global scheduling strategies to the upper-layer global strategy optimization sub-model.

[0119] The system may also include a data interaction interface for receiving real-time dynamic information such as road network status updates and new disaster reports from external data sources, such as traffic management centers, meteorological departments, and handheld terminals of on-site repair teams.

[0120] The system can also include a user interface module to display the Pareto optimal solution set to decision-makers and present the final selected scheduling scheme in a visual manner, such as an electronic map or Gantt chart. Decision-makers can issue scheduling instructions to the terminal devices of each repair vehicle through this interface.

[0121] The rest is the same as in Example 1.

[0122] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for dynamic resource scheduling for emergency repair of highways after disasters, characterized in that, Includes the following steps: Obtain emergency repair information; The emergency repair information is input into a two-layer optimization model to obtain the optimal global scheduling strategy. The two-layer optimization model includes an upper-layer global strategy optimization sub-model and a lower-layer loading and unloading plan dynamic optimization sub-model. The upper-layer global strategy optimization sub-model generates candidate global scheduling strategies based on the emergency repair information and iteratively evolves a population of multiple candidate global scheduling strategies based on their fitness to obtain the optimal global scheduling strategy. The lower-layer loading and unloading plan dynamic optimization sub-model executes specific driving path planning based on the candidate global scheduling strategies and performs simulation optimization and evaluation of the loading and unloading plans to obtain the total vehicle travel time and global scheduling performance penalty of the candidate global scheduling strategies. The total vehicle travel time and global scheduling performance penalty are then returned to the upper-layer global strategy optimization sub-model as the fitness value of the candidate global scheduling strategies.

2. The resource dynamic scheduling method for emergency repair of highways after disasters according to claim 1, characterized in that, The emergency repair information includes various aspects such as the initial road network topology, capacity distribution, and resource distribution.

3. The resource dynamic scheduling method for emergency repair of highways after disasters according to claim 1, characterized in that, The upper-level global policy optimization sub-model uses the NSGA-II algorithm to iteratively evolve a population composed of multiple candidate global scheduling policies based on their fitness, to obtain the optimal global scheduling policy. Specific steps include: S101. After obtaining the fitness of all candidate global scheduling strategies in the entire population, non-dominated sorting and crowding calculation are used to select excellent individuals in the population to enter the next generation. S102. Using a sequence-oriented crossover operator, randomly select a task segment delimited by a warehouse separator from each of the parent generation P1 and parent generation P2, and then swap these two task segments to generate child generation O1 and child generation O2. S103. Perturb the offspring sequence, the perturbation including randomly swapping the positions of two disaster points in the sequence, moving one disaster point to another position in the sequence, and randomly adding or deleting one visit to a certain disaster point. S104. Repeat steps S101 to S103 until the maximum number of iterations is reached or the solution set converges, and obtain the optimal global scheduling strategy.

4. The resource dynamic scheduling method for emergency repair of highways after disasters according to claim 1, characterized in that, The lower-level loading and unloading plan dynamic optimization sub-model adopts a rolling time-domain optimization framework. After receiving a candidate global scheduling strategy from the upper-level global strategy optimization module, it performs simulation optimization and evaluation on the candidate global scheduling strategy. The specific steps include: S201. At the current decision time t, set the prediction time domain [t, t+Tp] and the execution time domain [t, t+Te], and establish a vehicle loading and unloading plan optimization sub-problem within the prediction time domain [t, t+Tp] to obtain the vehicle access sequence; S202. For the vehicle delivery task within the predicted time domain [t, t+Tp], solve the vehicle loading and unloading plan optimization subproblem to determine the vehicle driving path and material loading and unloading plan in the future time. S203. Execute the vehicle travel route and material loading / unloading plan within the corresponding execution time domain [t, t+Te]. S204. After the execution time domain [t, t+Te] ends, advance the system time to t+Te, and update the road network accessibility status and the resource holding status of each disaster point, each warehouse, and each vehicle according to the plan execution results; S205. Take t+Te as the new current decision time, return and repeat sub-steps S201 to S204 until the preset simulation end condition is met.

5. The resource dynamic scheduling method for emergency repair of highways after disasters according to claim 4, characterized in that, The decision variables of the vehicle loading and unloading plan optimization subproblem include the quantity of emergency repair resources loaded by vehicles at the material storage center and the quantity of emergency repair resources unloaded by vehicles at the disaster site. The constraints are the current vehicle load, the real-time inventory of the warehouse, and the real-time remaining demand at the disaster site. The objective function of the vehicle loading and unloading plan optimization subproblem is to minimize the global scheduling performance penalty.

6. The resource dynamic scheduling method for emergency repair of highways after disasters according to claim 5, characterized in that, The minimum global scheduling performance penalty is: In the formula, Z2 is the global scheduling performance penalty, and C S The preset unit material shortage penalty factor, To determine the material shortage at disaster point f at time t, C H The pre-set penalty factor for stockpiling and leaving idle unit materials. Let F be the material redundancy at disaster point f at time t, where F is the set of disaster points and T is the time set.

7. The resource dynamic scheduling method for emergency repair of highways after disasters according to claim 4, characterized in that, In step S202, based on the solution results of the vehicle loading and unloading plan optimization subproblem and the road network topology state at the corresponding time, the shortest path from the current position of the vehicle to the next node in the access sequence is calculated to obtain the vehicle's travel path in the future time.

8. The resource dynamic scheduling method for emergency repair of highways after disasters according to claim 4, characterized in that, In step S204, the road network accessibility status and resource holding status are updated according to the plan execution results. The specific steps include: updating the resource inventory, cumulative consumption, shortage and idle quantity of each disaster point, as well as the vehicle's movement and location, according to the executed plan execution results; determining whether any disaster point's cumulative consumption has reached its maximum demand; if so, recording the time of its reopening and updating the road network topology status accordingly.

9. The resource dynamic scheduling method for emergency repair of highways after disasters according to claim 4, characterized in that, The preset simulation termination condition is that all disaster-damaged points have been repaired and the road network has been fully restored.

10. A dynamic resource scheduling system for emergency repair of highways after disasters, characterized in that, include: The emergency repair information acquisition module is used to acquire emergency repair information. The global scheduling strategy optimization module is used to input the emergency repair information into a two-layer optimization model to obtain the optimal global scheduling strategy. The two-layer optimization model includes an upper-layer global strategy optimization sub-model and a lower-layer loading and unloading plan dynamic optimization sub-model. The upper-layer global strategy optimization sub-model generates candidate global scheduling strategies based on the emergency repair information and iteratively evolves a population of multiple candidate global scheduling strategies based on the fitness of the candidate global scheduling strategies to obtain the optimal global scheduling strategy. The lower-layer loading and unloading plan dynamic optimization sub-model executes specific driving path planning based on the candidate global scheduling strategies and performs simulation optimization and evaluation of the loading and unloading plans to obtain the total vehicle driving time and global scheduling performance penalty of the candidate global scheduling strategies. The total vehicle driving time and global scheduling performance penalty are then returned to the upper-layer global strategy optimization sub-model as the fitness value of the candidate global scheduling strategies.

Images