An aircraft taxi scheduling method and system based on multi-objective optimization

Abstract

The application relates to the technical field of aircraft ground operation control, and discloses an aircraft taxiing scheduling method and system based on multi-target optimization, which comprises the following steps: a ground space-time topology model is constructed, ground nodes, taxiway sections and time discrete sets are combined to form space-time occupation units; a candidate taxiing scheme set containing multiple candidate paths is generated for each eVTOL, and binary decision variables are defined; a ground capacity and conflict constraint condition is constructed based on a capacity parameter, so that the number of occupations of any ground unit at any time is ensured to be less than an upper limit; a multi-target collaborative optimization model containing maximum throughput, minimum taxiing time and energy consumption and minimum conflict risk is constructed, a mixed integer linear programming method is used to solve the optimal taxiing scheme; and taxiing control instructions containing a time window, a path and a constraint time are generated and issued. The application realizes global collaborative scheduling and space-time conflict prevention of multiple eVTOLs.

Description

Technical Field

[0001] This invention relates to the field of aircraft ground operation control technology, and more specifically, to an aircraft taxiing scheduling method and system based on multi-objective optimization. Background Technology

[0002] With the development of urban air traffic and the commercialization of electric vertical takeoff and landing (eVTOL) aircraft, the concurrent taxiing scenarios of multiple eVTOLs at airports and vertical takeoff and landing sites are increasing. Existing ground taxiing control technologies mainly rely on manual operation by pilots or simple automatic path following to optimize the trajectory of a single aircraft or provide navigation assistance. Existing technologies have the following drawbacks: First, they cannot predict and avoid taxiing conflicts in advance in dynamic scenarios where multiple eVTOLs simultaneously enter and exit parking positions, boarding gates, and charging points, resulting in a high risk of spatiotemporal conflicts on the ground; second, they have not built a multi-aircraft collaborative taxiing scheduling logic, lack multi-aircraft collaborative scheduling capabilities, and cannot achieve capacity balancing, conflict resolution, and sequence optimization in high-density traffic scenarios; third, they have not established a capacity model for ground units and lack a unified capacity constraint mechanism, resulting in an unsystematic allocation of resources such as taxiway segments, intersections, and takeoff and landing pads, as well as congestion control. The root cause of the above problems is that existing technologies do not uniformly model the spatial location and time dimension of ground taxiing, and lack a global optimization scheduling mechanism for multi-aircraft collaboration. As a result, they cannot perform global collaborative scheduling, autonomous spatiotemporal conflict prevention and capacity balancing control for the ground taxiing of multiple eVTOLs under complex real-time conditions, leading to technical problems such as delays or safety hazards. Summary of the Invention

[0003] This invention provides an aircraft taxiing scheduling method and system based on multi-objective optimization, which solves the technical problems of ground capacity conflict, low taxiing efficiency and lack of collaborative optimization in the taxiing process of multiple eVTOLs in vertical take-off and landing sites in related technologies.

[0004] This invention provides an aircraft taxiing scheduling method based on multi-objective optimization, comprising: Obtain ground facility information of the vertical take-off and landing site, and construct a ground spatiotemporal topology model based on the ground facility information. The ground spatiotemporal topology model includes a ground topology network, a time discrete set, and a set of ground spatiotemporal occupancy units. Obtain taxiing mission information for multiple eVTOLs currently waiting to taxi; based on the ground spatiotemporal topology model and taxiing mission information, generate a set of candidate taxiing schemes for each eVTOL, and each candidate taxiing scheme contains the spatiotemporal occupancy sequence of the corresponding eVTOL. Based on the capacity parameters in the ground spatiotemporal topology model and the spatiotemporal occupancy sequence of each candidate taxiing scheme, ground capacity and conflict constraints are constructed. These constraints are used to limit the number of ground units occupied at any given time from not exceeding their capacity limit. Based on the ground capacity and conflict constraints of the candidate taxiing scheme set, a multi-objective collaborative optimization model is constructed to solve for the optimal decision variables and obtain the optimal taxiing scheme for each eVTOL. Under the premise of satisfying the constraints of ground capacity and conflict and the constraints of scheme selection, the unified optimization objective function is solved by the mixed integer linear programming method to obtain the optimal decision variables, and then determine the optimal taxiing scheme for each eVTOL. The multi-objective collaborative optimization model includes three optimization objectives: The primary optimization objective is to maximize the throughput of the ground process, which means maximizing the sum of the number of scheduled eVTOLs. The second optimization objective is to minimize coasting time and energy consumption, which means minimizing the weighted sum of the normalized coasting time and normalized energy consumption of each eVTOL; The third optimization objective is to minimize the risk of conflict. This is achieved by defining a congestion overflow function to represent the amount of overcapacity occupied by each ground unit in each time slice. The congestion overflow function is the larger of the difference between the actual occupied amount and the capacity limit and zero. The third optimization objective is to minimize the sum of the congestion overflow functions of all ground units in all time slices. The three optimization objectives are fused by weighting to construct a unified optimization objective function; For each candidate taxiing scheme of each eVTOL, a binary decision variable is defined. When the binary decision variable is 1, it means that the corresponding candidate taxiing scheme is selected. When the binary decision variable is 0, it means that the corresponding candidate taxiing scheme is not selected. For each eVTOL, a scheme selection constraint is set, which means that each eVTOL can select at most one taxiing scheme. Based on the optimal taxiing scheme of each eVTOL, taxiing control commands are generated and sent to the airborne controller of the corresponding eVTOL.

[0005] Furthermore, the ground topology network includes a set of ground nodes and a set of taxiway segments. The set of ground nodes includes parking position nodes, FATO landing pad nodes, charging position nodes, boarding gate nodes, and taxiway intersection nodes. Each element in the set of taxiway segments represents a taxiway between two adjacent ground nodes. The discrete time set includes multiple discrete time slices; The set of ground spatiotemporal occupancy units is used to represent the occupancy status of each ground unit in each time slice; The ground spatiotemporal topology model also includes capacity parameters for each ground unit, which represent the upper limit of the number of eVTOLs that a ground unit is allowed to occupy simultaneously at the same time.

[0006] Furthermore, generating a set of candidate taxiing schemes for each eVTOL includes: For each eVTOL, based on the starting position node and the target position node in its taxiing mission information, the shortest path algorithm is used to calculate the shortest path from the starting position node to the target position node and several shortest paths as candidate paths. By combining the taxiing speed parameters of each eVTOL, the spatiotemporal occupancy sequence corresponding to each candidate path is calculated, and a set of candidate taxiing schemes is generated.

[0007] Furthermore, the calculation process of the spatiotemporal occupancy sequence is as follows: For each ground cell in the candidate path, calculate the time required to pass through that ground cell based on the eVTOL's taxiing speed and the physical length of the ground cell; The time interval during which the eVTOL occupies the ground cell is determined based on the time when the eVTOL enters the ground cell and the time required for its passage. All discrete time slices within a time interval and their corresponding ground units are incorporated into the spatiotemporal occupancy sequence.

[0008] Furthermore, the construction of ground capacity and conflict constraints includes: Define an occupancy indication function, which is used to indicate whether a certain eVTOL occupies a certain ground cell in a certain time slice under a certain candidate taxiing scheme; Based on the occupancy indication function and the decision variables of each candidate taxiing scheme, ground capacity and conflict constraints are constructed. The ground capacity and conflict constraints mean that the weighted sum of the number of occupancy of the same ground unit by all eVTOLs under all candidate taxiing schemes in the same time slice does not exceed the capacity limit of that ground unit.

[0009] Furthermore, before constructing the multi-objective collaborative optimization model, the following steps are also included: The mean normalization method based on the range is adopted to normalize the gliding time and power consumption of each candidate gliding scheme to a preset range, so as to obtain the normalized gliding time and normalized power consumption. The coasting time is calculated by summing the ratios of the physical length of each ground unit in the candidate path to the eVTOL coasting speed. The energy consumption is calculated based on the product of the motor power and the coasting time during eVTOL coasting.

