Airport ground support vehicle collaborative scheduling optimization method and system

Abstract

The invention discloses an airport ground support vehicle cooperative scheduling optimization method and system, and belongs to the technical field of airport ground support vehicle cooperative scheduling, and the method comprises the steps: S1, selecting a model according to an optimization target; s2, for the selected model, performing model solution and feasible initial solution generation: relaxing vehicle constraints, and setting a large virtual vehicle upper limit for each motorcade to ensure that a solution space is feasible; changing an optimization target, and temporarily minimizing the total number of vehicles in the motorcade; aLNS is operated, iterative optimization is carried out by using a destruction operator and a repair operator, and the number of vehicles is reduced; terminating when the number of vehicles in each fleet does not exceed the actual number, and obtaining a feasible solution; target driving optimization: taking a feasible solution as a current solution, after an original target is recovered, running an enhanced ALNS for optimization, adaptively selecting a damage and repair operator, iteratively executing damage and repair operation, receiving a new solution based on a simulated annealing criterion, and dynamically updating the weight of the operator; and finally outputting an optimal scheduling scheme. And S3, outputting and executing an optimal scheduling scheme.

Description

Technical Field

[0001] This invention belongs to the field of airport ground support vehicle collaborative scheduling technology, and particularly relates to an optimization method and system for airport ground support vehicle collaborative scheduling. Background Technology

[0002] With the continued rapid development of the global civil aviation industry and the significant recovery in air transport demand, the complexity and scale of airport ground handling operations are expanding rapidly. Air traffic volume has gradually recovered to pre-pandemic levels and maintained stable growth, making airport ground handling tasks increasingly demanding in the current operating environment, placing higher requirements on scheduling efficiency, real-time performance, and coordination. Ground handling operations encompass multiple key aspects, including refueling, cabin cleaning, catering, baggage and cargo handling, and passenger shuttles. They require the coordinated operation of various types of specialized vehicles, and there are strict temporal and spatial constraints between different operations, resulting in extremely high scheduling complexity. Currently, among the many factors affecting flight punctuality, besides uncontrollable weather conditions, delays related to airport ground handling operations—including excessively long waiting times for ground handling vehicles, inefficient resource allocation, and disconnects between multiple services—have become one of the most significant controllable factors causing flight delays. This phenomenon directly indicates that a lack of systematic and coordinated scheduling will significantly drag down overall operational efficiency and may even directly translate into flight delays. Against this backdrop, how to build a scientific, efficient, real-time ground vehicle dispatching system that can coordinate multiple types of resources has become a core challenge for improving airport operational efficiency and management level.

[0003] In the field of airport ground support vehicle dispatching, the existing technical solutions most similar to this invention mainly include the following categories, each of which has its inherent and serious defects that urgently need to be overcome:

[0004] Heuristic algorithms based on single-job optimization (such as genetic algorithms and traditional vehicle routing problem solvers) are described as follows: These methods typically model the scheduling problem of a specific type of support vehicle in an airport (e.g., only for refueling trucks or only for shuttle buses) as a classic Vehicle Routing Problem (VRP) or its common variations (e.g., VRP with time windows), and use metaheuristic algorithms such as genetic algorithms, simulated annealing, and tabu search to solve it. Their optimization objectives are often relatively singular, typically focusing on minimizing the total travel distance of this type of vehicle or minimizing the task completion time.

[0005] Existing flaws: Isolated optimization, severely lacking in synergy: This type of method only considers the scheduling optimization of a single type of vehicle from a local perspective, completely ignoring the strict temporal dependencies between different support service types (e.g., cabin cleaning can only be carried out after passengers have disembarked and baggage has been unloaded, and catering can only be carried out after cleaning is completed; aircraft refueling must be carried out after most ground operations are completed and before pushback) and the problem of space resource contention in shared areas. This local optimization mode often leads to scheduling conflicts between subsystems, not only failing to achieve global optimization, but also making it difficult to form a practically executable collaborative operation plan. As a result, in actual operation, problems such as disordered service order, vehicles blocking each other, and delays in critical tasks occur, ultimately causing flight delays.

[0006] Based on multi-objective mathematical programming models and exact solvers (such as CPLEX and Gurobi), the approach describes a method where some studies attempt to integrate scheduling requirements by constructing mixed-integer linear programming (MILP) models that include multiple vehicle types and constraints, and then solve them using commercial mathematical programming solvers such as CPLEX and Gurobi. These models may simultaneously consider multiple objectives such as vehicle operating costs, total service time, and partial resource utilization, attempting to achieve comprehensive mathematical optimization.

[0007] Existing drawbacks: The computational complexity grows exponentially, making it difficult to use for actual real-time scheduling: Airport ground support scheduling is a typical NP-hard problem. As the number of flights (e.g., more than 20 flights / hour) and constraints increase, the problem size expands dramatically. The computation time of the exact solver can even grow exponentially. Often, due to memory overflow or computation timeout, it is impossible to obtain a feasible solution within the limited time required for actual airport operation (usually requiring the generation of a scheduling plan within a few minutes), which seriously fails to meet the requirements of real-time response.

[0008] Sensitive to initial feasible solutions and poor robustness: When dealing with highly constrained complex models, the performance of the exact solver is extremely dependent on the quality of the initial solution or the model relaxation strategy. Under strong constraints, obtaining only the first feasible solution may require a significant amount of computation time, becoming the bottleneck of the entire solution process. Furthermore, if the operating environment or input data changes slightly, the original solution may quickly become invalid, resulting in poor system adaptability.

[0009] A single-objective scheduling method based on the standard Adaptive Large Neighborhood Search (ALNS) algorithm is described as follows: This scheme generally models the ground support scheduling problem as a vehicle routing problem with a time window (VRPTW). Its core operating mechanism is as follows: an initial scheduling scheme is generated using a simple heuristic rule. Then, through iteration, a set of disruptive operators are used to randomly remove some assigned tasks. Repair operators are then used to reinsert the removed tasks into the solution. During this process, new solutions are typically accepted according to simulated annealing criteria, and the weights of the operators are adaptively adjusted based on their historical performance. Its optimization objective is mostly limited to minimizing the total travel distance of all vehicles.

