A multi-depot personnel and vehicle cooperative scheduling method and system

Abstract

The application discloses a multi-vehicle yard personnel and vehicle cooperative scheduling method and system, relates to the technical field of airport scheduling, and comprises the following steps: determining a scheduling scale according to a comparison result of a task quantity of a scheduling task associated with a scheduling request and a scale quantity threshold value; when the scheduling scale is a small or medium scale, taking minimum staff transfer time as a target, adopting personnel, a vehicle yard and the scheduling task as nodes, and constructing a scheduling network flow model based on a staff transfer time cost function; a commercial solver solves the scheduling network flow model based on the scheduling request under scheduling constraints and outputs a scheduling plan; when the scheduling scale is a large scale, a greedy algorithm is adopted to construct an initial solution based on the scheduling request and generate an initial vehicle yard record table according to the initial solution; and the initial solution is optimized based on the initial vehicle yard record table and the staff transfer time cost function by using a self-adaptive large neighborhood search algorithm to determine the scheduling plan. The scheduling strategy is flexible and applicable to scheduling services of different scales, and the reliability of personnel and vehicle separation service scheduling is improved.

Description

Technical Field

[0001] This invention relates to the field of airport dispatching technology, and in particular to a method and system for coordinated dispatching of personnel and vehicles across multiple parking lots. Background Technology

[0002] The rapid growth of global air traffic has brought greater challenges to airport ground services. Under the condition of limited airport ground service resources, efficient ground service scheduling is an important way to reduce flight delays, improve airport turnaround efficiency, and reduce corporate operating costs.

[0003] Based on flight service procedures and standards, services can be divided into arrival / departure services and transfer services. Arrival / departure services have strict arrival time requirements, such as shuttle buses, towing, passenger elevators, and baggage handling. Transfer services have latest completion time requirements, such as cabin cleaning, refueling, and catering. Arrival / departure services can be further categorized into driver-vehicle separation and driver-vehicle binding services based on the different occupancy times for the driver and vehicle. Driver-vehicle binding is mainly used for services such as shuttle buses, where the driver and vehicle occupy the same amount of time. In contrast, for services like passenger elevators, where driver and vehicle occupancy differs.

[0004] In the service of separating people and vehicles, such as Figure 1 As shown, taking remote gate passenger boarding and alighting as an example: For inbound flights, air traffic control will announce the aircraft's gate before landing. During the aircraft's taxiing, the task of docking the passenger stairs at the remote gate will be assigned to the corresponding special vehicle driver. The special vehicle driver will drive the special vehicle from the parking lot to the corresponding gate to wait for the aircraft to arrive. After the aircraft arrives at the gate and lowers the wheel chocks, the special vehicle driver will drive the special vehicle to dock. After docking and verification, the special vehicle will provide service, and the special vehicle driver will participate in other services in a work vehicle. The passenger stairs serve as the passage for passengers, crew members, cabin cleaning staff, etc. to enter and exit the aircraft. When the aircraft is about to depart, a special vehicle driver will be assigned to the gate to wait and complete the evacuation of the special vehicle before the cabin door closes. After the service is completed, the special vehicle driver will park the special vehicle in the parking lot or use it for docking in other services. In this business scenario, the resources needed to meet service requirements involve three types: vehicles, drivers, and parking lots. However, the current strategy mainly considers driver scheduling, assigning arrival and departure tasks to drivers and letting them decide the origin or destination of the vehicle. This ignores parking lot capacity limitations and the rational use of vehicle resources, which can easily lead to problems such as excessively long waiting times for drivers to coordinate between vehicle tasks. It also fails to effectively improve the utilization efficiency of vehicles and parking lots, resulting in low reliability of the human-vehicle separation service scheduling. Summary of the Invention

[0005] This invention provides a method and system for collaborative scheduling of personnel and vehicles in multiple parking lots, which solves the technical problem that the current personnel and vehicle separation service scheduling strategy mainly considers driver scheduling, resulting in low reliability of personnel and vehicle separation service scheduling.

[0006] The first aspect of the present application provides a multi-yard personnel vehicle cooperative scheduling method, comprising:

[0007] determining the scheduling scale according to a comparison result of a task quantity of a scheduling task associated with the scheduling request and a scale quantity threshold value;

[0008] when the scheduling scale is a small or medium scale, taking the minimum employee transfer time as the target, using personnel, yards and scheduling tasks as nodes and constructing a scheduling network flow model based on an employee transfer time cost function;

[0009] calling a commercial solver to solve the scheduling network flow model based on the scheduling request under scheduling constraints, and outputting a scheduling plan;

[0010] when the scheduling scale is a large scale, using a greedy algorithm to construct an initial solution based on the scheduling request, and generating an initial yard record table according to the initial solution;

[0011] optimizing the initial solution based on the initial yard record table and the employee transfer time cost function through an adaptive large neighborhood search algorithm to determine a scheduling plan.

[0012] Optionally, the step of optimizing the initial solution based on the initial yard record table and the employee transfer time cost function through the adaptive large neighborhood search algorithm to determine a scheduling plan comprises:

[0013] letting the initial solution be a current solution, taking the initial solution as an optimal solution, and initializing a destruction operator, a repair operator, an operator weight, an operator selection frequency and an operator selection score;

[0014] selecting a target destruction operator and a target repair operator through roulette according to the operator weights;

[0015] destroying the current solution based on the initial yard record table and the employee transfer time cost function using the target destruction operator to determine a removed task pair, a destroyed solution, a destroyed yard record table and a destroyed employee transfer time cost;

[0016] constructing a planned task pair of each removed node in the removed task pair, and judging the feasibility of the planned task pair based on the destroyed yard record table to determine a feasible task pair and a corresponding feasible insertion position;

[0017] repairing the destroyed solution by combining the destroyed yard record table, the employee transfer time cost function, the feasible task, the feasible insertion position and the destroyed employee transfer time cost through the target repair operator to output a new solution;

[0018] determine a new current solution based on a comparison result between the new solution and the optimal solution and the current solution, update the operator selection score and the operator selection times, and count the iteration times;

[0019] If the iteration times are less than a termination times threshold, update the operator weights according to the operator selection score and the operator selection times when the iteration times meet an operator update condition, and select a new target destruction operator and a new target repair operator through roulette according to the operator weights, until the iteration times equal the termination times threshold, and output the optimal solution as the scheduling plan.

[0020] Optionally, the destroying the current solution based on the initial yard record table and the employee transfer time cost function using the target destruction operator includes:

[0021] randomly selecting at least one task pair in the current solution as a removal task pair for removal, and determining a destruction solution;

[0022] Alternatively, removing a task pair associated with an employee transfer path with the largest employee transfer time cost in the current solution based on the employee transfer time cost function, and constructing a destruction solution;

[0023] Alternatively, calculating a similarity between task pairs based on a relative distance between nodes associated with the task pairs, a node start time difference, a node end time difference, and a node type difference in the current solution, removing a removal task pair based on the highest similarity between task pairs, and generating a destruction solution;

[0024] Alternatively, determining a removal cost reduction amount of each task pair in the current solution using the employee transfer time cost function, selecting a task pair with the largest removal cost reduction amount as a removal task pair for removal, and determining a destruction solution;

[0025] updating the initial yard record table based on the destruction solution to determine a destruction yard record table of the destruction solution, and calculating a destruction employee transfer time cost associated with the destruction solution according to the employee transfer time cost function.

[0026] Optionally, the constructing a plan task pair of each removal node in the removal task pair, and judging the feasibility of the plan task pair based on the destruction yard record table to determine a feasible task pair and a corresponding feasible insertion position includes:

[0027] constructing a plan task pair of each removal node in the removal task pair;

[0028] inserting any of the plan task pairs into the destruction solution, and updating the destruction yard record table to obtain a corresponding repair yard record table;

[0029] The feasible task pairs are obtained by using the repair yard record table association plan task pairs satisfying the non-negativity constraint, and the corresponding feasible insertion positions are determined.

[0030] Optionally, the damaged solution is repaired by combining the target repair operator with the damaged yard record table, the employee transfer time cost function, the feasible task, the feasible insertion position, and the damaged employee transfer time cost, and a new solution is output, including:

[0031] Each feasible task pair is inserted into any corresponding feasible insertion position in the damaged solution by using a removal set composed of each feasible task pair, and a corresponding cost increment is determined based on the employee transfer time cost function and the damaged employee transfer time cost.

[0032] The feasible task pair with the minimum cost increment is selected as an insertable task pair, and the insertable task pair is inserted into the damaged solution at the feasible insertion position associated with the minimum cost increment to determine a new damaged solution.

[0033] Alternatively, the difference between the minimum cost increment of each feasible task pair and the corresponding second minimum cost increment is taken as the regret value of each feasible task pair, the feasible task pair with the maximum regret value is taken as the insertable task pair, and the insertable task pair is inserted into the damaged solution at the feasible insertion position associated with the minimum cost increment of the insertable task pair to determine a new damaged solution.