[0010] Furthermore, the taxiing control command includes: specifying a taxiing time window, allocating an optimal taxiing path, and setting constraint times for passing each key node; The specified taxiing time window is extracted from the spatiotemporal occupancy sequence of the optimal taxiing scheme, and is the interval from the minimum time value to the maximum time value in the spatiotemporal occupancy sequence; The optimal gliding path is obtained by arranging the ground units in the spatiotemporal occupancy sequence in chronological order. The key nodes include taxiway crossing nodes and take-off and landing pad nodes, and the constraint time is obtained by looking up the time value corresponding to each key node in the spatiotemporal occupancy sequence.

[0011] Furthermore, it also includes a rolling replanning step: When a dynamic environmental change event is detected, a rolling replanning process is triggered. The dynamic environmental change events include a new eVTOL joining the taxi queue, a taxiing delay of an eVTOL, a fault exit of an eVTOL, or a change in the status of ground facilities. Obtain the actual location and status information of each eVTOL at the current time, and update the set to be scheduled and the starting position node of each eVTOL; The steps of generating candidate taxiing schemes, constructing ground capacity and conflict constraints, solving the multi-objective collaborative optimization model, and issuing taxiing control commands are repeated.

[0012] This invention provides an aircraft taxiing scheduling system based on multi-objective optimization, comprising: The ground spatiotemporal topology model construction module is used to obtain ground facility information of the vertical take-off and landing site and construct a ground spatiotemporal topology model based on the ground facility information; The candidate taxiing scheme generation module is used to obtain taxiing mission information of multiple eVTOLs currently waiting to taxi, and generate a set of candidate taxiing schemes for each eVTOL based on the ground spatiotemporal topology model and taxiing mission information; The constraint construction module is used to construct ground capacity and conflict constraints based on the capacity parameters in the ground spatiotemporal topology model and the spatiotemporal occupancy sequence of each candidate taxiing scheme. The optimization module is used to construct a multi-objective collaborative optimization model based on the candidate taxiing scheme set and ground capacity and conflict constraints. It solves for the optimal decision variables to obtain the optimal taxiing scheme for each eVTOL. The multi-objective collaborative optimization model includes three optimization objectives: the first objective is to maximize ground process throughput; the second objective is to minimize taxiing time and energy consumption; and the third objective is to minimize conflict risk. A mixed-integer linear programming method is used to solve the unified optimization objective function. The instruction generation and distribution module is used to generate taxiing control instructions based on the optimal taxiing scheme of each eVTOL and distribute them to the airborne controller of the corresponding eVTOL.

[0013] The beneficial effects of this invention are as follows: This invention unifies the modeling of spatial location and time dimension by constructing a ground spatiotemporal topology model. By generating a set of candidate taxiing schemes and defining decision variables, the multi-aircraft cooperative taxiing scheduling problem is transformed into a combinatorial optimization problem. By constructing ground capacity and conflict constraints, potential conflict risks are eliminated in the pre-planning stage. By constructing a multi-objective cooperative optimization model, comprehensive optimization of throughput, taxiing time, energy consumption, and conflict risk is achieved. By using a rolling replanning mechanism to adapt to dynamic environmental changes, this invention solves the technical problems of existing technologies being unable to perform global cooperative scheduling, autonomous spatiotemporal conflict prevention, and capacity balancing control for multiple eVTOLs. It achieves the technical effects of reducing ground spatiotemporal conflict risks, improving multi-aircraft cooperative scheduling capabilities, and realizing unified capacity constraint control. Attached Figure Description

[0014] Figure 1 This is a flowchart of the aircraft taxiing scheduling method based on multi-objective optimization of the present invention. Detailed Implementation

[0015] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0016] At least one embodiment of the present invention discloses an aircraft taxiing scheduling method based on multi-objective optimization, such as... Figure 1 As shown, it includes the following steps: Step 1: Construct a ground spatiotemporal topology model.

[0017] Obtain ground facility information for the vertical take-off and landing site, and construct a ground spatiotemporal topology model based on the ground facility information.

[0018] The ground spatiotemporal topology model includes a ground topology network, a time discrete set, and a set of ground spatiotemporal occupancy units.

[0019] The terrestrial topology network is represented as ,in For the set of ground nodes, This is a collection of taxiway sections.

[0020] Time discrete set representation is ,in Indicates the first A discrete time slice This indicates the total number of time slices.

[0021] The set of ground spatiotemporal occupancy units is represented as Used to represent a certain ground unit In time The state of being occupied.

[0022] It should be noted that the above set of ground nodes This includes parking stand nodes, FATO landing pad nodes, charging station nodes, boarding gate nodes, and taxiway intersection nodes. The above-mentioned taxiway segments are combined. Each element in the dataset represents a taxiway between two adjacent ground nodes. The time slices in the above discrete time set can be set to a fixed time interval, such as 1 second or 0.5 seconds.

[0023] It should be noted that the above-mentioned spatiotemporal topology model for the ground also includes capacity parameters for each ground unit. For each ground unit... Its maximum capacity is The capacity limit indicates the maximum number of eVTOLs that a ground unit can occupy simultaneously at any given time. The capacity limit for taxiway sections is typically set to 1, while the capacity limit for landing pads is set to 1 or 2 depending on their physical dimensions.

[0024] Step 2: Generate a set of candidate gliding schemes.

[0025] Obtain taxiing mission information for multiple eVTOLs currently scheduled to taxi, and generate a set of candidate taxiing schemes for each eVTOL based on the ground spatiotemporal topology model and taxiing mission information.

[0026] For the The eVTOL system's taxiing mission information includes the starting position node and the target position node. Using Dijkstra's shortest path algorithm, based on the physical distances of each taxiway segment in the ground topology network, the shortest path from the starting position node to the target position node, along with several sub-shortest paths, are calculated as candidate paths.

[0027] Combined with the The taxiing speed parameters of the eVTOL are used to calculate the spatiotemporal occupancy sequence corresponding to each candidate path, and a candidate taxiing scheme set is generated. ,in , , They represent the first The first, second, and third eVTOL units One candidate gliding scheme This represents the total number of candidate gliding schemes.

[0028] Each candidate gliding scheme Defined as eVTOL under candidate coasting scheme The spatiotemporal occupancy sequence is represented as:

[0029] Furthermore, the calculation process of the above spatiotemporal occupancy sequence is as follows: For each ground cell in the candidate path, according to the first... taxiing speed of eVTOL Physical length of ground unit Calculate the time required to pass through the ground unit. Thus determining the first The sequence of ground cells occupied by the eVTOL in each time slice. Specifically, if the eVTOL... At any moment Entering the ground unit Then it occupies ground units The time interval is All discrete time slices within the time interval are compared with ground units. Composition of spacetime units All are included in the spatiotemporal occupancy sequence.

[0030] Define binary decision variables for each candidate taxiing scheme for each eVTOL aircraft. ,when When, it indicates the selection of a candidate gliding scheme. This indicates that no candidate gliding scheme is selected.

[0031] For each eVTOL, set the scheme selection constraints:

[0032] in, This represents the total number of candidate taxiing schemes. The scheme selection constraint indicates that each eVTOL can select at most one taxiing scheme.

[0033] Furthermore, the generation of the aforementioned second-shortest paths is implemented using the K-shortest path algorithm. Specifically, after obtaining the shortest path, each taxiway segment in the shortest path is sequentially deleted. Dijkstra's algorithm is then re-run on the ground topology network after each deletion to obtain candidate paths. All candidate paths are sorted according to their length, and several paths with the second-shortest lengths are selected as the second-shortest paths. By generating the second-shortest paths, diverse taxiway path options can be provided for each eVTOL aircraft, increasing scheduling flexibility during the optimization process.