[0010] Existing shortcomings: The initial solution generation mechanism is unreliable: In complex, multi-constraint real-world applications, simple heuristic methods often fail to construct feasible initial solutions that satisfy all constraints, causing the entire optimization search process to fail to start effectively or get stuck in local optima with low convergence efficiency. The optimization objective is too singular: Focusing solely on vehicle travel distance as the primary optimization objective fails to directly address the most critical need in airport operations—minimizing flight delays—leading to a significant deviation from actual operational goals. The multi-service collaborative optimization capability is weak: The standard ALNS algorithm inherently lacks the ability to handle complex priority constraints and collaborative requirements among multiple service types. It lacks an effective global coordination mechanism and cannot provide trade-offs between different objectives, limiting the feasibility and practicality of the overall scheduling scheme. Summary of the Invention

[0011] To address the shortcomings of existing technologies in the field of airport ground support vehicle dispatching, this invention aims to solve the following four specific technical problems: Existing scheduling models suffer from incomplete constraint characterization and a singular optimization objective. Current methods typically only consider constraints such as vehicle capacity and time windows, failing to systematically integrate complex constraints in actual operations, such as flight service priority and multi-fleet operational coordination. This results in poor feasibility of scheduling solutions in practice. Furthermore, their optimization objectives are often limited to minimizing vehicle travel distance, failing to optimize for minimizing flight delay time as the direct objective, thus failing to meet the core requirement of airport operations for flight punctuality.

[0012] Generating initial feasible solutions to large-scale scheduling problems under strong constraints is a challenge. Under multiple strict constraints such as vehicle resources, time windows, and service order, the simple construction methods relied upon by traditional heuristic algorithms (such as the standard adaptive large neighborhood search algorithm) may fail to generate feasible initial scheduling schemes. This can lead to the optimization algorithm failing to start or providing extremely poor-quality initial solutions, severely impacting the convergence speed of subsequent optimization processes and the quality of the final solution.

[0013] Solving large-scale, complex scheduling problems is computationally inefficient and fails to meet real-time requirements. In large-scale scenarios involving the coordinated scheduling of dozens or even hundreds of flights and various support vehicles, the solution time of accurate solvers (such as CPLEX) increases exponentially, and they may even be unable to find a feasible solution within an acceptable time. Therefore, they cannot be applied to dynamic or real-time scheduling scenarios that require rapid response.

[0014] The problem lies in the lack of ability to effectively balance and coordinate multiple objectives. Existing methods lack the capacity to automatically generate a series of optimal trade-offs between the two key and conflicting objectives of "operating costs" (such as travel distance) and "service effectiveness" (such as flight delays). Decision-makers are unable to flexibly select different preferred scheduling schemes based on real-time operational conditions (such as peak and off-peak periods).

[0015] To achieve the above-mentioned objectives, the first objective of this invention is to provide a method for optimizing the coordinated scheduling of airport ground support vehicles, comprising: S1. Select the model based on the optimization objective: When the optimization objective is to minimize the total distance traveled by vehicles, the distance-optimized scheduling model is selected. When the optimization objective is to minimize the total delay time, the delay time optimization scheduling model is selected; When the optimization objective combines the total distance traveled by the vehicle and the total delay time, the Pareto multi-objective optimization model is selected. S2. For the selected model, solve the model, including: S201. Generation of feasible initial solution: Relax vehicle constraints and set a large virtual vehicle limit for each fleet k to ensure the solution space is feasible; change the optimization objective to temporarily minimize the total number of vehicles in the fleet; run ALNS and use destruction and repair operators to iteratively optimize and reduce the number of vehicles; terminate when the number of vehicles in each fleet does not exceed the actual number, and obtain a feasible solution. S202, Goal-Driven Optimization: Taking the feasible solution as the current solution, after restoring the original goal, the enhanced ALNS is run for optimization, adaptively selecting destruction and repair operators, iteratively executing destruction and repair operations, accepting new solutions based on the simulated annealing criterion, and dynamically updating operator weights; finally, the optimal scheduling scheme is output. S3. Output and execution of the optimal scheduling scheme.

[0016] A second objective of this invention is to provide an airport ground support vehicle collaborative scheduling optimization system, comprising: The human-computer interaction module allows users to input basic data, select a model, and display optimization results. The model database stores the scheduling model based on the shortest total vehicle travel distance, the scheduling model based on the minimum total delay time, and the Pareto multi-objective optimization model that integrates the total vehicle travel distance and the total delay time. The scheduling optimization module imports basic data into the selected model for model solving, including: Generate feasible initial solution: Relax vehicle constraints and set a large virtual vehicle limit for each fleet k to ensure the solution space is feasible; change the optimization objective to temporarily minimize the total number of vehicles in the fleet; run ALNS and use the destruction operator and repair operator to iteratively optimize and reduce the number of vehicles; terminate when the number of vehicles in each fleet does not exceed the actual number, and obtain a feasible solution; Goal-driven optimization: Using the feasible solution as the current solution, after restoring the original goal, the enhanced ALNS is run for optimization, adaptively selecting destruction and repair operators, iteratively executing destruction and repair operations, accepting new solutions based on the simulated annealing criterion, and dynamically updating operator weights; finally, the optimal scheduling scheme is output.

[0017] A third objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned airport ground support vehicle collaborative scheduling optimization method.

[0018] A fourth objective of this invention is to provide a computer program product, including a computer program that, when executed by a processor, implements the aforementioned airport ground support vehicle collaborative scheduling optimization method.

[0019] The advantages and positive effects of this application are: The initial solution generation method in this invention is robust, laying a solid and reliable foundation for the overall optimization process. By introducing an innovative two-stage algorithm framework, the first stage is specifically designed to optimize the scheduling feasibility problem under strong constraints, stably and efficiently generating high-quality initial scheduling schemes that meet all hard constraints. This mechanism effectively overcomes the common problems of traditional optimization algorithms failing to start the optimization process due to infeasible or low-quality initial solutions, or prematurely getting trapped in local optima. Therefore, it provides a superior starting point for the second stage, the deep optimization stage, significantly improving the overall algorithm's convergence efficiency and solution quality.

[0020] This invention achieves multi-objective collaborative optimization, significantly enhancing the system's decision support capabilities. It not only supports optimization modes with a single objective of minimizing total travel distance or total delay time, but also, by integrating an advanced Pareto optimization mechanism, automatically generates a series of non-dominated solution sets—the Pareto front—that achieve a balance between the two key performance indicators of "operating costs" and "flight punctuality." This function provides airport dispatchers with a wealth of scientific decision-making options in different operational scenarios. For example, during peak flight periods, priority can be given to ensuring punctuality, while during off-peak periods, the focus can be on reducing operating costs, thus flexibly responding to various practical needs.

[0021] This invention offers extremely high solution efficiency, effectively addressing the computational challenges of large-scale real-time scheduling tasks. Compared to traditional exact solvers such as CPLEX, the two-stage adaptive large neighborhood search algorithm (II-ALNS) employed in this invention significantly improves computational speed while maintaining high-quality solutions. The algorithm's runtime increases relatively smoothly with the problem size, performing excellently even when handling large-scale scheduling instances with more than 60 flights, demonstrating its potential for application in real-time and even dynamic scheduling environments.