[0034] The feasible task pair to which the removal node associated with the insertable task pair belongs is deleted from the removal set to generate a new removal set, and each feasible task pair in the new removal set is inserted into any corresponding feasible insertion position in the new damaged solution until the removal set is empty, and a new solution is output.

[0035] Optionally, the employee transfer time cost function includes:

[0036] ;

[0037] The scheduling network flow model includes:

[0038] ;

[0039] The scheduling constraint conditions include:

[0040]

[0041] wherein, ;

[0042] wherein, is the node , is the node , start time of node , end time of node , initial position of employee , transition value of employee between node and node , necessary transition time between node and node , initial position of employee , transition value of employee between initial position of employee and node , set of employees , set of location points of departure task, set of location points of arrival task, set of yards, set of nodes, set of location points of arrival task and departure task, transition value of employee to node , set of initial location points of employee , transition value of employee to node , node , transition value of employee between node and node , transition value of employee between node and node , yard, available vehicle number of yard before start time of node , initial available vehicle number of yard , set of location points of arrival task before start time of node , transition value of employee between yard and node , a set of evacuation task location points before the start time of the node , a value of the employee transfer between the node and the parking lot , a number of available parking spaces of the parking lot before the start time of the node , an initial number of available parking spaces of the parking lot , a value of the employee transfer between the node and the parking lot , a value of the employee transfer between the parking lot and the node , a total capacity , a time at which the employee leaves the node , a maximum value , a time at which the employee leaves the parking lot for the node , a necessary transfer time between the parking lot and the node , an end location of the employee , a set of location points of the docking task and the evacuation task before the start time of the node , a start time of the node , , a value of the employee transfer between the node and the parking lot .

[0043] The second aspect of the present application provides a multi-parking lot personnel and vehicle cooperative scheduling system, comprising:

[0044] a scale determination module configured to determine a scheduling scale according to a comparison result of a task quantity of a scheduling task associated with a scheduling request and a scale quantity threshold value;

[0045] a model construction module configured to, when the scheduling scale is a small or medium scale, construct a scheduling network flow model by taking a minimum employee transfer time as a target, taking personnel, a parking lot, and a scheduling task as nodes, and based on an employee transfer time cost function;

[0046] ​a first strategy determination module configured to call a commercial solver to solve the scheduling network flow model based on the scheduling request under scheduling constraints, and output a scheduling plan;

[0047] an initial data construction module configured to, when the scheduling scale is large-scale, construct an initial solution and an initial yard record table based on the scheduling request by using a greedy algorithm;

[0048] a second strategy determination module configured to determine a scheduling plan by optimizing the initial solution based on the initial yard record table and a staff transfer time cost function by using an adaptive large neighborhood search algorithm.

[0049] A computer device provided in a third aspect of the present application comprises a memory and a processor, the memory stores a computer program, and the computer program is executed by the processor to make the processor execute the steps of the multi-yard personnel and vehicle collaborative scheduling method according to any one of the above aspects.

[0050] A computer readable storage medium provided in a fourth aspect of the present application stores a computer program, and the computer program is executed to implement the multi-yard personnel and vehicle collaborative scheduling method according to any one of the above aspects.

[0051] A computer program product provided in a fifth aspect of the present application comprises computer programs / instructions, and the computer programs / instructions are executed by a processor to implement the multi-yard personnel and vehicle collaborative scheduling method according to any one of the above aspects.

[0052] It can be seen from the above technical solutions that the present application has the following advantages:

[0053] The above scheme of the present application provides a multi-vehicle yard personnel and vehicle collaborative scheduling method, which comprises: determining a scheduling scale according to a comparison result of a task quantity of a scheduling task associated with a scheduling request and a scale quantity threshold; when the scheduling scale is a small or medium scale, taking minimum staff transfer time as an objective, using personnel, yards and scheduling tasks as nodes and constructing a scheduling network flow model based on a staff transfer time cost function; calling a commercial solver to solve the scheduling network flow model based on the scheduling request under scheduling constraints, and outputting a scheduling plan; when the scheduling scale is a large scale, using a greedy algorithm to construct an initial solution based on the scheduling request, and generating an initial yard record table according to the initial solution; and optimizing the initial solution based on the initial yard record table and the staff transfer time cost function through an adaptive large neighborhood search algorithm to determine the scheduling plan. Based on the above scheme, the dynamic change of the number of vehicle parking spaces in the yard is tracked through the scheduling network flow model and the yard record table in the scheduling strategy, and the scheduling of staff and vehicles is considered at the same time based on the scheduling request from the scheduling task, while the utilization of the three types of resources of vehicles, yards and staff is optimized, the work coordination effect under the human-vehicle separation mode is improved, the corresponding scheduling strategy is applied to small or medium scale and large scale scheduling services respectively, the overall scheduling efficiency is improved on the basis of ensuring the scheduling plan accuracy, and the human-vehicle separation service scheduling reliability is improved. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0055] Figure 1 Schematic diagram of the scheduling framework for airport human-vehicle separation service;

[0056] Figure 2 Step flowchart of a multi-vehicle yard personnel and vehicle collaborative scheduling method provided by the embodiment of the present application;

[0057] Figure 3 Schematic diagram of a human-vehicle separation scheduling example of a scheduling plan provided by the embodiment of the present application;

[0058] Figure 4 Schematic diagram of personnel path pairing provided by the embodiment of the present application;

[0059] Figure 5 Schematic diagram of a yard record table updating mechanism provided by the embodiment of the present application;

[0060] Figure 6 Schematic diagram of a path damage updating mechanism provided by the embodiment of the present application;

[0061] Figure 7 The updating mechanism of the task insertion path provided by the embodiment of the present application is shown in the figure.

[0062] Figure 8 The convergence process diagram of the example S62-P15-V37 provided by the embodiment of the present application is shown in the figure.

[0063] Figure 9 The variance diagram of the ALNS example provided by the embodiment of the present application is shown in the figure.

[0064] Figure 10 The convergence process diagram of the ALNS example provided by the embodiment of the present application is shown in the figure.

[0065] Figure 11 The curve diagram of the constraint and the variable number of the single-stage model changing with the example size provided by the embodiment of the present application is shown in the figure.

[0066] Figure 12 The structural block diagram of a multi-depot personnel and vehicle collaborative scheduling system provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0067] The embodiment of the present application provides a multi-depot personnel and vehicle collaborative scheduling method and system, which is used for solving the technical problem that the current human-vehicle separation service strategy scheduling mainly considers the scheduling of drivers, resulting in low reliability of human-vehicle separation service scheduling.

[0068] In order to make the purpose, characteristics and advantages of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the following described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0069] Please refer to Figure 2 , Figure 2 The step flowchart of a multi-depot personnel and vehicle collaborative scheduling method provided by the embodiment of the present application is shown in the figure.

[0070] The multi-depot personnel and vehicle collaborative scheduling method provided by the present application comprises:

[0071] Step 101, determining the scheduling size according to the comparison result of the task quantity of the scheduling task associated with the scheduling request and the size quantity threshold.

[0072] It should be noted that when scheduling in response to a scheduling request, the following assumptions are met:

[0073] 1) Determined service information: all relevant service information is known within the decision period of the dispatch request, including dispatch service-related information such as yards, personnel, and vehicles;

[0074] 2) Available and homogeneous resources: all vehicles and yards included in the dispatch can complete subsequent tasks, and the resources are homogeneous;

[0075] 3) Ignoring the connection time: ignoring the connection time between employees during the process of taking and parking the special vehicle from the work trolley;

[0076] 4) Assumption of departure and docking tasks: assuming that employees will not first park the departure vehicle in a yard and then take a vehicle from another yard to perform the next docking task when performing the departure and docking tasks;

[0077] On the basis of the above assumption conditions, the corresponding dispatch scale is determined according to the dispatch request so as to select a suitable dispatch strategy to construct a dispatch plan; by analyzing the dispatch request, the relevant service information of the dispatch task, vehicle, yard, and employee associated with the human-vehicle separation problem can be determined, such as the number and description content, wherein the dispatch task includes a docking task and a departure task, the docking task refers to an employee driving a special vehicle to a corresponding parking space for docking, the departure task refers to an employee driving a special vehicle to a corresponding parking space for departure, and the task number of the dispatch task refers to the total number of docking tasks and departure tasks; in one implementation manner, the dispatch scale of the current dispatch request can be determined by comparing the task number of the dispatch task with a preset scale number threshold, and the scale number threshold can be determined by data experiments.

[0078] Step 102, when the dispatch scale is medium or small, taking the minimum employee transfer time as the target, using personnel, yards, and dispatch tasks as nodes and constructing a dispatch network flow model based on an employee transfer time cost function.