[0034] Step 3: Construct ground capacity and conflict constraints.

[0035] Based on the capacity parameters in the ground spatiotemporal topology model and the spatiotemporal occupancy sequence of each candidate taxiing scheme, ground capacity and conflict constraints are constructed.

[0036] Define the occupancy indicator function :

[0037] in, eVTOL In candidate gliding schemes In the middle, at time Occupied ground units .

[0038] Based on the occupancy indication function, the ground capacity and conflict constraints are constructed as follows:

[0039] in, This indicates the total number of eVTOLs waiting to glide. This represents the total number of candidate gliding schemes.

[0040] Ground capacity and conflict constraints ensure that at any time Arbitrary ground unit The actual number of units occupied does not exceed its capacity limit. .

[0041] Furthermore, the constraints in the aforementioned ground capacity and conflict constraints cover all ground units and all time slices in the ground spatiotemporal topology model, that is, the set of ground nodes. Combined with the taxiway section Each ground unit in the union and time discrete sets Each time slice All of them must meet the ground capacity and conflict constraints.

[0042] It should be noted that the above ground capacity and conflict constraints transform taxiing conflicts in the physical world into mathematical constraints: when the same ground cell is simultaneously occupied by multiple selected eVTOLs, if the number of occupied cells exceeds the capacity limit, the ground capacity and conflict constraints are violated. By satisfying the ground capacity and conflict constraints during the optimization process, the risk of taxiing conflicts can be fundamentally eliminated.

[0043] Step 4: Construct and solve a multi-objective collaborative optimization model.

[0044] Based on the ground capacity and conflict constraints of the candidate taxiing scheme set, a multi-objective collaborative optimization model is constructed to solve for the optimal decision variables and obtain the optimal taxiing scheme for each eVTOL.

[0045] Before constructing a multi-objective collaborative optimization model, the coasting time and energy consumption data are preprocessed. A mean normalization method based on the range is used to preprocess the coasting time. and power consumption Normalized to The interval is used to obtain the normalized glide time. and normalized power consumption This eliminates the impact of dimensional differences on the optimization calculation.

[0046] It should be noted that the above-mentioned gliding time The calculation formula is:

[0047] in, For ground unit The physical length, For eVTOL The gliding speed.

[0048] It should be noted that the above-mentioned power consumption The calculation formula is:

[0049] in, For eVTOL Motor power during gliding.

[0050] Furthermore, the aforementioned motor power The motor power is obtained from the eVTOL's technical specifications, provided by the eVTOL manufacturer or determined through ground testing experiments. In ground testing experiments, the eVTOL is run at a typical taxiing speed on a straight taxiway, and the power consumption per unit time is recorded by the onboard battery management system to calculate the average motor power during taxiing.

[0051] The multi-objective collaborative optimization model includes three optimization objectives: The primary optimization objective is to maximize the throughput of the ground process. (Definition) Indicates the first Whether the eVTOL is scheduled, among which This represents the total number of candidate gliding strategies. The first optimization objective is expressed as:

[0052] in, This indicates the total number of eVTOLs waiting to glide.

[0053] The second optimization objective is to minimize gliding time and energy consumption. For each candidate gliding scheme... Calculate its normalized glide time and normalized power consumption The second optimization objective is expressed as:

[0054] in, This indicates the total number of eVTOLs waiting to glide. This represents the total number of candidate gliding schemes. and These are the weighting coefficients.

[0055] Furthermore, the aforementioned weighting coefficients and To balance the relative importance of coasting time and energy consumption in the second optimization objective, the sum of the two satisfies All of these are non-negative real numbers. In practical applications, when operational efficiency is a greater priority, the following settings can be configured: When a greater emphasis is placed on energy conservation, it can be set... .

[0056] The third optimization objective is to minimize the risk of conflict. Define a congestion overflow function. The congestion overflow function represents the ground unit. In time Overcapacity usage:

[0057] in, This indicates the total number of eVTOLs waiting to glide. This represents the total number of candidate gliding schemes. Represents ground unit In time The actual number of units occupied Represents ground unit The capacity limit. When the actual number of occupants exceeds the capacity limit. A positive value indicates that congestion overflow has occurred; otherwise... It is 0.

[0058] The third optimization objective is expressed as:

[0059] Furthermore, the aforementioned congestion overflow function It plays a soft constraint role in the optimization solution process. When the ground capacity and conflict constraints are difficult to fully satisfy, the degree of overcapacity occupancy can be minimized by minimizing the congestion overflow function, thereby reducing the risk of conflict as much as possible while ensuring the feasibility of the system.

[0060] Furthermore, the aforementioned third optimization objective, by summing the congestion overflow functions of all ground units across all time slices, achieves a global quantitative assessment of conflict risk throughout the entire planning time domain. The time dimension in the third optimization objective is reflected in the discrete set of time. Each time slice The iterative summation ensures that conflict risk control is considered at every moment in the planning time domain.

[0061] The three optimization objectives are fused using weights to construct a unified optimization objective function:

[0062] in, This indicates the total number of eVTOLs waiting to glide. This represents the total number of candidate gliding schemes. , , The target weight coefficient.

[0063] Furthermore, the aforementioned target weight coefficients , , All are positive real numbers, used to adjust the relative importance of the three optimization objectives in the unified optimization objective function. In practical applications, when throughput is prioritized, a larger value should be set. Value; when prioritizing conflict risk control, a larger value should be set. Value; when prioritizing operational efficiency, a larger value should be set. value.

[0064] Under the premise of satisfying the constraints of ground capacity and conflict, as well as the constraints of scheme selection, a mixed-integer linear programming method is used to solve the unified optimization objective function to obtain the optimal decision variables. This allows for the determination of the optimal taxiing scheme for each eVTOL.

[0065] It should be noted that the input of the mixed-integer linear programming method is a unified optimization objective function, ground capacity and conflict constraints, and scheme selection constraints, and the output is the optimal decision variables. .

[0066] In this embodiment, when the number of eVTOLs is large or the number of candidate taxiing schemes is large, resulting in excessively long solution time, a heuristic search method can be used to solve the problem. Specifically, the eVTOLs are first sorted according to the urgency of the taxiing task. Then, for each eVTOL, a candidate taxiing scheme that does not conflict with the selected schemes and has the optimal unified optimization objective function value is selected, until all eVTOLs have completed the scheme selection or there are no feasible schemes available.

[0067] Furthermore, the urgency of the aforementioned taxiing tasks is determined based on the planned takeoff or arrival time of the eVTOL. For eVTOLs awaiting takeoff, the earlier the planned takeoff time, the higher the urgency; for eVTOLs taxiing after landing, the earlier the actual arrival time, the higher the urgency. By prioritizing tasks by urgency, taxiing resources can be allocated to eVTOLs with tighter time constraints, reducing the risk of delays.

[0068] Step 5: Generate and issue taxiing control commands.

[0069] Based on the optimal taxiing scheme of each eVTOL, taxiing control commands are generated and sent to the airborne controller of the corresponding eVTOL.

[0070] Taxi control commands include: specifying a taxi time window The optimal gliding path is assigned, and the time constraints for passing through each key node are determined.

[0071] After receiving the taxiing control command, the eVTOL's onboard controller controls the motor speed and direction, and enters the corresponding taxiway section at the specified time. When it arrives at the intersection or landing pad, it performs waiting or release operations according to the constraint time.

[0072] It should be noted that the above-mentioned specified coasting time window Extracted from the spatiotemporal occupancy sequence of the optimal gliding scheme, It is the minimum time value in the spatiotemporal occupancy sequence. This represents the maximum time value in the spatiotemporal occupancy sequence.