[0022] The algorithm architecture of this system possesses high flexibility and scalability, enabling it to adapt to various practical application scenarios. The core algorithm integrates multiple destruction and repair operators and dynamically adjusts the search strategy through an adaptive weighting mechanism to respond to different search states and problem characteristics at different stages. This design allows the algorithm to flexibly adjust according to the operational characteristics, constraint combinations, and scheduling scale of different airports, demonstrating excellent versatility and scalability, and providing a solid technical foundation for subsequent functional expansion and application scenario migration. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A flowchart of a preferred embodiment of the present invention is shown; Figure 2 A flowchart of S2 in a preferred embodiment of the present invention is shown; Figure 3 The diagram shows the framework of the II-ALNS algorithm in a preferred embodiment of the present invention; Figure 4 A simplified diagram of the airport layout in a preferred embodiment of the present invention is shown; Figure 5This diagram illustrates the second-stage convergence process of the total vehicle travel distance of the II-ALNS algorithm under different flight scales in a preferred embodiment of the present invention. Figure 6 The diagram illustrates the convergence process of the second-stage delay time of the II-ALNS algorithm under different flight sizes in a preferred embodiment of the present invention. Figure 7 The Pareto front plot for dual-objective optimization under a scale of 60 flights in a preferred embodiment of the present invention is shown. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The core of this invention is to provide an optimization method for the collaborative scheduling of airport ground support vehicles based on Two-stage Adaptive Large Neighborhood Search (II-ALNS). This method achieves a dynamic balance between global exploration and local development by combining an adaptive mechanism and a two-stage search strategy. This provides an efficient and robust solution to the complex constraints and optimization objectives in airport ground support vehicle scheduling, such as minimizing total vehicle travel distance and / or total delay time, maximizing resource utilization, and ensuring task coordination. Through an innovative algorithm framework design, this solution systematically addresses key issues raised in the background technology, such as resource allocation conflicts, time window compliance, multi-vehicle coordination, and real-time scheduling adjustments, significantly improving the efficiency and reliability of airport ground operations. The overall technical solution follows a logical thread of "modeling-model solving." In the modeling stage, the actual scheduling problem is transformed into a mathematical optimization model containing objective functions, decision variables, and constraints. In the model solving stage, the II-ALNS algorithm is used to iteratively improve the solution quality by adaptively selecting destruction and repair operators. The specific description is as follows: Please see Figures 1 to 7 The first embodiment, an optimization method for collaborative scheduling of airport ground support vehicles, mainly includes: S1. Select the model based on the optimization objective: When the optimization objective is to minimize the total distance traveled by vehicles, the distance-optimized scheduling model is selected. When the optimization objective is to minimize the total delay time, the delay time optimization scheduling model is selected; When the optimization objective combines the total distance traveled by the vehicle and the total delay time, the Pareto multi-objective optimization model is selected. S2. For the selected model, solve the model, including: S201. Generation of feasible initial solution: Relax vehicle constraints and set a large virtual vehicle limit for each fleet k to ensure the solution space is feasible; change the optimization objective to temporarily minimize the total number of vehicles in the fleet; run ALNS and use destruction and repair operators to iteratively optimize and reduce the number of vehicles; terminate when the number of vehicles in each fleet does not exceed the actual number, and obtain a feasible solution. S202, Goal-Driven Optimization: Taking the feasible solution as the current solution, after restoring the original goal, the enhanced ALNS is run for optimization, adaptively selecting destruction and repair operators, iteratively executing destruction and repair operations, accepting new solutions based on the simulated annealing criterion, and dynamically updating operator weights; finally, the optimal scheduling scheme is output. S3. Output and execution of the optimal scheduling scheme.

[0027] To better understand the technical solution of the present invention, the following non-limiting description is provided: System Modeling and Problem Definition The core objective of this technical solution is to ensure the smooth execution of flight ground support activities. Specifically, it requires allocating necessary ground resources for each flight. These resources are transported to the corresponding flight by specific support vehicles within a limited time window to complete the service operations. Due to the large number of flights to be supported, and the fact that each flight typically requires multiple different types of services, this problem can essentially be modeled as a multi-fleet vehicle routing problem with resource capacity constraints and time window limitations. The problem is defined on an undirected graph. Above, among which It is a collection of garage and machine station nodes. , node Represents the garage node. Represents the machine station node. This represents the set of edges between nodes.

[0028] convoy assembly , Each team is responsible for Different services, each provided by a separate fleet, and when Time indicates The priority of the service is higher than Each team have A fleet of vehicles providing the same service vehicle assembly They have the same capacity .

[0029] For the team One of the cars in the middle, It indicates that it is a flight The time required to provide the service It indicates that it comes from the node Drive to the node Time required. For flights Service hours. Represents a node Regarding the demand for resources, .

[0030] Each edge There is a related cost. .

[0031] This problem involves two types of decision variables, the first type of decision variable For a binary variable, if the team Vehicles in From node Drive to the node ,but =1, otherwise 0. The second type is a real number variable. It represents the convoy Vehicles in For flights The start time of service provision. These two variables determine the space and time for ensuring vehicle availability, respectively. For ease of explanation, the meanings of the mathematical symbols used in this invention are shown in Table 1.

[0032] Table 1 Meaning of Mathematical Symbols

[0033] A mathematical model that optimizes by minimizing the total distance traveled by the vehicle: (1) (2) (3) (4) (5) (6) (7)

[0034]

[0035]

[0036]

[0037]

[0038] in: Let be the cost function, representing the travel cost of fleet k on edge (i, j), where i and j are the gate nodes where the flight is located; For decision variables, the value is 1 if vehicle v in convoy k travels from gate node i to gate node j, and 0 otherwise; For the set of car teams k; , represents a specific vehicle in vehicle k of the convoy; This is a set of garage and machine station nodes; For the set of machine station nodes; As a decision variable, if vehicle v in convoy k travels from gate node i to gate node... The value is 1 if it is 1, otherwise it is 0. As a decision variable, if vehicle v in fleet k is from the machine station node The value is 1 if the vehicle reaches node j, otherwise it is 0. As a decision variable, if vehicle v in fleet k originates from garage node The value is 1 if the vehicle reaches node j, otherwise it is 0. Let vehicle v in fleet k be the decision variable, and if vehicle v travels from gate node i to garage node i... The value is 1 if it is 1, otherwise it is 0. Let be a real number decision variable, representing the start time of service for vehicle v in fleet k to flight at gate node j; Let be a real number decision variable, representing the start time of service for vehicle v in fleet k on the flight at gate node i; The duration of service for fleet k to flights at gate node i; Let be the time it takes for vehicle k to travel from node i to node j; The maximum time a vehicle v in fleet k is allowed to remain at its stand after completing service; The resource requirements of the flight at gate node i; The maximum amount of resources carried by vehicles in fleet k; The duration of service for fleet k to flights at gate node i; Let be the time it takes for vehicle k to travel from node i to node j; This is the lower bound of the arrival time of vehicles in convoy k at gate node i; Let be a real number decision variable, representing the fleet. The start time of service for vehicle v on the flight at gate node i; For the team The duration of service for the flight at gate node i; Let be a real number decision variable, representing the fleet. The start time of service for vehicle v on the flight at gate node i; This is the upper bound of the time when vehicles in fleet k arrive at station node i.