[0079] It should be noted that when the dispatch scale is medium or small, a single-stage mixed integer programming model is used to generate a dispatch plan as the first dispatch strategy: the human-vehicle separation dispatch problem is constructed as a network flow model , wherein the vertices in the graph are all nodes , in the graph; a series of employee transfer path arcs are included; there are dispatch tasks with resource dispatch requirements in the dispatch request , the dispatch tasks are divided into two task sets, docking and departure , that is , and each dispatch task has a corresponding start time , end time , resource scheduling request of each dispatch task needs to be completed, where the docking needs to be from the yard , call special vehicles or use special vehicles after evacuation, and evacuation tasks need to park vehicles at the evacuation point to a yard or for docking tasks, and the initial available number of vehicles in the yard , the initial available number of parking spaces is , the total capacity is , the above scheduling problem needs to allocate dispatch tasks to employees and plan the order of each employee to execute the task, the source or destination of the vehicle, comply with the capacity constraints of the yard, ensure the rationality of picking up and dropping off, and for multiple visits to the yard, use to record the time of the employee after visiting the yard; the related parameter symbols are shown in Table 1:

[0080] Table 1: Set and parameter symbol description

[0081]

[0082] It can be understood that the cost of airport ground service is divided into fixed cost and variable cost, the fixed cost mainly refers to the employment of drivers, the purchase and maintenance of vehicles, and other costs unrelated to the number of service customers and the operation route, and the variable cost mainly comes from fuel cost, travel time, etc.; under the framework of rolling time domain, ground staff scheduling belongs to short-period decision, and the configuration of vehicle number, employee number and yard facility is usually based on the load situation in a longer period, and in the short-period decision, how to arrange each employee to complete the task efficiently based on the existing resource configuration is mainly considered, and more consideration is given to variable cost;

[0083] To reduce variable cost, the transfer time of employees should be shortened as much as possible in the separate scheduling of employees and vehicles, but when the employee directly uses the vehicle at the evacuation point for the next docking point, it will cause continuous occupation of employee and vehicle resources, and increase the transfer time cost, so in actual operation, when the time interval between the evacuation task and the next docking task is long, the ground staff usually parks the vehicle safely in the yard and drives the special vehicle to the corresponding position before the next task starts; the related decision variables are shown in Table 2:

[0084] Table 2: Decision variable information table

[0085]

[0086] ​In summary, considering the minimization of employee transfer time, and taking personnel, parking lots, and scheduling tasks as nodes, a scheduling network flow model is constructed based on the employee transfer time cost function, as described below:

[0087] The employee transfer time cost function includes:

[0088] ;

[0089] The scheduling network flow model includes:

[0090] (1)

[0091] in, ;

[0092] In the formula, For nodes , For nodes , For nodes The start time, For nodes End time, For employees , For employees At the node and nodes Value transfer between intervals, For nodes and nodes The necessary transfer time between them For employees Initial position, For employees In employees Initial position and nodes Value transfer between intervals, For employees gather, This is the set of locations for the evacuation mission. This is the set of locations where the task is to be completed. Gather at the parking lot, For a set of nodes, This is the set of locations for both the rendezvous and evacuation missions.

[0093] Step 103: Call the commercial solver to solve the scheduling network flow model based on the scheduling request under scheduling constraints, and output the scheduling plan.

[0094] It should be noted that, since the single-stage method has a clear mathematical model, after determining the relevant scheduling parameters required for the scheduling network flow model based on the scheduling request, a scheduling plan can be output by using mathematical programming methods under scheduling constraints using commercial solvers such as Gurobi. The scheduling constraints include:

[0095]

[0096] In the formula, For employees to nodes Transfer value, For employees The set of initial position points For employees To the node Transfer value, For nodes , For employees At the node and nodes Value transfer between intervals, For employees At the node and nodes Value transfer between intervals, For parking lot, For nodes parking lot before start time The number of available vehicles, For parking lot The initial number of available vehicles, For nodes The set of locations where the tasks were to be picked up before the start time. For employees At the parking lot and nodes Value transfer between intervals, For nodes The set of locations of the evacuation mission prior to the start time. For employees at the node Car Park Value transfer between intervals, For nodes parking lot before start time Available number of parking spaces For parking lot The initial number of available parking spaces, For employees At the node Car Park Value transfer between intervals, For employees At the parking lot and nodes the transfer between the node and the depot, is the total capacity, is the employee leaves the node at the time, is the maximum value, is the employee leaves the depot goes to the node at the time, is the depot and the node necessary transfer time, is the end position of the employee, is the set of location points of the docking task and the withdrawal task before the start time of the node , is the node , is the employee transfers between the node and the depot ;

[0097] In the above constraint conditions, formula (2) indicates that each task has an employee to complete, formula (3) and formula (4) indicate the supply and demand relationship of the transfer between the employees, formula (5) indicates the flow balance of the employees with respect to the depot points and the task points, formula (6), formula (7) and formula (8) ensure the capacity conservation of each depot before each task, the process of the vehicle dynamic parking and picking up is described in the form of a static equation, formula (9), formula (10), formula (11) and formula (12) indicate the time flow description of the personnel in the graph, formula (13) indicates that the employee can only start from the initial position once in the decision-making round, formula (14) and formula (15) solve the constraint description of the employee breaking the sub-circle for multiple visits to the depot point, if the employee arrives at the depot point from a task point and there is a task point with earlier start time visited from the depot point, then the employee is required to have earlier inflow and later outflow with respect to the depot, formula (16), formula (17) and formula (18) are the prohibition of the employee visiting route, that is, there is no docking-to-docking, withdrawal-to-withdrawal, and depot-to-depot visiting arc line in the supply and demand pair; In the scheduling plan determination process, the multi-depot with limited capacity is included, the multiple visits to the depot caused by the multiple use of the vehicle are considered, and the characteristics of the dynamic change of the number of vehicles and parking spaces in the depot are abstractly expressed through the mathematical model, which simplifies the complex problem into a form easier to understand and analyze, which helps to find the optimal solution in a limited time, thereby providing an accurate scheduling plan for the scheduling problem.

[0098] The scheduling plan sample is as follows Figure 3As shown, there are two employees, two yards, 12 nodes, and 12 personnel transfer path arcs. The size of the scheduling task number represents the order of the start of the scheduling task. Task No. 3 is the corresponding evacuation task of the docking task No. 1, that is, the vehicle that docks in this decision cycle needs to be evacuated. The first employee starts from the initial position of the current decision and arrives at the yard, drives the vehicle to complete task No. 1, then takes the working trolley to return to the yard to drive the vehicle to complete task No. 2, and then takes the working trolley to task No. 4 to evacuate the special vehicle to the yard, and end the tasks in the current decision cycle. The second employee takes the working trolley from the initial position to task No. 3 to evacuate the vehicle for task No. 5, then takes the working trolley to the yard after completing the task to drive the vehicle to complete task No. 6, wherein the virtual point is the end of the journey of the employee, and no actual transfer is required.

[0099] Step 104, when the scheduling scale is large, a greedy algorithm is used to construct an initial solution based on the scheduling request, and an initial yard record table is generated according to the initial solution.

[0100] It should be noted that the man-vehicle separation problem can be regarded as a variant of the many-to-many pickup and delivery problem, and the relevant solution strategies of the vehicle routing problem can be referred to. GD Konstantakopoulos and others have made a detailed summary of the algorithm for the VRP problem. In practical applications, an approximate solution strategy is often used to find a satisfactory solution to meet the timeliness requirements in actual operation. In this embodiment, a second scheduling strategy based on an adaptive large neighborhood search algorithm is considered to cope with large-scale scheduling problems. The algorithm framework of the adaptive large neighborhood search (ALNS) algorithm can be referred to Table 3.

[0101] Table 3 ALNS flow information table

[0102]

[0103] In the man-vehicle separation problem, the task vehicle source or destination is determined by pairing, that is, the destruction and repair of the problem are carried out on the pair nodes. According to the vehicle trip, the task pair of the vehicle trip is as follows: Figure 4The three cases shown can be divided into three cases: 1) the yard relies on pairing: moving from the yard point to the task point of the approach; 2) the yard pairing: moving from the task point of the departure to the yard point; 3) the yard pairing: moving from the task point of the departure to the task point of the approach; Therefore, in the ALNS algorithm: first, an initial solution of the problem is constructed by a greedy algorithm, and it is set as the current solution, then in each subsequent iteration, remove the q task pairs from the current solution in the form of roulette, then randomly select the insertion algorithm to re-insert the nodes of the q task pairs into the current solution, and then obtain a new solution, in the judgment of the new solution, the annealing mechanism is introduced to avoid the algorithm from falling into local optimum too early, when the algorithm iterates to the specified number of times, the algorithm is terminated, and the result is output;

[0104] In specific implementation, the main construction steps of the initial solution include: 1) initializing the candidate task pair set , wherein each candidate task pair is one of the aforementioned three pairings, ; 2) randomly selecting a task pair from the set , traversing all resource paths and calculating the cost increment of inserting into each path position, finding the position with the smallest increment, inserting the task pair into the corresponding path and updating the path state, removing from the set , setting , and jumping to step 2); 3) until , output the initial solution;