[0073] It should be noted that the above-mentioned optimal taxiing path is extracted from the spatiotemporal occupancy sequence of the optimal taxiing scheme. The ground units in the spatiotemporal occupancy sequence are arranged in chronological order to obtain an ordered sequence of ground units from the starting position to the target position, which is the optimal taxiing path.

[0074] It should be noted that the constraint time of each key node is extracted from the spatiotemporal occupancy sequence of the optimal taxiing scheme. Key nodes include taxiing intersection nodes and take-off and landing pad nodes. The time value corresponding to each key node is found from the spatiotemporal occupancy sequence and used as the constraint time for passing through the key node.

[0075] Furthermore, the process of the airborne controller executing taxiing control commands includes two control stages: path tracking and time synchronization. The path tracking stage obtains the current position of the eVTOL using the airborne GPS or UWB positioning system based on the ground unit sequence in the optimal taxiing path, calculates the lateral deviation from the target path, and adjusts the front wheel steering angle or differential motor speed to achieve path following. The time synchronization stage calculates the deviation between the expected arrival time of the eVTOL at each key node and the constraint time, based on the constraint time for passing each key node. When the expected arrival time is earlier than anticipated, the taxiing speed is reduced or the eVTOL waits briefly in a safe area; when the expected arrival time is later than anticipated, the taxiing speed is appropriately increased within a safe range to ensure that the eVTOL passes through the key nodes according to the constraint time.

[0076] In this embodiment, to improve the robustness of the system in dynamic environments, a rolling replanning step is also included. When a dynamic environment change event is detected, the rolling replanning process is triggered, steps 2 to 5 are re-executed, and new coasting control commands are generated and issued.

[0077] The aforementioned dynamic environmental change events include: a new eVTOL joining the taxi queue, a taxiing delay of an eVTOL, a malfunction and withdrawal of an eVTOL, and changes in the status of ground facilities.

[0078] The specific process of the above rolling replanning steps is as follows: First, obtain the actual position and status information of each eVTOL at the current time; then, remove the eVTOLs that have completed the taxiing task from the set to be scheduled, and add the newly added eVTOLs to the set to be scheduled; next, update the starting position node of each eVTOL to the node corresponding to the current actual position; finally, re-execute the steps of generating candidate taxiing scheme set, constructing ground capacity and conflict constraints, solving the multi-objective cooperative optimization model, and issuing taxiing control commands.

[0079] In this embodiment of the application, to further reduce computational complexity, the rolling replanning step adopts an incremental optimization method. Specifically, for eVTOLs unaffected by dynamic environmental change events, their original optimal taxiing scheme remains unchanged; only for affected eVTOLs, a new set of candidate taxiing schemes is generated and optimized; during the optimization process, the spatiotemporal occupancy of unaffected eVTOLs is used as a known constraint, reducing the number of decision variables that need to be optimized.

[0080] Furthermore, the method for determining the affected eVTOLs is as follows: for eVTOLs experiencing taxiing delays or malfunctions, they are marked as affected; for eVTOLs newly added to the taxiing queue, they are also marked as affected; for other eVTOLs, it is checked whether their original optimal taxiing scheme's spatiotemporal occupancy sequence overlaps with the original spatiotemporal occupancy sequence of the affected eVTOLs. If there is an overlap, they are marked as affected; otherwise, they are marked as unaffected. This method for determining affected eVTOLs accurately identifies the range of eVTOLs that need to be replanned, avoiding unnecessary replanning of all eVTOLs.

[0081] This implementation method constructs a ground spatiotemporal topology model to uniformly represent the spatial location and time dimension of ground sliding, transforming the ground sliding conflict problem into an overlapping occupation problem of the same ground unit at the same time, thus providing a unified mathematical basis for subsequent conflict detection and constraint modeling.

[0082] This implementation transforms the multi-aircraft cooperative taxiing scheduling problem into a combinatorial optimization problem by generating a set of candidate taxiing schemes for each eVTOL and defining binary decision variables. This enables the scheduling algorithm to make unified decisions on the taxiing schemes of each eVTOL from a global perspective, overcoming the limitation of existing technologies that can only perform local optimization on a single machine.

[0083] This implementation method incorporates conflict detection into the pre-planning stage by constructing ground capacity and conflict constraints. During the optimization process, it ensures that the number of any ground unit occupied at any time does not exceed its capacity limit, thereby eliminating potential conflict risks when the optimal taxiing scheme is determined. This overcomes the problem of high conflict risk caused by existing technologies relying on online obstacle avoidance.

[0084] This implementation eliminates the influence of parameters with different dimensions in the optimization calculation by normalizing the coasting time and power consumption in the preprocessing. By constructing a multi-objective collaborative optimization model that includes throughput, coasting time, energy consumption and conflict risk, multiple optimization objectives are integrated into a unified optimization objective function through weights. This enables the optimization solution process to simultaneously take into account operating efficiency and resource security, and overcomes the problem of other performance indicators deterioration that may be caused by single-objective optimization.

[0085] This implementation introduces a rolling replanning mechanism, which triggers the rolling replanning process when dynamic environmental change events are detected. This enables the system to perform online replanning under disturbances such as dynamic addition, delay, or failure of eVTOL, overcoming the problem that static planning cannot adapt to dynamic environmental changes.

[0086] Based on the aforementioned implementation method, it can be understood that the executing entity of the airport ground taxiway scheduling method is a computing device with data processing capabilities. The computing device pre-stores the topology data of the airport taxiway network, including the location information and connection relationships of each network node. Therefore, the method also includes the following implementation steps: Step 100: Calculate the load density value of the network cell.

[0087] Obtain aircraft distribution data for each network node and road segment in the current time period, as well as short-term traffic forecast data for the future. Calculate the load density value of each network unit based on the aircraft distribution data and short-term traffic forecast data for the future.

[0088] It should be noted that a network unit is the basic topological element in a taxiway network, including network nodes representing taxiway intersections or critical locations, and road segments representing connection paths between adjacent network nodes. Load density is a quantitative indicator characterizing the current and expected load on a network unit.

[0089] It should be noted that the load density value is calculated as follows: for each network node, the number of aircraft currently located on the network node and its adjacent road segments is counted as the current load, and the number of aircraft expected to pass through the network node within a preset time window in the future is obtained as the expected load. The weighted sum of the current load and the expected load is used as the load density value of the network node. For each road segment, the average of the load density values ​​of the network nodes at both ends of the road segment is used as the load density value of the road segment.

[0090] Furthermore, the preset time window has a duration of 5 to 15 minutes, which can cover the typical time span for an aircraft to taxi from its current position to the target node, ensuring that the statistics of the expected load are of practical significance.

[0091] Furthermore, in the weighted sum of the current load and the expected load, the weight of the current load ranges from 0.6 to 0.8, and the weight of the expected load is 1 minus the weight of the current load. The sum of the two weights is 1. The configuration of the two weights makes the load density value focus more on the current actual carrying capacity, while also taking into account future traffic trends.

[0092] Step 200: Perform region partitioning and assign version identifiers.

[0093] Based on the spatial distribution of load density values, a weighted graph segmentation algorithm is applied to calculate the sub-region partitioning scheme. The calculated sub-region partitioning scheme is compared with the current partitioning. If the difference exceeds the re-partitioning threshold, the region boundary adjustment is triggered. A unique version identifier is assigned to the new partitioning scheme, and the boundary definition and the set of network nodes contained in each sub-region under the version identifier are recorded.

[0094] It should be noted that the weighted graph partitioning algorithm is a conventional graph partitioning algorithm. The input of the weighted graph partitioning algorithm is the taxiway network topology and the load density value of each network node, and the output of the weighted graph partitioning algorithm is the sub-region partitioning scheme of the taxiway network.