[0039] (1) The objective function aims to minimize the total cost (distance). (2) Ensures that each flight is served by only one vehicle from each fleet. (3) Guarantees route continuity. (4) Guarantees that the number of vehicles used in each fleet does not exceed the number of available vehicles. (5) Guarantees that the number of vehicles entering and leaving the garage is equal. (6) Limits the maximum time a vehicle is allowed to remain at the gate after completing its service. (7) Determines the capacity constraints for each fleet. (8) and (9) ensure that the service start time is within their respective time windows. Equation (10) stipulates that if there is a priority relationship, the operation of one fleet must precede the operation of another fleet.

[0040] Delay time optimization scheduling models have significant practical value in actual airport operations. To simplify the problem, this invention focuses only on the impact of airport ground support on flight delays. Based on the Civil Aviation Administration's flight safety operation standards, this invention defines a flight delay as follows: for a given flight, the latest of the following three times—catering completion time, passenger door closing time, and cargo door closing time—is subtracted from the expected wheel chock removal time. If the subtraction is greater than 0, the flight's support activities are delayed; if it is less than or equal to 0, no delay has occurred. The scheduling optimization objective is to minimize the total flight support activity delay time. For flights without delays, the delay time is 0, not a negative number. Only the catering completion time, passenger door closing time, and cargo door closing time are considered because these three activities are the latest to be completed.

[0041] The following modifications are made to the mathematical model of this invention. New constraints are added: Calculation of completion time for each service (13) (14)

[0042] Calculate the delay time: Introducing auxiliary variables Indicates the latest completion time of flight i:

[0043]

[0044]

[0045] Introducing binary variables When flight i is delayed ,otherwise ; The objective function is: ,

[0046]

[0047]

[0048]

[0049] in: for ; for ; for ; Let v be a real number decision variable, representing the start time of service for vehicle v in the catering fleet to the flight at gate node i; The duration of catering service for flights at gate node i; Let v be a real number decision variable, representing the start time of service for vehicle v in the freight truck fleet to the flight at gate node i; The duration of the cargo service for flights at gate node i; Let v be a real number decision variable, representing the start time of service for vehicle v in the shuttle bus fleet to the flight at gate node i; The duration of the shuttle service for flights at gate node i; For camera position ; M is ; for ; Objective function: ,

[0050] Retain constraints (2)-(12); Add new constraints (13)-(21); Ensure by constraints (16)-(18) The delay time is equal to the maximum of the three service completion times; constraints (19)-(21) ensure the delay time. ;when This indicates no delay; when hour, This indicates a delay.

[0051] The time window constraint of the model is a soft constraint, which means that the vehicles may arrive earlier or later than expected.

[0052] The revised model shifts the focus of scheduling from minimizing travel distance to minimizing flight delays, better reflecting the actual needs of airport operations.

[0053] The Pareto multi-objective optimization model addresses the inherent multi-objective nature of airport ground support scheduling, where decision-makers must simultaneously weigh operating costs (represented by travel distance) against service efficiency (represented by delay time). To characterize the inherent conflict between these dual objectives, this invention constructs the following dual-objective optimization model:

[0054]

[0055] in:

[0056]

[0057] here, The objective is a two-dimensional vector. The solution to this problem is not a single solution, but rather a set of mutually weighing solutions within the objective space. For vector optimization problems, the traditional concept of "optimal solution" is replaced by "non-dominated solution" or "Pareto optimal solution".

[0058] Domination relationship: Let , There are two feasible scheduling schemes. If the following conditions are met: and Then it is called a solution. Dominant Solution .

[0059] Pareto optimal solution: If a feasible solution Not by any other feasible solution What is controlled is called It is a Pareto optimal solution (or a non-dominated solution).

[0060] Pareto Front: The objective vectors corresponding to all Pareto optimal solutions lie in the objective space. The set formed by these elements is called the Pareto front (PF).

[0061] Each point on the Pareto front represents an optimal trade-off between total travel distance and total delay time, which cannot be optimized simultaneously.

[0062] Multi-objective optimization strategy and scalarization method: Since directly solving the analytical solution of the Pareto front is computationally infeasible (NP-hard problem), this invention adopts a decomposition-based heuristic search strategy. Its core idea is to transform the original bi-objective problem into a series of structured single-objective sub-problems for collaborative optimization. Specifically, the Chebyshev scalarization method is used because it has good approximation ability for non-convex fronts.

[0063] For a given weight vector ,in and and a reference point = Construct the following Chebyshev scaled quantum problem:

[0064]

[0065] in,

[0066] It can be proven that, under mild conditions, by uniformly sampling a set of weight vectors The corresponding subproblems are solved separately, and the resulting solution set can approximate the real Pareto front well in the target space.

[0067] In solving the model, this invention not only addresses practical constraints in airport ground support, such as time windows, capacity, or priority constraints, but also handles the heterogeneous operations of a large number of aircraft. This invention also notes that the problem definition is not merely an extension of the traditional path planning problem, but rather a good reflection of the complex background of actual basic ground support.

[0068] All constraints in the formula have their own physical meaning. Meanwhile, airport ground support has its own unique properties; for example, it typically requires decisions regarding operational routes and start times, making it more practical and challenging than traditional path planning problems. Solving the airport ground support problem in such a complex context is quite challenging. Therefore, this invention uses the heuristic algorithm ALNS to solve this problem.

[0069] The core idea is: destructive solution, repair solution, dynamic weight adjustment and selection (adaptive).

[0070] In large-scale, multi-constraint real-world problems, initial solution generation is often more difficult than the optimization process. To address the issue that the general ALNS algorithm's initial solution generation method may fail to produce an initial solution, this paper proposes an improved two-stage ALNS framework to solve the initial solution generation problem. Finally, seven breaking operators are used to perform breaking operations. This variety of breaking operations significantly expands the feasible solution space, making the locally optimal solutions obtained by the algorithm more likely to approximate the global optimum.