[0105] In the scheduling plan determination process, selecting a yard for scheduling tasks at each time will involve the return of the usage of a certain yard, such as returning the vehicles of the yard for the approach task at a certain time to reduce the number of parking spaces, and passing this impact to the subsequent records, and for the departure task at a certain time, it is to return the parking spaces of the yard and reduce the number of vehicles in the yard, and pass this impact to the records of all subsequent times, so it is necessary to judge whether the records of the most recent previous time based on the pickup or parking time and all subsequent times can meet the feasibility requirements; This feasibility judgment mainly considers two factors, one is the limitation of the number of parking spaces and vehicles in the yard itself, and the other is whether the current pickup or parking will affect the subsequent time period, by dynamically recording the yard usage and performing feasibility judgment when selecting a yard, the rationality of the employees' pickup and parking during the entire operation process can be ensured;

[0106] Therefore, after determining the initial solution, an initial parking lot record table is generated. This table records the number of available vehicles and the number of available parking spaces remaining at each time point after a vehicle has been picked up or parked in the parking lot during scheduling according to the initial solution. For example, the parking lot record table is dynamically updated based on the initial number of available vehicles and the initial number of available parking spaces recorded at the initial time t0 before the initial solution is constructed, according to the path specified in the initial solution. Figure 5 As shown on the left, if there is a vehicle pick-up task at time t2, it is first necessary to determine whether the number of available vehicles remaining at time t1 is greater than 1. If this condition is met, it is then determined whether the vehicle retrieval at time t2 will affect the usage at subsequent times t3 and t4. If the retrieval condition is met, the vehicle can be retrieved from this parking lot. After retrieval, the change in the number of parking spaces at time t2 is recorded. If the number of vehicles decreases, the number of parking spaces increases, and this update is passed on to the records at subsequent times to ensure the uniformity of the overall parking lot capacity. Conversely, as... Figure 5 As shown on the right, if there is an evacuation task requiring vehicles at time t2, it is first necessary to determine whether the number of available parking spaces remaining at time t1 is greater than 1. If this condition is met, it is then determined whether the use of parking spaces at time t2 will affect the use at subsequent times t3 and t4. If the use condition is met, parking can be selected in this parking lot, and the change in the number of parking spaces at time t2 is recorded. If the number of parking spaces decreases, the number of vehicles increases, and this update is passed on to the records of subsequent times.

[0107] Step 105: Optimize the initial solution based on the initial parking lot record table and the employee transfer time cost function using the adaptive large neighborhood search algorithm to determine the scheduling plan.

[0108] It should be noted that, based on the iterative optimization of the adaptive large neighborhood search algorithm, appropriate vehicle sources or destinations are selected for each scheduling task in the scheduling request. By introducing a parking lot record table to track changes in parking lot capacity, it is ensured that vehicles and parking lot resources can be reasonably allocated for each task scheduling. At the same time, the scheduling constraints caused by parking lot capacity limitations are simplified, so that the increase in problem size does not lead to a multiple increase in solution time, which helps to complete the scheduling task more smoothly; Step 105 includes the following sub-steps:

[0109] S1. Let the initial solution be the current solution, take the initial solution as the optimal solution, and initialize the destruction operator, repair operator, operator weight, operator selection count, and operator selection score.

[0110] S2. Select the target destruction operator and the target repair operator by roulette wheel according to the weight of each operator.

[0111] S3, destroying the current solution based on the initial yard record table and the employee transfer time cost function by using a target destruction operator, determining the removal task pair, the destroyed solution, the destroyed yard record table and the destroyed employee transfer time cost.

[0112] It should be noted that the destruction operator is any one of the random removal method, the worst path removal method, the Shaw removal method and the worst value removal method, and their common purpose is to determine q task pairs from the current solution as removal task pairs, and to form a destroyed path, i.e. a destroyed solution, through a path update mechanism after removing the removal task pairs, wherein a random parameter p is introduced to increase the randomness of the Shaw removal method and the worst value removal method, so as to expand the search range and avoid falling into a local optimum.

[0113] In one specific embodiment, the sub-step S3 comprises:

[0114] S31, randomly selecting at least one task pair from the current solution as a removal task pair for removal, and determining a destroyed solution.

[0115] It should be noted that the basic process of the random removal method is to randomly select q task pairs from the current solution, and then remove the q task pairs from the current solution to obtain a determined destroyed solution.

[0116] S32, or, based on the employee transfer time cost function, removing the task pair associated with the employee transfer path with the maximum employee transfer time cost in the current solution as the removal task pair, and constructing a destroyed solution.

[0117] It should be noted that the basic process of the worst path removal method is to select the path with the highest cost from the existing paths of the current solution, remove all task pairs on the path, if the number of removed task pairs is less than q pairs, then select the path with the second highest cost for removal, and repeat the process until the total number of removed task pairs is not less than q pairs.

[0118] S33, or, based on the relative distance of each task pair associated node, the node start time, the node end time and the node type in the current solution, calculating the similarity between task pairs, determining the removal task pair based on the highest similarity between task pairs for removal, and generating a destroyed solution.

[0119] It should be noted that the Shaw removal method is proposed by Shaw in 1997, in the present embodiment, the similarity of different task pairs is calculated through the common attributes between task pairs, removing the task with high similarity from the current path can help the subsequent path repair to obtain a better solution, and removing the task with low similarity will obtain a poor or original result; since the removed is the pair pair, there are starting point and ending point between the two nodes constituting a pair pair, when calculating the similarity between task pairs, the differences of starting point and ending point position difference, starting time difference, ending time difference and pair pair type difference are considered, and the corresponding weight is introduced for calculation:

[0120]

[0121] In the formula, is the similarity between task pairs, is the distance weight, is the relative distance of the node, is the starting time weight, is the ending time weight, is the type weight, is the type of node is the type of node

[0122] S34, or, the removal cost reduction amount of each task pair in the current solution is determined by using the employee transfer time cost function, the task pair with the maximum removal cost reduction amount is selected as the removal task pair for removal, and the destroyed solution is determined.

[0123] It should be noted that the worst value removal method first calculates the first employee transfer time cost of the current solution by using the employee transfer time cost function, evaluates the second employee transfer time cost after removing a certain task pair through the employee transfer time cost function, and determines a plurality of removal cost reduction amounts by using the difference operation of the first employee transfer time cost and the second employee transfer time cost, selects the task pair with the most cost reduction as the removal task pair in the current solution to remove and generate the destroyed solution, and determines the destroyed solution.

[0124] S35, updating the initial yard record table based on the destroyed solution to determine the destroyed yard record table of the destroyed solution, and calculating the destroyed employee transfer time cost associated with the destroyed solution according to the employee transfer time cost function.

[0125] It should be noted that in the process of generating the destroyed solution, whether the yard involved in the path needs to be reselected after removing a certain task pair depends on whether the vehicle of the task is at the original yard to pick up or park, exemplarily, for example Figure 6 ​​​As shown, when the destroyed task pair is located in the middle of the current path of the employee, the previous task pair is the withdrawal parking pair, and the next task pair is the parking approach pair. After the withdrawal approach pair is destroyed, the task vehicles in the previous and next pairs may have better choices in terms of the source and destination, and thus the reselection of the parking lot is involved in the process. The principle of the reselection of the parking lot is to find a parking lot point with less transfer time cost among three points to improve the time transfer cost after the removal of the task pair. After the parking lots of the previous and next tasks are selected, the withdrawal parking pair and the parking approach pair can be formed. However, the two tasks themselves can also form a withdrawal approach pair, and thus the transfer time of the two pairs is compared, and the pair with less transfer time cost is selected as the path after the pair is destroyed. This can be evaluated by the algorithm inside the destruction operator.

[0126] In addition to the use of the parking lot record table in the selection of the parking lot for the task, the parking lot record table is also used in the destruction of the path, and the return of the use amount of the parking lot is involved. Therefore, the update of the initial parking lot record table based on the path of the destruction solution can determine the destruction parking lot record table.

[0127] S4, constructing a planned task pair of each removal node in the removal task pair, and judging the feasibility of the planned task pair based on the destruction parking lot record table to determine the feasible task pair and the corresponding feasible insertion position.

[0128] In one specific embodiment, the sub-step S4 includes:

[0129] S41, constructing a planned task pair of each removal node in the removal task pair;

[0130] S42, inserting any planned task pair into the destruction solution to update the destruction parking lot record table to obtain the corresponding repair parking lot record table;

[0131] S43, using the planned task pair associated with the repair parking lot record table satisfying the non-negativity constraint as the feasible task pair, and determining the corresponding feasible insertion position.