[0095] Furthermore, the execution process of the weighted graph segmentation algorithm is as follows: the siding network is represented as an undirected graph, network nodes are the vertices of the undirected graph, road segments are the edges of the undirected graph, and the weight of each vertex is the load density value of the corresponding network node; the weighted graph segmentation algorithm divides the vertex set of the undirected graph into several subsets through iterative optimization, so that the sum of the vertex weights in each subset tends to be balanced, while minimizing the size of the edge cut set between subsets; each subset corresponds to a sub-region, and the vertices in the subset are the network nodes contained in the sub-region.

[0096] It should be noted that the degree of difference is measured as follows: the proportion of network nodes whose regional affiliation has changed between the new partitioning scheme and the current partitioning is calculated out of the total number of nodes, and this proportion is used as the degree of difference value; the repartition threshold is a preset proportion threshold, and when the degree of difference value exceeds the repartition threshold, it is determined that regional boundary adjustment needs to be performed.

[0097] Furthermore, the re-partitioning threshold ranges from 0.15 to 0.25. When the difference value is lower than the re-partitioning threshold, the current partitioning remains unchanged to avoid frequent region adjustments due to minor changes. When the difference value exceeds the re-partitioning threshold, region boundary adjustments are performed to ensure that the sub-region partitioning scheme can respond promptly to significant changes in load distribution.

[0098] It should be noted that the version identifier is a unique code that distinguishes the regional division schemes of different periods, and is generated by a combination of timestamp and serial number; the boundary definition includes the set of boundary nodes of each sub-region and the set of boundary edges between adjacent sub-regions.

[0099] Furthermore, the version identifier is generated as follows: obtain the timestamp of the moment when the boundary adjustment of the triggering area is performed, and represent it in the format of year, month, day, hour, minute, and second; maintain a globally incrementing sequence number counter, and increment the sequence number counter by 1 each time a version identifier is generated; connect the timestamp and the sequence number with an underscore to form a unique version identifier string. The generation method of the version identifier ensures that even if multiple adjustments are triggered at the same time, each version identifier remains unique.

[0100] Furthermore, the boundary node set is the set of network nodes within a sub-region that are directly connected to adjacent sub-regions by road segments; the boundary edge set is the set of road segments that connect the boundary nodes of adjacent sub-regions, and the boundary edges in the boundary edge set need to satisfy cross-regional coordination constraints when solving in parallel.

[0101] In this embodiment of the application, in order to avoid the impact of frequent changes in the region division scheme on the system stability, a lag judgment is introduced when the region boundary adjustment is triggered: when the degree of difference first exceeds the re-division threshold, it is marked as a pre-adjustment state; if the degree of difference continues to exceed the re-division threshold within a preset number of calculation cycles, the region boundary adjustment is confirmed to be executed; if the degree of difference falls back to below the re-division threshold during the pre-adjustment state, the pre-adjustment state is canceled and the current division remains unchanged.

[0102] Furthermore, the preset quantity ranges from 3 to 5 calculation cycles. The preset quantity setting ensures that only continuous load changes can trigger the regional adjustment, filtering out false triggers caused by instantaneous fluctuations.

[0103] Step 300: Construct a node ownership mapping table and perform state migration.

[0104] Calculate the node affiliation mapping table between the old and new regional division versions, and record the correspondence between each network node from the old version regional code to the new version regional code in the node affiliation mapping table; identify network nodes whose regional affiliation has changed based on the node affiliation mapping table, perform state migration on aircraft located on network nodes whose regional affiliation has changed, and transfer the aircraft's scheduling information from the atomic region data structure to the new sub-region data structure.

[0105] It should be noted that the node affiliation mapping table is a data structure that records the regional codes to which the same network node belongs under different regional versions. The key of the node affiliation mapping table is the network node identifier, and the value of the node affiliation mapping table contains mapping pairs of old version regional codes and new version regional codes.

[0106] Furthermore, the node affiliation mapping table is constructed as follows: traverse all network nodes in the taxiway network, for each network node, query the sub-region code to which the network node belongs from the current partitioning scheme as the old version region code, and query the sub-region code to which the network node belongs from the new partitioning scheme as the new version region code; use the network node identifier as the key and the mapping pair consisting of the old version region code and the new version region code as the value, and write it into the node affiliation mapping table.

[0107] It should be noted that the scheduling information includes the aircraft's current position, target position, planned path, estimated arrival time, and interval constraints with other aircraft; the state transition is performed as follows: read the aircraft's scheduling information from the atomic region data structure, write the scheduling information into a new sub-region data structure, and delete the scheduling information entry from the atomic region data structure.

[0108] In this embodiment of the application, in order to support cache retrieval across multiple historical versions, the node attribution mapping table adopts a linked storage structure: for each network node, the regional code sequence of the network node in the most recent preset number of regional versions is stored to form a version chain; when it is necessary to query the regional code change of a network node from any historical version to the current version, the complete mapping path can be obtained by tracing along the version chain.

[0109] Furthermore, the number of recently preset regional versions ranges from 5 to 10 versions. This setting ensures cross-version retrieval coverage while controlling the storage overhead and query complexity of the node affiliation mapping table.

[0110] Furthermore, the version chain is stored as follows: for each network node, an ordered list is maintained. Each element in the ordered list contains a version identifier and the region code of the version corresponding to the version identifier. The ordered list is sorted from newest to oldest according to the version generation time. When the version chain length exceeds a preset number, the oldest version record at the end of the ordered list is deleted to keep the version chain length from exceeding the preset number.

[0111] Furthermore, the tracing method of the mapping path is as follows: given the starting historical version and the target current version, locate the region code corresponding to the starting historical version from the version chain of the network node as the starting code, and locate the region code corresponding to the target current version as the ending code; if the starting code and the ending code are the same, the region affiliation of the network node between the starting historical version and the target current version remains unchanged; if the starting code and the ending code are different, record the mapping relationship of the network node from the starting code to the ending code.

[0112] Step 400: Generate the feature vector of the scheme.

[0113] For the current detailed task to be processed, obtain the network node identifiers of the aircraft's origin and destination, query the region code to which the origin and destination network nodes belong based on the current region version, and generate the scheme feature vector based on the region code.

[0114] It should be noted that the scheme feature vector is a multi-dimensional vector used for cache retrieval, including four dimensions: starting area code, ending area code, coarse-grained time period code, and congestion status feature code. The coarse-grained time period code is the code of the time period to which the current moment belongs after dividing the day into several time periods. The congestion status feature code is the coded value that represents the current level of congestion in the starting and ending point path areas.

[0115] It should be noted that the congestion status feature code is calculated as follows: obtain the load density value of each network unit in the starting region and the ending region, calculate the average value of the load density value in the starting region and the ending region, determine the congestion level according to the preset interval in which the average value is located, and use the code value corresponding to the congestion level as the congestion status feature code.

[0116] Furthermore, the congestion level is divided into three levels: low, medium, and high, with corresponding coding values ​​of 1, 2, and 3, respectively. When the average load density is less than the first threshold, it is classified as low level; when it is between the first and second thresholds, it is classified as medium level; and when it is greater than the second threshold, it is classified as high level. The first threshold ranges from 0.3 to 0.4, and the second threshold ranges from 0.6 to 0.7. This method of classifying the congestion level discretizes the continuous load density value into a finite number of levels, which facilitates feature matching during cache retrieval.

[0117] Step 500: Perform hierarchical cache retrieval.

[0118] When performing a retrieval operation in the hierarchical cache, firstly, search for cache entries in the cache partition corresponding to the current region version whose distance from the scheme feature vector is less than the matching threshold; if there is no matching cache entry in the cache partition corresponding to the current region version, then use the node affiliation mapping table to reverse the regional code in the scheme feature vector to the code under the historical region version, and search for matching cache entries in the cache partition corresponding to the historical region version.