[0071] General Adaptive Large Neighborhood Search Algorithm ALNS By designing multiple sets of destruction and repair operators, the search range of the solution space is expanded, and the current solution is improved. The destruction and repair methods with good performance receive higher scores and higher weights.

[0072] In each iteration, the various destruction and repair operators are selected and their weights adjusted based on past performance. An efficient combination method is used to improve the algorithm's optimization ability, thereby finding the optimal solution.

[0073] ALNS consists of destruction and repair operations. The destruction operation involves removing the flight nodes of the original solution, typically removing 40% of the nodes. The repair operation attempts to find a new solution using repair methods. If the obtained solution is better, it is updated, and this process is repeated until the maximum number of iterations is reached.

[0074] To address the initial solution problem of ALNS mentioned above, a greedy strategy may not be able to generate a complete initial solution. This invention must find a method to guarantee the stable generation of the initial solution.

[0075] The pseudocode for the ALNS algorithm is shown in Table 2.

[0076] Table 2 shows the adaptive large neighborhood search algorithm.

[0077] The two-stage algorithm II-ALNS based on ALNS addresses the crucial role of initial solution generation in iterative algorithms. In this algorithm, initial solution generation is not only the starting point of the search process but also a key factor influencing convergence and the quality of the final solution. Especially in large-scale scheduling problems with multiple constraints, initial solution generation itself is a complex feasibility problem. Traditional ALNS algorithms often get stuck in local optima or even fail to start due to initial solution generation failures or low quality. To address this, this invention proposes a two-stage II-ALNS algorithm. The first stage ensures the generation of a feasible initial solution through a relaxation model, and the second stage performs objective optimization based on this. This method theoretically solves the feasibility dilemma in initial solution generation, providing a stable and efficient solution framework for cooperative scheduling under complex constraints.

[0078] In the first stage, this invention first changes the optimization objective to minimizing the number of vehicles used, and sets the number of vehicles in each fleet to be large enough to ensure an initial solution. There is no need to worry about the number of vehicles used not reflecting reality, because as the algorithm progresses, the number of vehicles used will be optimized to a sufficiently low number. The algorithm terminates when the number of vehicles used in each fleet is less than or equal to the actual number of vehicles available, and returns the solution at this point.

[0079] In the second stage, the present invention uses the solution output from the first stage as the initial solution of the ALNS algorithm with the goal of minimizing the travel distance, thereby solving the problem that the greedy strategy may not be able to generate an initial solution.

[0080] The II-ALNS algorithm framework diagram is as follows: Figure 3 As shown in the diagram, the framework clearly illustrates the overall structure of the algorithm and the flow of the two stages. The mathematical models of the first and second stages maintain a consistent basic framework, differing only in the initial available vehicle quantity constraint, in addition to the optimization objective. Specifically, the first stage focuses on optimizing the number of vehicles used, while the second stage introduces a more complex optimization objective. Importantly, as long as the first stage can optimize the number of vehicles used to reach the vehicle quantity constraint required by the second stage, the mathematical models of the two stages will have the same solution. This demonstrates that the algorithm can achieve seamless transitions between stages and consistency in solutions while satisfying resource constraints.

[0081] The II-ALNS-MO multi-objective optimization algorithm, based on II-ALNS, addresses the dual-objective optimization problem in airport ground support scheduling, where traditional single-objective II-ALNS algorithms are not directly applicable. Therefore, this invention proposes a Pareto multi-objective optimization algorithm (II-ALNS-MO) based on II-ALNS, whose core design concept is as follows: The original algorithm retains its two-stage structure, especially the robustness of generating feasible initial solutions in the first stage. In the second stage, a multi-objective optimization framework is introduced to approximate the true Pareto front by maintaining an elite solution set. The Chebyshev decomposition method is used to transform the bi-objective problem into multiple single-objective sub-problems, and global exploration is achieved through population co-evolution.

[0082] Table 3 shows the pseudocode and algorithm of the II-ALNS-MO algorithm.

[0083] The core idea of ​​destruction operators is to create a new search space by removing some nodes from the current solution. By disrupting the structure of the current solution, the algorithm can explore different neighborhoods, thereby escaping local optima. This invention uses seven destruction operators.

[0084] Random Removal, let the current solution be... The service node set is The damage ratio is :

[0085] in and These are the preset minimum and maximum damage ratios, respectively.

[0086] Number of nodes removed:

[0087] Probability distribution of randomly selected nodes:

[0088] in The set of nodes that were removed. .

[0089] Shaw removal (ShawRemoval) serves as the set of service nodes. Shaw removes nodes based on their similarity, and the similarity function is defined as:

[0090] in: The Euclidean distance between nodes i and j , Let i be the start time of the service for nodes i and j. , Let i be the demand for nodes i and j; For the weight parameters, satisfying ; Normalized similarity:

[0091] Normalization scales the three different metrics—distance, time difference, and demand difference—to the [0,1] interval, making the weight parameters more uniform. It can effectively control the relative importance of each indicator and prevent a single indicator with a large dimension from dominating the entire similarity calculation. Thus, the normalized similarity values The smaller the value, the more similar nodes i and j are.

[0092] Given a seed node s, select the k-1 nodes with the highest similarity:

[0093] The set of nodes that were removed.

[0094] Worst Removal: For node i in path r, define the distance saving value for removing node i.

[0095] in and These are the predecessor and successor nodes of node i, respectively.

[0096] Remove collection:

[0097] RouteRemoval, assuming the selected path is... The starting position is The length of the continuous segment is :

[0098]

[0099] Remove the set of nodes:

[0100] Cluster Removal The nodes are divided using the K-means algorithm. There are clusters, among which .

[0101] It is a set of nodes, containing a group of service nodes that are geographically close together, for clustering. Define cluster centers:

[0102] Cluster Center Clustering The geometric centroid (center of mass) of all nodes in the plane is the average position of these points on the plane.

[0103] node Belongs to clustering Conditions:

[0104] Representing vectors The Euclidean distance, with the subscript 2 indicating that it is the L2 norm, i.e., the Euclidean distance.

[0105] Select the largest cluster:

[0106] Remove m nodes randomly selected from the largest cluster:

[0107] TimeWindowRemoval, defining a node Time window width:

[0108] in and They are nodes The earliest and latest service times.

[0109] Remove the set (select the k nodes with the tightest time window):

[0110] DemandRemoval Remove nodes with high demand to free up capacity. Adjust the load to meet vehicle capacity constraints.

[0111] Calculate the demand for each node and remove the node with the highest demand. k Each node.