[0132] It should be noted that each removal task pair includes two removal nodes, and the removal node is a node that constitutes a task pair. Taking the withdrawal approach pair as an example, the removal nodes refer to the withdrawal task and the approach task. The pairing form of the planned task pair is consistent with the pairing form of the task pair described above. Exemplarily: 1) if the removal node is a withdrawal task, a withdrawal approach pair and a withdrawal parking pair are formed; 2) if the removal node is an approach task, a withdrawal approach pair and a parking approach pair are formed; 3) if the removal node is a parking lot, a withdrawal parking pair and a parking approach pair are formed.

[0133] For the repair operator, inserting a task in the existing staff path can be understood as a look-back operation, and for this purpose, any path position in the broken solution is inserted into the planned task pair, and the feasibility of inserting the planned task pair is judged. When the available vehicle number and the available vehicle position in the repair yard record table obtained by inserting and updating the broken yard record table do not appear negative values at each time, it is considered that the planned task pair can be inserted as a feasible task pair at the corresponding path position as a feasible insertion position for repair.

[0134] S5, combining the target repair operator with the broken yard record table, the staff transfer time cost function, the removed task pair and the broken staff transfer time cost to repair the broken solution, and outputting a new solution.

[0135] It should be noted that, similar to the path update after the damage, in the repair operator, the current task to be inserted and the adjacent tasks before and after are also sought for better vehicle sources or destinations according to the position of the insertion point; for example, as shown in Figure 7 , the staff needs to park the vehicle of the previous withdrawal task at the yard, and then come to the withdrawal task to withdraw the vehicle of the task, and the next task is a docking task, and naturally considers whether the vehicle of the previous withdrawal task has other destinations, and whether the vehicle of the next docking task has other sources; after selecting the best vehicle sources for the tasks before and after, different pairing schemes are formed, and according to the repair operator, the removed removed task pair is re-inserted into the current path according to certain priority rules to select a suitable repair scheme to output a new solution. In this embodiment, the target repair operator includes the optimal insertion method and the regret value insertion method.

[0136] In one specific embodiment, the sub-step S5 includes:

[0137] S51, using each feasible task pair to form a removed set, inserting each feasible task pair in the broken solution into any corresponding feasible insertion position, and determining the corresponding cost increment based on the staff transfer time cost function and the broken staff transfer time cost.

[0138] It should be noted that, regardless of whether the target repair operator adopts the optimal insertion method or the regret value insertion method, the cost increment of each feasible task pair inserted into the broken solution needs to be calculated. In the specific implementation process, for each feasible task pair, the repair staff transfer time cost of inserting each feasible insertion position in the broken solution is calculated based on the staff transfer time cost function, and the repair staff transfer time cost and the broken staff transfer time cost are used for difference operation to determine the cost increment of each feasible task pair inserted into the broken solution.

[0139] S52, select the feasible task pair with the minimum cost increment as the insertable task pair, and insert the insertable task pair into the broken solution at the feasible insertion position associated with the minimum cost increment to determine a new broken solution.

[0140] It should be noted that in the optimal insertion method, the cost increment of the total cost after inserting all feasible task pairs in the removal set into the existing path according to the corresponding feasible insertion position is calculated, the feasible task pair corresponding to the minimum cost increment among all feasible task pairs in the removal set is selected as the insertable task pair, and the feasible insertion position corresponding to the minimum cost increment represents the optimal insertion position of the insertable task pair, so the insertable task pair is inserted into the corresponding position, thereby determining a new broken solution.

[0141] S53, or, the difference between the minimum cost increment and the second minimum cost increment of each feasible task pair is taken as the regret value of each feasible task pair, the feasible task pair with the maximum regret value is taken as the insertable task pair, and the insertable task pair is inserted into the broken solution at the feasible insertion position associated with the minimum cost increment of the insertable task pair to determine a new broken solution.

[0142] It should be noted that when the target repair operator adopts the regret value insertion method, the increase in the overall path cost after inserting a certain feasible task pair needs to be determined to determine the best insertion point, and the relevant regret value needs to be determined to determine which feasible task pair to insert; in this embodiment, for each feasible task pair The removal set calculates the repair staff transfer time cost based on the staff transfer time cost function, and determines the cost increment of each feasible task pair inserted into the broken solution by difference operation of each repair staff transfer time cost and the broken staff transfer time cost , sorts each cost increment in ascending order according to the feasible task pair, and takes the difference between the minimum cost increment and the second minimum cost increment of each feasible task pair as the regret value of each feasible task pair , the feasible task pair with the maximum regret value is taken as the insertable task pair, and the insertable task pair is inserted into the position corresponding to the minimum cost increment to determine a new broken solution.

[0143] S54, delete the feasible task pair to which the removal node associated with the insertable task pair belongs from the removal set to generate a new removal set, and insert the feasible task pairs in the new removal set into any corresponding feasible insertion position in the new broken solution until the removal set is empty, and output the new solution.

[0144] It should be noted that, regardless of whether the target repair operator adopts the optimal insertion method or the regret value insertion method, after determining the removable set any feasible task pair as an insertable task pair and inserting it, all feasible task pairs to which the removable nodes associated with the insertable task pair belong in the removable set are deleted to generate a new removable set, and the feasible task pairs in the new removable set are inserted in the new destruction solution to perform insertion repair, until the removable set is empty, at which time each removable node of the removable task pair is re-inserted, and a new solution is output.

[0145] S6, based on the comparison result between the new solution and the optimal solution, the current solution, determine the new current solution, update the operator selection score and the operator selection times, and count the iteration times.

[0146] It should be noted that, in specific implementation, if the new solution is better than the optimal solution, the new solution is the new optimal solution, the new solution is the new current solution, the operator selection score and the operator selection times are updated, and the iteration times are counted; if the new solution is better than the current solution, the new solution is the new current solution, the operator selection score and the operator selection times are updated, and the iteration times are counted; if the new solution is not better than the current solution, the new solution is accepted or rejected according to the simulated annealing criterion, if accepted, the new solution is the new current solution, the operator selection score and the operator selection times are updated, and the iteration times are counted.

[0147] S7, if the iteration times are less than the termination times threshold, when the iteration times meet the operator update condition, the operator weights are updated according to the operator selection score and the operator selection times, and the new target destruction operator and the new target repair operator are selected through roulette according to the operator weights, until the iteration times equal the termination times threshold, and the optimal solution is output as the scheduling plan.

[0148] It should be noted that the termination condition of the adaptive large neighborhood search algorithm iteration is set to the iteration times reaching the termination times threshold, and the optimal solution is output as the scheduling plan when the iteration is terminated; when the iteration times are less than the termination times threshold, if the iteration times of updating the destruction operator and the repair operator are reached, the operator weights are updated according to the operator selection score and the operator selection times; in the iteration process, the operator weight updating of the destruction operator and the repair operator is mainly based on the performance of the new solution, and the scores corresponding to the new optimal solution, the solution better than the current solution, and the solution worse than the current solution but accepted are respectively 、 、 In specific implementation, the operator weight updating can be a period of 30 iterations, and the weights of each destruction operator and each repair operator are adjusted after a period ends, and the updating process is as follows:

[0149] ;

[0150] In the formula, is the operator, is the first period, is the impact factor, is the operator is the first operator weight in the period, is the operator is the first operator weight in the period, is the operator is the first operator selection frequency in the period, is the operator is the first total score of operator selection score in the period.

[0151] In order to verify the optimization effect of the multi-vehicle personnel and vehicle collaborative scheduling method proposed in this embodiment, actual business data of a certain international airport in South China is used for data experiment:

[0152] Select data of different time periods for several days as each example for testing. For relevant service information, the corresponding parking lot information is divided into 6 parking lots with a total capacity of 72 according to the existing airport corresponding open-air parking lot, special vehicle garage, etc. The others are tested according to the number of available personnel and available vehicles at different time periods. Small-scale examples are provided with W12-P6-H15 and W28-P9-H40, medium-scale examples are provided with W42-P10-H30 and W62-P15-H37, and large-scale examples are provided with W80-P18-H44, W99-P22-H54 and W122-P22-H54. The corresponding example number "W" represents the task, "P" represents the personnel, "H" represents the vehicle, and the number after it represents the number of the resource. For example, "W12-P6-H15" represents that the example information is 12 tasks, 6 employees and 15 vehicles.

[0153] All models and methods are realized by java programming. The single-stage method has a clear mathematical model and is a mixed integer programming model, so Gurobi is called to solve it respectively. It is worth mentioning that similar to the method of storing node exploration information to the disk after the physical memory is used up in CPLEX, the node exploration file exceeding the specified size can also be written to the disk by setting NodefileStart in Gurobi. In addition, the scheduling is based on the scheduling sub-problem under the rolling horizon scheduling framework. In order to ensure the timeliness of the solution, the overall solution time limit is set to 360s. The cpu used in this test is 12th Gen Intel(R) Core(TM) i5-12500H 2.50 GHz, and the RAM is 16G.