[0119] It should be noted that the hierarchical cache is a multi-layered cache structure organized by region version. Each layer of the hierarchical cache corresponds to a region version, and each layer of the hierarchical cache stores the refined scheme generated under the region version. The feature vector distance is calculated as follows: for the corresponding dimension encoding values ​​of two feature vectors, if the dimension is a categorical encoding, the difference is 0 when the encoding values ​​are the same and 1 when they are different. If the dimension is a numerical encoding, the difference is the absolute value of the difference between the two encoding values. The weighted sum of the differences of each dimension is used as the feature vector distance.

[0120] Furthermore, the matching threshold ranges from 0.5 to 1.0. When the feature vector distance is less than the matching threshold, the cached entry is determined to match the current task. The matching threshold setting ensures matching accuracy while allowing cached entries with certain differences in features to be reused, thereby improving the cache hit rate.

[0121] It should be noted that the reverse transformation is performed as follows: query the regional codes of the starting network node and the ending network node in the target historical version from the node affiliation mapping table, replace the current regional code in the scheme feature vector with the historical regional code obtained from the query, and generate a transformation feature vector for historical version retrieval.

[0122] Furthermore, the method for determining the target historical version is as follows: obtain the version chain of each network node from the node affiliation mapping table, extract all historical version identifiers except the current version from the version chain, sort them from newest to oldest according to the version generation time, and use them as the target historical versions for reverse conversion and retrieval.

[0123] Furthermore, the query method for historical region codes is as follows: given the target historical version and network node identifier, obtain the version chain of the network node from the node affiliation mapping table, search for the region code record corresponding to the target historical version in the version chain, and extract the region code value in the region code record as the historical region code.

[0124] In this embodiment of the application, in order to improve the efficiency of cross-version retrieval, a priority sorting strategy is adopted when performing historical version retrieval: the version similarity is calculated based on the degree of change in node affiliation between each historical version and the current version, and the retrieval is performed in the cache partition of each historical version in descending order of version similarity. When a matching cache entry is found, the retrieval process of subsequent versions is terminated.

[0125] Furthermore, the version similarity is calculated as follows: the proportion of network nodes whose regional affiliation remains unchanged between the historical version and the current version is used as the version similarity value. The higher the version similarity value, the closer the regional division of the two versions is, and the greater the probability that the cached entries under the historical version will still be valid after conversion.

[0126] Step 600: Perform feasibility verification.

[0127] Perform feasibility verification on cached entries that are retrieved, and check whether the solutions in the cached entries are feasible under the current constraints. For cached entries retrieved across versions, additional boundary coupling constraint compatibility checks are performed to determine whether the start and end points of the solutions in the cached entries still belong to the same region or still cross the same region boundary under the current regional version.

[0128] It should be noted that the current constraints include aircraft spacing constraints, taxiway capacity constraints, and time window constraints. The feasibility verification is performed as follows: obtain the estimated transit time of each path node in the caching scheme, check whether the constraints at each time point are met, and if all constraints are met, the caching scheme is deemed feasible; otherwise, the caching scheme is deemed infeasible.

[0129] Furthermore, the aircraft spacing constraint is the minimum time interval that adjacent aircraft must maintain on the same road segment or network node. The verification method for the aircraft spacing constraint is as follows: for each path node in the caching scheme, obtain the estimated passage time of the path node, query the time of other aircraft passing through the path node in the same time period, calculate the time difference, and if the time difference is less than the preset minimum interval time, it is determined that the aircraft spacing constraint is not satisfied.

[0130] Furthermore, the taxiway capacity constraint is the maximum number of aircraft allowed to pass through each segment at the same time. The verification method for the taxiway capacity constraint is as follows: for each path segment in the caching scheme, obtain the time period occupied by each aircraft on the path segment, count the number of aircraft existing on the path segment at each time, and if the number of aircraft exceeds the segment capacity limit, it is determined that the taxiway capacity constraint is not met.

[0131] Furthermore, the time window constraint is the constraint that the aircraft must reach the target location within a specified time range. The verification method for the time window constraint is as follows: obtain the estimated time for the aircraft to reach the destination in the caching scheme, and compare it with the target arrival time window of the aircraft. If the estimated arrival time is earlier than the start time of the time window or later than the end time of the time window, it is determined that the time window constraint is not satisfied.

[0132] It should be noted that boundary coupling constraints describe the coordination constraints that aircraft must follow when crossing regional boundaries in a caching scheme, including the passage time constraints of boundary nodes and the interval constraints with aircraft in adjacent regions. The execution method of boundary coupling constraint compatibility check is as follows: query the regional affiliation of the start and end points of the caching scheme according to the current regional version, compare it with the regional affiliation when the caching scheme was generated, and if the same region or cross-regional relationship of the start and end points changes, it is determined that the boundary coupling constraints are incompatible.

[0133] Furthermore, the same-region or cross-regional relationship between the origin and destination includes three cases: the origin and destination are located in the same sub-region, the origin and destination are located in adjacent sub-regions, and the origin and destination are located in non-adjacent sub-regions. The boundary coupling constraint compatibility check verifies whether the three relationship types are consistent with the current version when the caching scheme is generated. If the relationship type changes, the domain and constraint form of the boundary coupling constraint may become invalid.

[0134] Furthermore, the method for verifying the consistency of relation types is as follows: read the region codes of the start and end points when the cache scheme was generated from the metadata of the cache entry, and determine whether the two region codes are the same to determine the relation type at the time of generation; query the region codes of the start and end points from the current region version, and determine whether the two region codes are the same to determine the current relation type; compare the relation type at the time of generation with the current relation type, and if the relation types are the same, determine that the relation is consistent; if the relation types are different, determine that the relation has changed.

[0135] In this embodiment of the application, in order to utilize cache entries that are incompatible with boundary coupling constraints but whose main body of the caching scheme still has reference value, a scheme repair attempt is performed when the boundary coupling constraint compatibility check fails: the path segments in the caching scheme affected by boundary changes are identified, local replanning is performed on the path segments affected by boundary changes, and if the local replanning is successful and the replanned caching scheme passes the feasibility verification, the repaired caching scheme is used as the retrieval result.

[0136] Furthermore, the path segments affected by boundary changes are those that cross regional boundaries in the caching scheme. When the regional boundary crossed by a path segment changes location or no longer exists in the current version, the path segment needs to perform local replanning to adapt to the new boundary definition.

[0137] Furthermore, the method for identifying path segments affected by boundary changes is as follows: traverse the sequence of path nodes in the caching scheme, and for two adjacent path nodes, query the region affiliation of the two path nodes in the caching scheme generation version and the current version; if the two path nodes belong to different regions in the caching scheme generation version but belong to the same region in the current version, or if the two path nodes belong to the same region in the caching scheme generation version but belong to different regions in the current version, then the path segment connecting the two path nodes is a path segment affected by boundary changes.

[0138] Furthermore, the execution method of local replanning is as follows: for each path segment affected by boundary changes, the starting point and ending point of the path segment affected by boundary changes are used as the starting and ending points of local planning. Under the topology and constraints of the current region version, a path planning algorithm is applied to calculate new path segments. The newly calculated path segments replace the path segments affected by boundary changes in the original caching scheme, while keeping other path segments unchanged, to form a repaired caching scheme.

[0139] Step 700: Output the refined results and update the cache.

[0140] If the feasibility verification passes, the caching scheme is adopted as the refined result, and the regional code of the caching scheme is updated to the current version; if the feasibility verification fails or no cache entry is hit, a complete refinement optimization solution is performed to generate a new scheme, and the new scheme is written to the hierarchical cache with the code of the current regional version.

[0141] It should be noted that the refinement optimization solution is an optimization process that refines the coarse-grained layer scheme in time and space. The input of the refinement optimization solution is the aircraft's origin and destination information and the path skeleton of the coarse-grained layer scheme. The output of the refinement optimization solution is a refined scheme that includes the specific path node sequence and the passage time of each path node.