[0112] Removed in descending order of demand:

[0113] Repair operators and dynamically adjust weights, then select (adaptive). Greedy Repair Algorithm For the node to be inserted and location (Position p on path r), the insertion cost increment is:

[0114] in, Let p and p be the nodes at positions p-1 and p respectively in path r. For nodes Service hours This represents the penalty coefficient for violating the time window.

[0115] Optimal insertion position:

[0116] in, Let i be the set of all feasible insertion positions for node i.

[0117] Regret Repair Algorithm For the node to be inserted ,definition: Optimal insertion cost:

[0118] Suboptimal insertion cost:

[0119] Regret value is defined as:

[0120] Insert the node with the highest regret value first:

[0121] Where U is the set of currently unassigned nodes.

[0122] The operator weights are updated according to the following formula:

[0123] In the formula, For operator weights, For operator fractions, The number of times the operator is used. These are the weight update coefficients (controlling the rate of weight change). The operator weights are linked to past performance.

[0124] Table 4 shows the algorithm for updating the weights of the destruction operator.

[0125] The weights of operators are dynamically adjusted based on their performance during the search process. This mechanism ensures that the algorithm can maintain a balance between exploration (trying different operators) and utilization (preferring operators that perform well), thereby improving search efficiency and solution quality.

[0126] An airport ground support vehicle collaborative scheduling optimization system is provided to implement the method of the above embodiments. The system includes: The human-computer interaction module allows users to input basic data, select a model, and display optimization results. The model database stores the scheduling model based on the shortest total vehicle travel distance, the scheduling model based on the minimum total delay time, and the Pareto multi-objective optimization model that integrates the total vehicle travel distance and the total delay time. The scheduling optimization module imports basic data into the selected model for model solving, including: Generate feasible initial solution: Relax vehicle constraints and set a large virtual vehicle limit for each fleet k to ensure the solution space is feasible; change the optimization objective to temporarily minimize the total number of vehicles in the fleet; run ALNS and use the destruction operator and repair operator to iteratively optimize and reduce the number of vehicles; terminate when the number of vehicles in each fleet does not exceed the actual number, and obtain a feasible solution; Goal-driven optimization: Using the feasible solution as the current solution, after restoring the original goal, the enhanced ALNS is run for optimization, adaptively selecting destruction and repair operators, iteratively executing destruction and repair operations, accepting new solutions based on the simulated annealing criterion, and dynamically updating operator weights; finally, the optimal scheduling scheme is output.

[0127] The specific usage process of the above system includes: Step 1: System initialization and data input.

[0128] Receive input data from the airport dispatch system, including: Flight schedule collection: Includes each flight number, parking position, estimated arrival time, estimated wheel chock removal time, and required completion time for critical services (catering, closing passenger cabin doors, closing cargo cabin doors).

[0129] Support fleet: Define M different types of support fleets (e.g., unloading, refueling, cleaning, freight, catering, water supply), and define the number of vehicles, vehicle capacity, and service time of each fleet.

[0130] Spatiotemporal constraint parameters: including travel time from node i to node j, service time window of flight i to fleet k, and service priority order between fleets. 4. Optimization objective configuration: select a single optimization objective (minimize total travel distance or minimize total delay time) or a dual-objective Pareto optimization mode based on the actual operation stage (such as peak flight period or off-peak period).

[0131] Step 2: Construct a scheduling optimization model.

[0132] Based on the configuration in step 1, construct the corresponding mathematical model in the computing system.

[0133] If the goal is to minimize the travel distance, then the model is constructed as shown in formulas (1)-(12) in the background art.

[0134] If the goal is to minimize the delay time, then a delay optimization model is constructed that incorporates formulas (13)-(21) from the background technique.

[0135] If bi-objective optimization is chosen, a Pareto optimization model is constructed as described in the background section.

[0136] Step 3: Solve the problem by executing the two-stage adaptive large neighborhood search (Ⅱ-ALNS) algorithm.

[0137] This step is the core of the implementation of this invention, and its first and second stage processes are as follows: Figure 2 As shown.

[0138] Phase 1: Generation of feasible initial solutions.

[0139] Relax vehicle constraints: Set a virtual vehicle limit for each fleet k that is much larger than the actual number of vehicles |V^k| (e.g., 50 vehicles) to ensure that the solution space contains feasible solutions.

[0140] Change the optimization objective: Temporarily replace the original model's optimization objective (such as driving distance) with minimizing the total number of vehicles actually used by all fleets.

[0141] The ALNS algorithm performs a feasibility search: it employs an adaptive large neighborhood search framework, using a set of disruptive operators (such as random removal) and repair operators (such as greedy insertion), and iteratively optimizes under this relaxed model. During optimization, the algorithm gradually reduces the number of vehicles used by each fleet.

[0142] Termination Criterion: The first phase terminates when, in the scheduling scheme obtained after iterative optimization, the number of vehicles used by each fleet k is less than or equal to its actual number of vehicles. At this point, the obtained solution S_feasible is a feasible scheduling scheme that satisfies all hard constraints.

[0143] Phase Two: Goal-Driven Optimization.

[0144] Initialization: Use the feasible solution S_feasible output from the first stage as the current solution for this stage.

[0145] Restore the original objective: Restore the optimized objective to the true objective configured in step S2 (such as minimizing the total delay time).

[0146] Run enhanced ALNS for deep optimization: Adaptive operator selection: Dynamically select a destruction operator from a set of 7 destruction operators, including random removal, worst-case removal, time window removal, Shaw removal, path removal, clustering removal, and demand removal, based on historical performance weights; dynamically select a repair operator from greedy repair and regret value repair operators.

[0147] Iterative search: In each iteration, a "break-and-repair" operation is performed on the current solution to generate a new solution. High-quality solutions are accepted, or occasionally low-quality solutions are accepted, based on the simulated annealing criterion to escape local optima.

[0148] Weight update: The weights of each operator are dynamically updated based on the quality of the new solution generated by the operator (such as finding a new global optimum, improving the current solution, or accepting a non-improved solution), so as to achieve the adaptive search strategy.

[0149] Output: After reaching the preset maximum number of iterations, output the best-performing scheduling scheme S_best found during the entire search process.

[0150] Step S4: Output and execute the scheduling scheme.

[0151] The optimized scheduling scheme S_best obtained in step S3 is converted into executable instructions, including: the detailed operation path of each support vehicle (service flight sequence); the planned start time of each support service for each flight; the estimated total flight delay time and the total vehicle travel distance.

[0152] This plan can be distributed to each support vehicle fleet through the airport ground operations management system.