[0154] 1) Single-stage model convergence verification

[0155] The scheduling network flow model is applied to the example W62-P15-H37, as shown in the convergence process diagram Figure 8 The solver finds the initial feasible solution at the root relaxation stage 182s due to the large scale of constraints and the complexity of the model. The actual GAP at the final time is 13.80% due to the scaling. In addition, the large feasible region caused by the large number of variables and constraints is the main factor that causes the slow convergence of the relaxation solution. Looking at the constraints and structure of the entire model, the yard sub-circle constraint is a combination constraint that grows exponentially with the increase of the problem size. Therefore, in the subsequent consideration, the adaptive large neighborhood search algorithm commonly used in vehicle routing problems is used to solve this problem.

[0156] 2) Stability and convergence of ALNS results

[0157] We analyze the stability and convergence of the algorithm under different example sizes. The corresponding parameters of the algorithm are shown in Table 4:

[0158] Table 4 ALNS parameter information table

[0159]

[0160] Overall, in the task similarity calculation, only the same type of tasks have the possibility of mutual comparison, so we give a larger value. In order to ensure the solving efficiency within a certain time at the overall algorithm iteration level, 420 is selected as the termination threshold of the overall iteration.

[0161] Through the ALNS example variance diagram as shown in Figure 9 It is obvious that with the increase of the overall example size, the overall solution variance increases with the increase of the size, but the overall variance is around 2.

[0162] From the ALNS example convergence process as shown in Figure 10 After the optimal insertion generates the initial solution, the algorithm will gradually converge with the number of iterations. Overall, the algorithm has a large improvement space at the beginning, and gradually converges to a certain value in the subsequent.

[0163] 3) Changes of single-stage constraints and variable numbers with example size

[0164] Since the single-stage method has a clear mathematical model, it can be solved by mathematical programming. The single-stage scheduling network flow model is solved by the commercial solver Gurobi under different examples, and the variable number and constraint number of the single-stage model are obtained. The variable number and constraint number change curve with the example size is shown in Figure 11 .

[0165] Depend on Figure 11 It can be seen that for small to medium-sized problems, the number of constraints and variables in the single-stage model remains at a low level, and the growth trend of both is relatively stable as the problem size increases. At this time, although the number of constraints is greater than the number of variables, the overall size has not exceeded the processing capacity of commercial solvers (such as Gurobi), so the optimal solution can be found in a reasonable time. This shows that for small to medium-sized problems, the single-stage solution strategy exhibits high solution efficiency and can meet the needs of scheduling problems.

[0166] However, as the problem size increases, the number of constraints and variables in the single-stage model exhibit different growth trends due to the combined constraints involved in selecting the sub-loops of the parking lot in a single stage. The number of constraints grows exponentially with increasing case size, while the number of variables grows relatively slowly. This results in the number of constraints in the single-stage model being far greater than the number of variables. This imbalance causes the solver to handle a large number of constraints, increasing the complexity of the solution and significantly slowing down the solution speed. At this point, the efficiency of the single-stage model begins to decline, and it becomes impossible to find a high-quality solution within a reasonable time.

[0167] Therefore, single-stage models can solve problems efficiently and perform well in small and medium-sized problems. However, in large-scale problems, the solution efficiency drops significantly as the number of constraints increases rapidly. Therefore, when dealing with large-scale scheduling problems, optimization algorithms (such as the ALNS algorithm) are needed to improve the solution efficiency and reduce the solution time.

[0168] 4) Comparison of the two in terms of solution methods and efficiency

[0169] The solution results determined by the first scheduling strategy using the Gurobi algorithm are compared with the solution results determined by the second scheduling strategy using the ALNS algorithm. The comparison results are shown in Table 5.

[0170] Table 5 ALNS Parameter Information Table

[0171]

[0172] In Table 5, "UB" and "LB" represent the feasible solution and lower bound of Gurobi results respectively, "Time" is the solving time of each strategy, "Gap" is calculated by the feasible solution and lower bound, "--" means that the solving strategy failed to find a feasible solution, "*" means that Gurobi found the optimal solution of the current example, the bold data represents the better one of the two results, "Best" represents the optimal value of the 10 times solving results of the ALNS algorithm, "Avg" represents the average value of the 10 times solving results of the ALNS algorithm, "RPD" (relative percentage deviation) represents the deviation percentage of the solving results of each strategy from the current optimal solution, and the calculation process is shown in formula (19):

[0173] (19)

[0174] We compare the relative situation of the two strategies by comparing the situation of each strategy:

[0175] As shown in Table 5, in small-scale examples (such as W12-P6-H15, W28-P9-H40), the single-stage model solving and ALNW algorithm can find the same optimal solution (for example, Obj is 28.5 and 33.84). This shows that in small-scale problems, the single-stage solving strategy of Gurobi and the ALNW algorithm have equivalent solving accuracy; at the same time, the solving time of the single-stage model is slightly longer than that of ALNW (for example, in W12-P6-H15, the single-stage is 0.73 seconds, while the ALNW is only 0.2 seconds), which shows that ALNW has a certain time advantage in small-scale problems;

[0176] In medium-scale problems (such as W42-P10-H30, W62-P15-H37), the single-stage model can still find a solution close to the optimal solution, for example, the single-stage result Obj of W42-P10-H30 is 59.5, and the optimal solution of ALNW is 68.8, although there is a gap between them, but the Gap of the single-stage model is 0, which shows that the accuracy of its solution is high; but as the problem size increases (such as W62-P15-H37), the solving error of the single-stage model begins to increase significantly, and the Gap reaches 13.80%, but the optimal solution of the single-stage model is still better than that of the ALNS in terms of solving accuracy, and it also provides a lower bound (LB) for further optimization. At this time, the ALNW algorithm shows good flexibility and solving ability, with an RPD of only 5.27%, and the solving time is significantly reduced compared with the single-stage, such as the 33.3 seconds in W62-P15-H37 and the 360 seconds in the large-scale problem W62-P15-H37, there is a big gap;

[0177] ​In large-scale problems (such as W80-P18-H44, W99-P22-H54, W122-P22-H54), the single-stage model fails to find a feasible solution (such as W80 and subsequent examples are shown as “--”), while the ALNW algorithm can still give a feasible solution in a shorter time, and as the problem size increases, the deviation of the ALNW solution from the optimal solution is smaller, and the RPD is controlled within 11.72% (such as W80) or even close to the optimal (such as the RPD of W99 is 0.53%);

[0178] Therefore, the single-stage model shows high solving accuracy in small and medium-sized problems, and can find or approach the optimal solution. However, in large-scale problems, the solving ability is significantly reduced due to the rapid increase of constraints, and even no feasible solution can be found. The ALNS algorithm shows obvious time advantage in each scale problem, and as the problem size increases, the ALNS algorithm gradually dominates in solving efficiency and solution accuracy. Especially in large-scale problems, ALNS can effectively balance the solution time and solution quality, and show superior solving ability.

[0179] In the embodiment of the present application, the dynamic change of the number of vehicle parking spaces in the yard is tracked by scheduling the network flow model and the yard record table in the scheduling strategy, and the scheduling of employees and vehicles is considered based on the scheduling request starting from the scheduling task, while optimizing the utilization of three types of resources of vehicles, yards and employees. In the environment of multiple yards and limited capacity, the problem of vehicle congestion entering and leaving the yard can be effectively solved, resource conflicts are reduced, the utilization efficiency of the yard is improved, the work coordination effect under the mode of separation of people and vehicles is improved, and corresponding scheduling strategies are applied to small and medium-sized and large-scale scheduling services. On the basis of ensuring the accuracy of the scheduling plan, the overall scheduling efficiency is improved, thereby improving the scheduling reliability of the people-vehicle separation service, and the complex scheduling demand of the peak period in the actual business of the airport can be well adapted.

[0180] Please refer to Figure 12 , Figure 12 The structural block diagram of the multi-yard personnel and vehicle collaborative scheduling system provided by the embodiment of the present application is shown.

[0181] The multi-yard personnel and vehicle collaborative scheduling system provided by the embodiment of the present application comprises:

[0182] The scale determination module 1201 is configured to determine the scheduling scale according to a comparison result of the task quantity of the scheduling task associated with the scheduling request and the scale quantity threshold value.

[0183] The model construction module 1202 is configured to, when the scheduling scale is small and medium-sized, take the minimum employee transfer time as the target, adopt personnel, yards and scheduling tasks as nodes, and construct a scheduling network flow model based on an employee transfer time cost function.

[0184] The first strategy determination module 1203 is configured to invoke a commercial solver to solve the scheduling network flow model based on the scheduling request under the scheduling constraint condition, and output a scheduling plan.

[0185] The initial data construction module 1204 is configured to, when the scheduling scale is large-scale, construct an initial solution and an initial yard record table based on the scheduling request by using a greedy algorithm.

[0186] The second strategy determination module 1205 is configured to optimize the initial solution based on the initial yard record table and the employee transfer time cost function by using an adaptive large neighborhood search algorithm, and determine the scheduling plan.