[0142] Furthermore, the detailed optimization process is as follows: Based on the path skeleton provided by the coarse-grained layer scheme, the approximate taxiing path of the aircraft is determined. Under the constraints of the path skeleton, an optimization model is constructed that includes path selection variables and time scheduling variables. The objective function of the optimization model is... ,in The total number of aircraft to be dispatched. For aircraft indexing, For the first Total taxiing time of the aircraft For the first The cost of path conflict for aircraft As time weight, The conflict weights are used; the constraints of the optimization model include path continuity constraints, aircraft spacing constraints, taxiway capacity constraints, and time window constraints; the mixed integer programming solver is called to solve the optimization model to obtain the optimal path node sequence and the passage time of each path node as a refinement scheme.

[0143] Furthermore, time weighting The value ranges from 0.3 to 0.5, and the conflict weight is... The value ranges from 0.5 to 0.7, and the sum of the time weight and the conflict weight is 1. The configuration of the time weight and the conflict weight enables the optimization process to prioritize avoiding path conflicts while reducing the total gliding time.

[0144] Furthermore, total gliding time For the first The actual taxiing time of an aircraft from the origin to the destination is calculated as the time taken to reach the destination node minus the departure time from the origin node; path conflict cost. For the first The penalty value for when an aircraft overlaps with another aircraft in time and space on a path node or segment. The path conflict cost is generated when two aircraft occupy the same network unit at the same time.

[0145] It should be noted that the execution method for updating the region code is as follows: the starting region code and the ending region code in the caching scheme are replaced with the corresponding region code under the current version, while keeping other contents of the caching scheme unchanged; the updated caching scheme creates a new index entry in the current version's cache partition.

[0146] Furthermore, the new scheme is written to the hierarchical cache as follows: extract the starting area code, ending area code, coarse-grained time period code, and congestion status feature code of the new scheme, and construct the scheme feature vector as the cache index key; use the path node sequence of the new scheme, the passage time of each path node, and the regional version identifier at the time of generation as the content of the cache entry; in the cache partition corresponding to the current regional version, write the cache entry with the scheme feature vector as the key and the content of the cache entry as the value.

[0147] Step 800: Solve the scheduling subproblem in parallel and output the scheduling scheme.

[0148] For each sub-region, a local scheduling sub-problem is constructed based on the refined results. The alternating direction multiplier method is then used to solve the local scheduling sub-problem of each sub-region in parallel. Each sub-region performs optimization calculations independently under boundary constraints. After the solution results of each sub-region converge, the local scheduling schemes of each sub-region are integrated to output the overall scheduling scheme.

[0149] It should be noted that the local scheduling subproblem is an optimization problem that takes the refined paths of aircraft within a sub-region as input and aims to minimize conflicts and delays within the sub-region.

[0150] It should be noted that the alternating direction multiplier method is a conventional distributed optimization algorithm. The input of the alternating direction multiplier method is the local scheduling subproblem and boundary constraints of each sub-region, and the output of the alternating direction multiplier method is the local scheduling scheme of each sub-region. The alternating direction multiplier method updates the local solution and boundary Lagrange multipliers of each sub-region iteratively, so that the boundary variables of each sub-region gradually satisfy the global consistency constraint.

[0151] Furthermore, the objective function for the local scheduling subproblem is: ,in The number of aircraft in the sub-region. For aircraft indexing within a sub-region, For the first The cost of a conflict involving aircraft For the first The delay time of the aircraft As the conflict cost weight, To delay the weighting.

[0152] Furthermore, conflict cost weighting The value ranges from 0.6 to 0.8, and the delay weight is... The value of is 1 minus the conflict cost weight. The sum of the conflict cost weight and the delay weight is 1. The configuration of the conflict cost weight and the delay weight makes the optimization process prioritize eliminating conflicts between aircraft, and then minimize the delay time under the premise of no conflict.

[0153] Furthermore, the cost of conflict For the first The cumulative penalty value for spatiotemporal overlap between the aircraft and other aircraft in the same sub-region at path nodes or road segments; delay time. For the first The difference between the actual arrival time and the planned arrival time of an aircraft is the time difference between the actual arrival time and the planned arrival time. When the actual arrival time is later than the planned arrival time, a delay occurs.

[0154] It should be noted that the boundary constraints include the consistency of aircraft passage time and the aircraft interval constraint at the boundary nodes of adjacent sub-regions; the convergence of the solution is determined by calculating the change of boundary variables of each sub-region between adjacent iterations, and the solution is considered to have converged when the change of all boundary variables is less than the convergence threshold.

[0155] Furthermore, the convergence threshold ranges from 0.01 to 0.05 seconds. When the change in the boundary variable is less than the convergence threshold, it indicates that the boundary state of each sub-region has become stable, and continuing the iteration will have limited improvement on the results. At this point, the iteration process is terminated to balance the solution accuracy and computational efficiency.

[0156] Furthermore, the method for integrating the local scheduling schemes of each sub-region is as follows: collect the aircraft path node sequence and passage time output by each sub-region, and merge them according to the aircraft identifier; for aircraft that span multiple sub-regions, connect the path segments of the aircraft in each sub-region in chronological order to form a complete taxiway path; summarize the complete taxiway paths and time scheduling of all aircraft to generate an overall scheduling scheme covering the entire taxiway network.

[0157] In this embodiment of the application, in order to periodically maintain cache quality and evaluate the rationality of region division, the following steps are also included: based on cache usage statistics, identify cache entries whose access frequency is lower than a preset frequency threshold or whose generation time is earlier than a preset time threshold, and perform cleanup operations on the identified low-frequency and outdated cache entries; periodically obtain load distribution data of each sub-region, evaluate the changing trend of load balancing, and decide the triggering time of the next round of region adjustment based on the evaluation results.

[0158] Furthermore, the preset frequency threshold ranges from 2 to 5 accesses per hour. When the access frequency of a cache entry is lower than the preset frequency threshold, it is determined to be a low-frequency cache entry. The preset time threshold ranges from 24 to 48 hours after its generation. When the generation time of a cache entry is earlier than the preset time threshold, it is determined to be an outdated cache entry. Cleaning up low-frequency and outdated cache entries releases cache space and provides storage resources for high-value cache entries.