[0153] For the implementation of the dual-objective optimization mode (Ⅱ-ALNS-MO): In the second stage of step S3 above, the II-ALNS-MO algorithm is replaced (its process can be found in the pseudocode of Algorithm 2 in the background section): The algorithm maintains both a population (corresponding to multiple weight vectors) and an elite archive (used to store Pareto non-dominated solutions). Each search thread (corresponding to a weight vector) uses a Chebyshev scalarization function to transform its bi-objective problem into a single-objective problem and independently performs the II-ALNS search. During the search process, newly discovered non-dominated solutions are continuously added to the elite archive, and dominated solutions are removed. Finally, all solutions in the elite archive are output, forming an approximate Pareto front, which the decision-maker uses to select the final solution based on the real-time operational strategy.

[0154] To verify the effectiveness of the present invention, data from a hub airport was used as an example for implementation.

[0155] Example scenario: Data scale: Select 60 inbound and outbound flights during the time period from 08:00 to 12:00 on a certain day.

[0156] Support services: Assume that each flight requires 6 types of ground services: disembarkation, refueling, cleaning, cargo, catering, and water supply.

[0157] Resource constraints: Each type of fleet has 5 vehicles of the same type.

[0158] Optimization objectives: Test minimizing the total travel distance (M1) and minimizing the total delay time (M2).

[0159] Figure 4 This is a basic layout diagram of the airport, where dots represent aircraft stands, numbers represent stand numbers, and lines represent ground support roads.

[0160] Phase 1 Validation: Table 5 shows the convergence process of the number of vehicles used by each fleet in the first stage of the algorithm under a scale of 60 flights.

[0161] Table 5 shows the experimental results, verifying the effectiveness of the first stage of the II-ALNS algorithm. To ensure the generation of the initial solution, each fleet was initialized with 50 vehicles. Then, iterative optimization was performed with the number of vehicles used as the optimization objective until the number of vehicles used by each fleet met the actual requirements. For the generation of the ALNS initial solution, if only a greedy strategy was used, the generated initial solution would require each fleet to use 6 vehicles, which does not satisfy the constraint that each fleet should only have 5 vehicles. However, the II-ALNS algorithm successfully solved this problem. As can be seen from Table 5, after 2447 iterations, the algorithm successfully controlled the number of vehicles used by all fleets within the constraints, and ultimately optimized them to a relatively optimal configuration of 4 vehicles each.

[0162] Second-stage optimization results: Driving distance optimization: For target M1, the algorithm convergence curve is shown in the attached figure. Figure 5 As shown. For a scale of 60 sorties, the algorithm can quickly decrease and stabilize near the optimal value.

[0163] Delay time optimization: For target M2, the algorithm convergence curve is shown in the attached figure. Figure 6 As shown, it also exhibits fast convergence characteristics.

[0164] Dispatch scheme visualization: The optimized Gantt chart for the support tasks of 60 flights is attached. Figure 5 As shown, the timing and coordination of different support services for each flight are clearly displayed, and it can be seen that high-priority services (such as the following flight) are always executed first.

[0165] Bi-objective optimization results: Running the II-ALNS-MO algorithm, a Pareto front consisting of 12 non-dominated solutions was successfully obtained, as shown in the attached figure. Figure 7 As shown, this frontier clearly demonstrates the trade-off between travel distance and delay time, providing decision-makers with a variety of optimal scheduling strategies ranging from "minimum delay" to "shortest distance".

[0166] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned airport ground support vehicle collaborative scheduling optimization method.

[0167] A computer program product includes a computer program that, when executed by a processor, implements the aforementioned airport ground support vehicle collaborative scheduling optimization method.

[0168] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented, in whole or in part, as a computer program product, the computer program product includes one or more computer instructions. When the computer program instructions are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line, or wireless (e.g., infrared, wireless, microwave, etc.) means). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0169] The above description is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for optimizing the collaborative scheduling of airport ground support vehicles, characterized in that, include: S1. Select the model based on the optimization objective: When the optimization objective is to minimize the total distance traveled by vehicles, the distance-optimized scheduling model is selected. When the optimization objective is to minimize the total delay time, the delay time optimization scheduling model is selected; When the optimization objective combines the total distance traveled by the vehicle and the total delay time, the Pareto multi-objective optimization model is selected. S2. For the selected model, solve the model, including: S201. Generation of feasible initial solution: Relax vehicle constraints and set a large virtual vehicle limit for each fleet k to ensure the solution space is feasible; change the optimization objective to temporarily minimize the total number of vehicles in the fleet; run ALNS and use destruction and repair operators to iteratively optimize and reduce the number of vehicles; terminate when the number of vehicles in each fleet does not exceed the actual number, and obtain a feasible solution. S202, Goal-Driven Optimization: Taking the feasible solution as the current solution, after restoring the original goal, the enhanced ALNS is run for optimization, adaptively selecting destruction and repair operators, iteratively executing destruction and repair operations, accepting new solutions based on the simulated annealing criterion, and dynamically updating operator weights; finally, the optimal scheduling scheme is output. S3. Output and execution of the optimal scheduling scheme.

2. The airport ground support vehicle collaborative scheduling optimization method according to claim 1, characterized in that, The mathematical expression for the distance-optimized scheduling model is: （1） （2） （3） （4） （5） （6） （7） in: Let be the cost function, representing the travel cost of fleet k on edge (i, j), where i and j are the gate nodes where the flight is located; For decision variables, the value is 1 if vehicle v in convoy k travels from gate node i to gate node j, and 0 otherwise; For the set of car teams k; , represents a specific vehicle in vehicle k of the vehicle fleet; This is a set of garage and machine station nodes; For the set of machine station nodes; As a decision variable, if vehicle v in convoy k travels from gate node i to gate node... The value is 1 if it is 1, otherwise it is 0. As a decision variable, if vehicle v in fleet k is from the machine station node The value is 1 if the vehicle reaches node j, otherwise it is 0. As a decision variable, if vehicle v in fleet k originates from garage node The value is 1 if the vehicle reaches node j, otherwise it is 0. Let vehicle v in fleet k be the decision variable, and if vehicle v travels from gate node i to garage node i... The value is 1 if it is 1, otherwise it is 0. Let be a real number decision variable, representing the start time of service for vehicle v in fleet k to flight at gate node j; Let be a real number decision variable, representing the start time of service for vehicle v in fleet k on the flight at gate node i; The duration of service for fleet k to flights at gate node i; Let be the time it takes for vehicle k to travel from node i to node j; The maximum time a vehicle v in fleet k is allowed to remain at its stand after completing service; The resource requirements of the flight at gate node i; The maximum amount of resources carried by vehicles in fleet k; The duration of service for fleet k to flights at gate node i; Let be the time it takes for vehicle k to travel from node i to node j; This is the lower bound of the arrival time of vehicles in convoy k at gate node i; Let be a real number decision variable, representing the fleet. The start time of service for vehicle v on the flight at gate node i; For the team The duration of service for the flight at gate node i; Let be a real number decision variable, representing the fleet. The start time of service for vehicle v on the flight at gate node i; This is the upper bound of the time when vehicles in fleet k arrive at station node i.