[0187] Further, the second strategy determination module 1205 comprises:

[0188] The initialization unit is configured to set the initial solution as the current solution, set the initial solution as the optimal solution, and initialize a destruction operator, a repair operator, an operator weight, an operator selection frequency and an operator selection score;

[0189] The operator selection unit is configured to select a target destruction operator and a target repair operator by roulette according to the operator weights;

[0190] The destruction operation unit is configured to destroy the current solution based on the initial yard record table and the employee transfer time cost function by using the target destruction operator, determine a removed task pair, a destroyed solution, a destroyed yard record table and a destroyed employee transfer time cost;

[0191] The insertion feasibility judgment unit is configured to construct a planned task pair of each removed node in the removed task pair, and perform feasibility judgment on the planned task pair based on the destroyed yard record table, to determine a feasible task pair and a corresponding feasible insertion position;

[0192] The repair operation unit is configured to repair the destroyed solution by using the target repair operator in combination with the destroyed yard record table, the employee transfer time cost function, the feasible task, the feasible insertion position and the destroyed employee transfer time cost, and output a new solution;

[0193] The evaluation unit is configured to determine a new current solution based on a comparison result between the new solution and the optimal solution and the current solution, update the operator selection score and the operator selection frequency, and count an iteration frequency;

[0194] The plan output unit is configured to, if the iteration frequency is less than a termination frequency threshold, update the operator weights according to the operator selection score and the operator selection frequency when the iteration frequency meets an operator update condition, and select a new target destruction operator and a new target repair operator by roulette according to the operator weights, until the iteration frequency is equal to the termination frequency threshold, and output the optimal solution as the scheduling plan.

[0195] Further, the destruction operation unit is specifically configured to:

[0196] randomly selecting at least one task pair in the current solution as a removal task pair for removal, and determining a destruction solution;

[0197] Alternatively, a task pair associated with a worker transfer path with the largest worker transfer time cost in the current solution is removed as a removal task pair based on a worker transfer time cost function, and a destruction solution is constructed;

[0198] Alternatively, similarity between task pairs is calculated based on relative distance of nodes associated with the task pairs, node start time difference, node end time difference, and node type difference in the current solution, a removal task pair is determined based on the highest similarity between task pairs, and a destruction solution is generated;

[0199] Alternatively, removal cost reduction of each task pair in the current solution is determined using a worker transfer time cost function, a task pair with the largest removal cost reduction is selected as a removal task pair for removal, and a destruction solution is determined;

[0200] The initial yard record table is updated based on the destruction solution to determine a destruction yard record table of the destruction solution, and a destruction worker transfer time cost associated with the destruction solution is calculated according to a worker transfer time cost function.

[0201] Further, the insertion feasibility judgment unit is specifically configured to:

[0202] constructing a planned task pair of each removal node in the removal task pair;

[0203] inserting any planned task pair into the destruction solution, and updating the destruction yard record table to obtain a corresponding repair yard record table;

[0204] using a planned task pair associated with the repair yard record table satisfying the non-negativity constraint as a feasible task pair, and determining a corresponding feasible insertion position.

[0205] Further, the repair operation unit is specifically configured to:

[0206] using each feasible task pair to form a removal set, inserting each feasible task pair into any corresponding feasible insertion position in the destruction solution, and determining a corresponding cost increment based on the worker transfer time cost function and the destruction worker transfer time cost;

[0207] selecting a feasible task pair with the smallest cost increment as an insertable task pair, and inserting the insertable task pair into the destruction solution according to the feasible insertion position associated with the smallest cost increment to determine a new destruction solution;

[0208] Or, the difference between the minimum cost increment of each feasible task pair and the corresponding second minimum cost increment is taken as the regret value of each feasible task pair, the feasible task pair with the maximum regret value is taken as the insertable task pair, and the insertable task pair is inserted into the broken solution according to the feasible insertion position associated with the minimum cost increment of the insertable task pair to determine a new broken solution;

[0209] The feasible task pair to which the removal node associated with the insertable task pair belongs is deleted from the new removal set to generate a new removal set, and the feasible task pairs in the new removal set are inserted into any corresponding feasible insertion position in the new broken solution until the removal set is empty, and the new solution is output.

[0210] Further, the employee transfer time cost function includes:

[0211] ;

[0212] The scheduling network flow model includes:

[0213] ;

[0214] The scheduling constraint conditions include:

[0215] ;

[0216] ;

[0217] ;

[0218] ;

[0219] ;

[0220] ;

[0221] ;

[0222] ;

[0223] ;

[0224] ;

[0225] ;

[0226] ;

[0227] ;

[0228] ;

[0229] ;

[0230] ;

[0231] ;

[0232] wherein, ;

[0233] wherein, is the start time of node , is the end time of node , is the start time of node , is the end time of node , is the employee , is the employee takes values between node and node , is the necessary transfer time between node and node , is the initial position of employee , is the set of values that the employee takes between the initial position of employee and node , is the set of employees , is the set of location points where the departure tasks are located, is the set of location points where the arrival tasks are located, is the set of depots, is the set of nodes, is the set of location points where the departure and arrival tasks are located, is the set of values that the employee takes to node , is the set of initial location points of employee , is the set of values that the employee takes to node , is the node , is the set of values that the employee takes between node and node , is the set of values that the employee takes between node and node the number of transfers between the yard and the node, the start time of the yard, the start time of the node the number of available vehicles of the yard before the start time of the node, the initial number of available vehicles of the yard, the start time of the node the set of location points where the docking task before the start time of the node is located, the start time of the node the set of location points where the departure task before the start time of the node is located, the start time of the node the set of location points where the docking task before the start time of the node is located, the start time of the node the set of location points where the departure task before the start time of the node is located, the start time of the node the number of available parking spaces of the yard before the start time of the node, the start time of the node the number of available parking spaces of the yard before the start time of the node, the start time of the node the number of available parking spaces of the yard before the start time of the node, the initial number of available parking spaces of the yard, the start time of the node the initial number of available parking spaces of the yard, the start time of the node the number of available parking spaces of the yard before the start time of the node, the start time of the node the number of available parking spaces of the yard before the start time of the node, the start time of the node the start time of the node the number of available parking spaces of the yard before the start time of the node, the start time of the node the number of available parking spaces of the yard before the start time of the node, the total capacity, the time at which the employee leaves the node, the maximum value, the time at which the employee leaves the yard to go to the node, the necessary transfer time between the yard and the node, the end location of the employee, the set of location points where the docking task and the departure task before the start time of the node are located, the start time of the node the set of location points where the docking task and the departure task before the start time of the node are located, the start time of the node the set of location points where the docking task and the departure task before the start time of the node are located, the start time of the node the start time of the node the set of location points where the docking task and the departure task before the start time of the node are located, the start time of the node the set of location points where the docking task and the departure task before the start time of the node are located, the start time of the node the set of location points where the docking task and the departure task before the start time of the node are located, the start time of the node the set of location points where the docking task and the departure task before the start time of the node are located, the start time of the node the start time of the node

[0234] The embodiment of the present application further provides a computer device, comprising a memory and a processor, the memory stores a computer program; the computer program is executed by the processor, so that the processor executes the steps of the multi-depot personnel and vehicle cooperative scheduling method according to any one of the above embodiments.

[0235] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program / instruction, and the computer program / instruction is executed by a processor to realize the steps of the multi-depot personnel and vehicle cooperative scheduling method according to any one of the above embodiments.

[0236] The embodiment of the present application further provides a computer program product, which comprises a computer program / instruction, and the computer program / instruction is executed by a processor to realize the steps of the multi-depot personnel and vehicle cooperative scheduling method according to any one of the above embodiments.

[0237] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, module and unit can refer to the corresponding process in the foregoing method embodiments, and will not be described here.

[0238] In several embodiments provided in the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, system or unit, and can be electrical, mechanical or other forms.

[0239] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0240] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of software functional unit.

[0241] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0242] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for coordinated scheduling of personnel and vehicles across multiple parking lots, characterized in that, include: The scheduling scale is determined by comparing the number of tasks associated with the scheduling request with the scale threshold. When the scheduling scale is small to medium, the goal is to minimize employee transfer time. Personnel, parking lots and scheduling tasks are used as nodes and a scheduling network flow model is constructed based on the employee transfer time cost function. The commercial solver is invoked to solve the scheduling network flow model based on the scheduling request under scheduling constraints, and a scheduling plan is output. When the scheduling scale is large, a greedy algorithm is used to construct an initial solution based on the scheduling request, and an initial parking lot record table is generated based on the initial solution. The initial solution is optimized based on the initial parking lot record table and the employee transfer time cost function using an adaptive large neighborhood search algorithm to determine the scheduling plan. The step of optimizing the initial solution based on the initial parking lot record table and the employee transfer time cost function using an adaptive large neighborhood search algorithm to determine the scheduling plan includes: Let the initial solution be the current solution, take the initial solution as the optimal solution, and initialize the destruction operator, repair operator, operator weight, operator selection count, and operator selection score; Based on the weights of each operator, the target destruction operator and the target repair operator are selected by roulette wheel selection; The target destruction operator is used to destroy the current solution based on the initial parking lot record table and the employee transfer time cost function, thereby determining the task pair to be removed, the solution to be destroyed, the parking lot record table to be destroyed, and the employee transfer time cost to be destroyed. Construct planned task pairs for each removal node in the removal task pair, and make a feasibility judgment on the planned task pairs based on the damage parking lot record table to determine feasible task pairs and corresponding feasible insertion positions. The damaged solution is repaired by combining the damaged parking lot record table, the employee transfer time cost function, the feasible task, the feasible insertion position, and the damaged employee transfer time cost with the target repair operator, and a new solution is output. Based on the comparison results between the new solution, the optimal solution, and the current solution, a new current solution is determined, the operator selection score and the number of operator selections are updated, and the number of iterations is counted. If the number of iterations is less than the termination threshold, when the number of iterations meets the operator update condition, the operator weight is updated according to the operator selection score and the number of operator selections. Then, a new target destruction operator and a new target repair operator are selected by roulette wheel selection according to the weight of each operator until the number of iterations equals the termination threshold. The optimal solution is then output as the scheduling plan.