[0159] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. An aircraft taxiing scheduling method based on multi-objective optimization, characterized in that, Includes the following steps: Obtain ground facility information of the vertical take-off and landing site, and construct a ground spatiotemporal topology model based on the ground facility information. The ground spatiotemporal topology model includes a ground topology network, a time discrete set, and a set of ground spatiotemporal occupancy units. The taxiing mission information of multiple eVTOLs currently waiting to taxi is obtained. Based on the ground spatiotemporal topology model and taxiing mission information, a candidate taxiing scheme set is generated for each eVTOL. Each candidate taxiing scheme contains the spatiotemporal occupancy sequence of the corresponding eVTOL. In the process of generating the candidate taxiing scheme set, the generation of the second shortest path is implemented using the K-shortest path algorithm. After obtaining the shortest path, each taxiway segment in the shortest path is deleted in turn. Dijkstra's algorithm is re-run on the ground topology network after each deletion to obtain candidate paths. All candidate paths are sorted according to their path length, and several paths with the second shortest length are selected as the second shortest paths. Based on the capacity parameters in the aforementioned ground spatiotemporal topology model and the spatiotemporal occupancy sequences of each candidate taxiing scheme, ground capacity and conflict constraints are constructed. These constraints limit the number of ground units occupied at any given time to their capacity limit. The construction of these constraints includes: defining an occupancy indicator function to indicate whether a certain eVTOL occupies a certain ground unit in a certain time slice under a certain candidate taxiing scheme; and based on the occupancy indicator function and the decision variables of each candidate taxiing scheme, ground capacity and conflict constraints are constructed, indicating that the weighted sum of the number of times all eVTOLs occupy the same ground unit in the same time slice under all candidate taxiing schemes does not exceed the capacity limit of that ground unit. Based on the ground capacity and conflict constraints of the candidate taxiing scheme set, a multi-objective collaborative optimization model is constructed to solve for the optimal decision variables and obtain the optimal taxiing scheme for each eVTOL. Under the premise of satisfying ground capacity and conflict constraints and scheme selection constraints, a mixed integer linear programming method is used to solve the unified optimization objective function to obtain the optimal decision variables, and then determine the optimal taxiing scheme for each eVTOL. When the number of eVTOLs is large or the number of candidate taxiing schemes is large, resulting in excessive solution time, a heuristic search method is used to solve the problem. First, the eVTOLs are sorted according to the urgency of the taxiing task. Then, for each eVTOL, a candidate taxiing scheme that does not conflict with the selected scheme and has the best unified optimization objective function value is selected in turn, until all eVTOLs have completed the scheme selection or there is no feasible scheme to choose from. The multi-objective collaborative optimization model includes three optimization objectives: The primary optimization objective is to maximize the throughput of the ground process, which means maximizing the sum of the number of scheduled eVTOLs. The second optimization objective is to minimize coasting time and energy consumption, which means minimizing the weighted sum of the normalized coasting time and normalized energy consumption of each eVTOL; The third optimization objective is to minimize the risk of conflict. This is achieved by defining a congestion overflow function to represent the amount of overcapacity occupied by each ground unit in each time slice. The congestion overflow function is the larger of the difference between the actual occupied amount and the capacity limit and zero. The third optimization objective is to minimize the sum of the congestion overflow functions of all ground units in all time slices. The three optimization objectives are fused using weights to construct a unified optimization objective function; the unified optimization objective function is as follows: Where K represents the total number of eVTOLs to be taxied, and M represents the total number of candidate taxiing schemes. , , The target weight coefficient, Indicates whether the k-th eVTOL has been scheduled. For binary decision variables, and These are the weighting coefficients. This is the normalized glide time. Normalized energy consumption For congestion overflow function; For each candidate taxiing scheme of each eVTOL, a binary decision variable is defined. When the binary decision variable is 1, it means that the corresponding candidate taxiing scheme is selected. When the binary decision variable is 0, it means that the corresponding candidate taxiing scheme is not selected. For each eVTOL, a scheme selection constraint is set, which means that each eVTOL can select at most one taxiing scheme. Based on the optimal taxiing scheme of each eVTOL, taxiing control commands are generated and sent to the airborne controller of the corresponding eVTOL.

2. The aircraft taxiing scheduling method based on multi-objective optimization according to claim 1, characterized in that, The ground topology network includes a set of ground nodes and a set of taxiway segments. The set of ground nodes includes parking position nodes, FATO landing pad nodes, charging position nodes, boarding gate nodes, and taxiway intersection nodes. Each element in the set of taxiway segments represents a taxiway between two adjacent ground nodes. The discrete time set includes multiple discrete time slices; The set of ground spatiotemporal occupancy units is used to represent the occupancy status of each ground unit in each time slice; The ground spatiotemporal topology model also includes capacity parameters for each ground unit, which represent the upper limit of the number of eVTOLs that a ground unit is allowed to occupy simultaneously at the same time.

3. The aircraft taxiing scheduling method based on multi-objective optimization according to claim 1, characterized in that, The process of generating a set of candidate taxiing schemes for each eVTOL includes: For each eVTOL, based on the starting position node and the target position node in its taxiing mission information, the shortest path algorithm is used to calculate the shortest path from the starting position node to the target position node and several shortest paths as candidate paths. By combining the taxiing speed parameters of each eVTOL, the spatiotemporal occupancy sequence corresponding to each candidate path is calculated, and a set of candidate taxiing schemes is generated.

4. The aircraft taxiing scheduling method based on multi-objective optimization according to claim 3, characterized in that, The calculation process of the spatiotemporal occupancy sequence is as follows: For each ground cell in the candidate path, calculate the time required to pass through that ground cell based on the eVTOL's taxiing speed and the physical length of the ground cell; The time interval during which the eVTOL occupies the ground cell is determined based on the time when the eVTOL enters the ground cell and the time required for its passage. All discrete time slices within a time interval and their corresponding ground units are incorporated into the spatiotemporal occupancy sequence.

5. The aircraft taxiing scheduling method based on multi-objective optimization according to claim 1, characterized in that, Before constructing the multi-objective collaborative optimization model, the following is also included: The mean normalization method based on the range is adopted to normalize the gliding time and power consumption of each candidate gliding scheme to a preset range, so as to obtain the normalized gliding time and normalized power consumption. The coasting time is calculated by summing the ratios of the physical length of each ground unit in the candidate path to the eVTOL coasting speed. The energy consumption is calculated based on the product of the motor power and the coasting time during eVTOL coasting.

6. The aircraft taxiing scheduling method based on multi-objective optimization according to claim 1, characterized in that, The taxiing control commands include: specifying a taxiing time window, allocating an optimal taxiing path, and setting time constraints for passing each key node; The specified taxiing time window is extracted from the spatiotemporal occupancy sequence of the optimal taxiing scheme, and is the interval from the minimum time value to the maximum time value in the spatiotemporal occupancy sequence; The optimal gliding path is obtained by arranging the ground units in the spatiotemporal occupancy sequence in chronological order. The key nodes include taxiway crossing nodes and take-off and landing pad nodes, and the constraint time is obtained by looking up the time value corresponding to each key node in the spatiotemporal occupancy sequence.

7. The aircraft taxiing scheduling method based on multi-objective optimization according to any one of claims 1 to 6, characterized in that, It also includes a rolling replanning step: When a dynamic environmental change event is detected, a rolling replanning process is triggered. The dynamic environmental change events include a new eVTOL joining the taxi queue, a taxiing delay of an eVTOL, a fault exit of an eVTOL, or a change in the status of ground facilities. Obtain the actual location and status information of each eVTOL at the current time, and update the set to be scheduled and the starting position node of each eVTOL; The steps of generating candidate taxiing schemes, constructing ground capacity and conflict constraints, solving the multi-objective collaborative optimization model, and issuing taxiing control commands are repeated.

8. An aircraft taxiing scheduling system based on multi-objective optimization, used to execute the aircraft taxiing scheduling method based on multi-objective optimization as described in any one of claims 1 to 7, characterized in that, include: The ground spatiotemporal topology model construction module is used to obtain ground facility information of the vertical take-off and landing site and construct a ground spatiotemporal topology model based on the ground facility information; The candidate taxiing scheme generation module is used to obtain taxiing mission information of multiple eVTOLs currently waiting to taxi, and generate a set of candidate taxiing schemes for each eVTOL based on the ground spatiotemporal topology model and taxiing mission information; The constraint construction module is used to construct ground capacity and conflict constraints based on the capacity parameters in the ground spatiotemporal topology model and the spatiotemporal occupancy sequence of each candidate taxiing scheme. The optimization solution module is used to construct a multi-objective collaborative optimization model based on the candidate taxiing scheme set and ground capacity and conflict constraints, and solve for the optimal decision variables to obtain the optimal taxiing scheme for each eVTOL. The multi-objective collaborative optimization model includes three optimization objectives: the first optimization objective is to maximize ground process throughput, the second optimization objective is to minimize taxiing time and energy consumption, and the third optimization objective is to minimize conflict risk; a mixed integer linear programming method is used to solve the unified optimization objective function. The instruction generation and distribution module is used to generate taxiing control instructions based on the optimal taxiing scheme of each eVTOL and distribute them to the airborne controller of the corresponding eVTOL.

Images