3. The airport ground support vehicle collaborative scheduling optimization method according to claim 2, characterized in that, The mathematical expression of the delay time optimization scheduling model includes: Calculate the completion time for each service: （13） （14） Calculate the delay time: Introducing auxiliary variables Indicates the latest completion time of flight i: Introducing binary variables When flight i is delayed ,otherwise ; The objective function is: , in: for ; for ; for ; Let v be a real number decision variable, representing the start time of service for vehicle v in the catering fleet to the flight at gate node i; The duration of catering service for flights at gate node i; Let v be a real number decision variable, representing the start time of service for vehicle v in the freight truck fleet to the flight at gate node i; The duration of the cargo service for flights at gate node i; Let v be a real number decision variable, representing the start time of service for vehicle v in the shuttle bus fleet to the flight at gate node i; The duration of the shuttle service for flights at gate node i; For camera position ; M is ; for ; Objective function: , Retain constraints (2)-(12); Add new constraints (13)-(21); Ensure by constraints (16)-(18) The delay time is equal to the maximum of the three service completion times; constraints (19)-(21) ensure the delay time. ;when This indicates no delay; when hour, This indicates a delay.

4. The airport ground support vehicle collaborative scheduling optimization method according to claim 1, characterized in that, The destruction operators include: random removal, Shaw removal, worst-case removal, path removal, clustering removal, time window removal, and demand removal.

5. The airport ground support vehicle collaborative scheduling optimization method according to claim 1, characterized in that, The repair operators include a greedy repair algorithm and a regret value repair algorithm.

6. An airport ground support vehicle collaborative scheduling optimization system, characterized in that, include: The human-computer interaction module allows users to input basic data, select a model, and display optimization results. The model database stores the scheduling model based on the shortest total vehicle travel distance, the scheduling model based on the minimum total delay time, and the Pareto multi-objective optimization model that integrates the total vehicle travel distance and the total delay time. The scheduling optimization module imports basic data into the selected model for model solving, including: Generate feasible initial solution: Relax vehicle constraints and set a large virtual vehicle limit for each fleet k to ensure the solution space is feasible; change the optimization objective to temporarily minimize the total number of vehicles in the fleet; run ALNS and use the destruction operator and repair operator to iteratively optimize and reduce the number of vehicles; terminate when the number of vehicles in each fleet does not exceed the actual number, and obtain a feasible solution; Goal-driven optimization: Using the feasible solution as the current solution, after restoring the original goal, the enhanced ALNS is run for optimization, adaptively selecting destruction and repair operators, iteratively executing destruction and repair operations, accepting new solutions based on the simulated annealing criterion, and dynamically updating operator weights; finally, the optimal scheduling scheme is output.

7. The airport ground support vehicle collaborative scheduling optimization system according to claim 6, characterized in that, The mathematical expression for the distance-optimized scheduling model is: （1） （2） （3） （4） （5） （6） （7） in: Let be the cost function, representing the travel cost of fleet k on edge (i, j), where i and j are the gate nodes where the flight is located; For decision variables, the value is 1 if vehicle v in convoy k travels from gate node i to gate node j, and 0 otherwise; For the set of car teams k; , represents a specific vehicle in vehicle k of the vehicle fleet; This is a set of garage and machine station nodes; For the set of machine station nodes; As a decision variable, if vehicle v in convoy k travels from gate node i to gate node... The value is 1 if it is 1, otherwise it is 0. As a decision variable, if vehicle v in fleet k is from the machine station node The value is 1 if the vehicle reaches node j, otherwise it is 0. As a decision variable, if vehicle v in fleet k originates from garage node The value is 1 if the vehicle reaches node j, otherwise it is 0. Let vehicle v in fleet k be the decision variable, and if vehicle v travels from gate node i to garage node i... The value is 1 if it is 1, otherwise it is 0. Let be a real number decision variable, representing the start time of service for vehicle v in fleet k to flight at gate node j; Let be a real number decision variable, representing the start time of service for vehicle v in fleet k on the flight at gate node i; The duration of service for fleet k to flights at gate node i; Let be the time it takes for vehicle k to travel from node i to node j; The maximum time a vehicle v in fleet k is allowed to remain at its stand after completing service; The resource requirements of the flight at gate node i; The maximum amount of resources carried by vehicles in fleet k; The duration of service for fleet k to flights at gate node i; Let be the time it takes for vehicle k to travel from node i to node j; This is the lower bound of the arrival time of vehicles in convoy k at gate node i; Let be a real number decision variable, representing the fleet. The start time of service for vehicle v on the flight at gate node i; For the team The duration of service for the flight at gate node i; Let be a real number decision variable, representing the fleet. The start time of service for vehicle v on the flight at gate node i; This is the upper bound of the time when vehicles in fleet k arrive at station node i.

8. The airport ground support vehicle collaborative scheduling optimization system according to claim 7, characterized in that, The mathematical expression of the delay time optimization scheduling model includes: Calculate the completion time for each service: （13） （14） Calculate the delay time: Introducing auxiliary variables Indicates the latest completion time of flight i: Introducing binary variables When flight i is delayed ,otherwise ; The objective function is: , in: for ; for ; for ; Let v be a real number decision variable, representing the start time of service for vehicle v in the catering fleet to the flight at gate node i; The duration of catering service for flights at gate node i; Let v be a real number decision variable, representing the start time of service for vehicle v in the freight truck fleet to the flight at gate node i; The duration of the cargo service for flights at gate node i; Let v be a real number decision variable, representing the start time of service for vehicle v in the shuttle bus fleet to the flight at gate node i; The duration of the shuttle service for flights at gate node i; For camera position ; M is ; for ; Objective function: , Retain constraints (2)-(12); Add new constraints (13)-(21); Ensure by constraints (16)-(18) The delay time is equal to the maximum of the three service completion times; constraints (19)-(21) ensure the delay time. ;when This indicates no delay; when hour, This indicates a delay.

9. A computer program product, characterized in that, The invention includes a computer program that, when executed by a processor, implements the airport ground support vehicle collaborative scheduling optimization method as described in any one of claims 1-5.

10. A computer-readable storage medium storing a computer program, characterized in that, When the program is executed by the processor, it implements the airport ground support vehicle collaborative scheduling optimization method as described in any one of claims 1-5.

Images