2. The multi-parking lot personnel and vehicle collaborative scheduling method according to claim 1, characterized in that, The step of using the target destruction operator to destroy the current solution based on the initial parking lot record table and the employee transfer time cost function, and determining the task pairs to be removed, the solution to be destroyed, the parking lot record table to be destroyed, and the employee transfer time cost to be destroyed, includes: Randomly select at least one task pair from the current solution as a removal task pair to remove it, thereby determining the corrupted solution; Alternatively, based on the employee transfer time cost function, the task pair associated with the employee transfer path with the highest employee transfer time cost in the current solution can be removed as a removal task pair to construct a destructive solution; Alternatively, the similarity between task pairs can be calculated based on the relative distance between nodes associated with each task pair in the current solution, the difference in node start time, the difference in node end time, and the difference in node type. The task pair to be removed is determined based on the highest similarity between task pairs, and a destructive solution is generated. Alternatively, the employee transfer time cost function can be used to determine the reduction in removal cost of removing each task pair in the current solution, and the task pair with the largest reduction in removal cost can be selected as the task pair to be removed to determine the destructive solution. The initial parking lot record table is updated based on the destructive solution to determine the destructive parking lot record table of the destructive solution, and the destructive employee transfer time cost associated with the destructive solution is calculated according to the employee transfer time cost function.

3. The multi-parking lot personnel and vehicle collaborative scheduling method according to claim 1, characterized in that, The process of constructing planned task pairs for each removal node in the removal task pair, and performing feasibility assessments on the planned task pairs based on the vehicle destructor record table to determine feasible task pairs and corresponding feasible insertion positions includes: Construct planned task pairs for each removal node in the removal task pair; Insert any of the planned tasks into the destruction solution and update the destruction parking lot record table to obtain the corresponding repair parking lot record table; The planned task pairs associated with the repair yard record table that satisfy the non-negativity constraint are used as feasible task pairs, and the corresponding feasible insertion positions are determined.

4. The multi-parking lot personnel and vehicle collaborative scheduling method according to claim 1, characterized in that, The process involves using the target repair operator in conjunction with the damaged parking lot record table, the employee transfer time cost function, the feasible task, the feasible insertion position, and the damaged employee transfer time cost to repair the damaged solution and output a new solution, including: Each of the feasible task pairs is used to form a removal set. Each of the feasible task pairs is inserted into any corresponding feasible insertion position in the destructive solution, and the corresponding cost increment is determined based on the employee transfer time cost function and the destructive employee transfer time cost. Select the feasible task pair with the minimum cost increment as the insertable task pair, and insert the destructive solution into the insertable task pair according to the feasible insertion position associated with the minimum cost increment to determine the new destructive solution; Alternatively, the difference between the minimum cost increment and the corresponding second-smallest cost increment of each feasible task pair can be used as the regret value of each feasible task pair. The feasible task pair with the maximum regret value can be used as the insertable task pair. The insertable task pair can be inserted into the breaking solution according to the feasible insertion position associated with the minimum cost increment of the insertable task pair to determine the new breaking solution. Remove the feasible task pairs associated with the removal nodes of the insertable task pairs from the removal set to generate a new removal set. Insert the feasible task pairs in the new removal set into any corresponding feasible insertion position in the new broken solution until the removal set is empty. Then output the new solution.

5. The multi-parking lot personnel and vehicle collaborative scheduling method according to claim 1, characterized in that, The employee transfer time cost function includes: ； The scheduling network flow model includes: ； The scheduling constraints include: ； ； ； ； ； ； ； ； ； ； ； ； ； ； ； ； ； in, ; In the formula, For nodes , For nodes , For nodes The start time, For nodes End time, For employees , For employees At the node and nodes Value transfer between intervals, For nodes and nodes The necessary transfer time between them For employees Initial position, For employees In employees Initial position and nodes Value transfer between intervals, For employees gather, This is the set of locations for the evacuation mission. This is the set of locations where the task is to be completed. Gather at the parking lot, For a set of nodes, This is a set of locations for both the rendezvous and evacuation missions. For employees to nodes Transfer value, For employees The set of initial position points For employees To the node Transfer value, For nodes , For employees At the node and nodes Value transfer between intervals, For employees At the node and nodes Value transfer between intervals, For parking lot, For nodes parking lot before start time The number of available vehicles, For parking lot The initial number of available vehicles, For nodes The set of locations where the tasks were to be picked up before the start time. For employees At the parking lot and nodes Value transfer between intervals, For nodes The set of locations of the evacuation mission prior to the start time. For employees at the node Car Park Value transfer between intervals, For nodes parking lot before start time Available number of parking spaces For parking lot The initial number of available parking spaces, For employees At the node Car Park Value transfer between intervals, For employees At the parking lot and nodes Value transfer between intervals, For total capacity, For employees Leave node Time, It is the maximum value. For employees Leaving the parking lot Go to node Time, For parking lot and nodes The necessary transfer time between them For the employee's final position, For nodes The set of locations for the docking and evacuation missions prior to the start time. For nodes , For employees At the node Car Park Values ​​are transferred between intervals.

6. A multi-parking lot personnel and vehicle collaborative dispatching system, characterized in that, include: The size determination module is used to determine the scheduling size based on the comparison between the number of tasks associated with the scheduling request and the size threshold. The model building module is used to construct a scheduling network flow model when the scheduling scale is small to medium scale, with the goal of minimizing employee transfer time, using personnel, parking lots and scheduling tasks as nodes and based on the employee transfer time cost function; The first strategy determination module is used to call a commercial solver to solve the scheduling network flow model based on the scheduling request under scheduling constraints and output a scheduling plan; The initial data construction module is used to construct an initial solution and an initial parking lot record table based on the scheduling request using a greedy algorithm when the scheduling scale is large-scale. The second strategy determination module is used to optimize the initial solution based on the initial parking lot record table and the employee transfer time cost function using an adaptive large neighborhood search algorithm to determine the scheduling plan. The second strategy determination module includes: An initialization unit is used to set the initial solution as the current solution, take the initial solution as the optimal solution, and initialize the destruction operator, repair operator, operator weight, operator selection count, and operator selection score. The operator selection unit is used to select the target destruction operator and the target repair operator by roulette wheel according to the weights of each operator. The destruction calculation unit is used to destroy the current solution based on the initial parking lot record table and the employee transfer time cost function using the target destruction operator, and to determine the task pair to be removed, the solution to be destroyed, the parking lot record table to be destroyed, and the employee transfer time cost to be destroyed. The insertion feasibility judgment unit is used to construct the planned task pair for each removal node in the removal task pair, and to make a feasibility judgment on the planned task pair based on the damage parking lot record table to determine the feasible task pair and the corresponding feasible insertion position. The repair operation unit is used to repair the damaged solution by combining the target repair operator with the damaged parking lot record table, the employee transfer time cost function, the feasible task, the feasible insertion position, and the damaged employee transfer time cost, and output a new solution; An evaluation unit is used to determine a new current solution based on the comparison results between the new solution, the optimal solution, and the current solution, update the operator selection score and the number of operator selections, and count the number of iterations. The planning output unit is used to update the operator weights according to the operator selection score and the number of operator selections when the number of iterations meets the operator update condition, if the number of iterations is less than the termination number threshold, and select new target destruction operators and new target repair operators by roulette wheel selection according to each operator weight, until the number of iterations equals the termination number threshold, and output the optimal solution as the scheduling plan.

7. A computer device, characterized in that, The system includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the multi-parking lot personnel and vehicle collaborative scheduling method as described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instruction is executed by the processor, it implements the steps of the multi-parking lot personnel and vehicle collaborative scheduling method as described in any one of claims 1-5.

9. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the steps of the multi-parking lot personnel and vehicle collaborative scheduling method as described in any one of claims 1-5.

Images