Scheduling method and system for unmanned ferry vehicles in airport

Abstract

The invention discloses an airport unmanned ferry vehicle scheduling method and system, and relates to the technical field of intelligent traffic, and the method comprises the steps: obtaining flight information, the state information of all unmanned ferry vehicles, and the traffic information of an airport road, and carrying out the fusion to generate an airport operation panoramic dynamic data snapshot; based on the airport operation panoramic dynamic data snapshots, calculating a dynamic passenger travel tightness index of each passenger to be served, and generating a service request queue sorted according to the dynamic passenger travel tightness indexes; the service request queue is a core input to carry out multi-objective optimization calculation, a vehicle scheduling instruction set is generated, and the unmanned ferry vehicle executes the vehicle scheduling instruction set and plans a local path to drive to a passenger boarding point; when the unmanned ferry vehicle is about to arrive, real-time transport capacity and requirements are met. According to the method, the core problem that the priority judgment standard in the prior art is rough and cannot accurately respond to the highly urgent demand is solved, and the optimal distribution of the transport capacity at the critical moment is realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a scheduling method and system for unmanned airport shuttle buses. Background Technology

[0002] With the deepening of smart airport construction, the application of unmanned driving technology in the passenger shuttle process has become a key direction for improving operational efficiency and service quality. The intelligent scheduling method of airport unmanned shuttle buses is one of the core research areas. Under the development trend of smart airports, the intelligent scheduling of unmanned shuttle buses is crucial to improving ground transportation efficiency. Existing technologies are mainly based on static rules or single optimization objectives for vehicle scheduling. These methods have achieved automation to a certain extent, but the precision and dynamic adaptability of the scheduling strategy are still limited.

[0003] The shortcomings of existing technologies are that their scheduling priority judgment criteria are relatively crude, and they cannot accurately quantify and dynamically respond to the urgency of passengers' journeys. The system fails to comprehensively consider the differences in passengers' journey attributes and real-time spatial locations, which makes it impossible to accurately identify and prioritize serving passengers with the most urgent journeys when capacity is tight, thus limiting the further improvement of scheduling efficiency and service quality. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a scheduling method for unmanned airport shuttle buses to solve the problem of inaccurate scheduling caused by the inability to precisely quantify the closeness of passenger journeys.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for scheduling unmanned shuttle buses at airports, which includes acquiring flight information, status information of all unmanned shuttle buses and traffic information of airport roads, and fusing them to generate a snapshot of dynamic panoramic data of airport operations. Based on the panoramic dynamic data snapshot of airport operations, the dynamic passenger trip tracking index of each passenger to be served is calculated, and a service request queue sorted by the dynamic passenger trip tracking index is generated. The service request queue serves as the core input for multi-objective optimization calculations, generating a vehicle dispatch instruction set. The driverless shuttle bus executes the vehicle dispatch instruction set and plans a local route to the passenger boarding point. When the driverless shuttle bus is about to arrive, the service granularity of multiple passenger service requests is dynamically adjusted according to real-time capacity and demand, and a task encapsulation instruction for the shared pick-up point is generated. The system receives task completion information from the driverless shuttle bus and optimizes and updates the dynamic passenger journey tracking index and the strategy model for dynamic service granularity adjustment based on the task completion information.

[0007] As a preferred embodiment of the airport unmanned shuttle bus scheduling method of the present invention, the method includes the following steps: acquiring flight information, status information of all unmanned shuttle buses, and traffic information of airport roads, and fusing them to generate a panoramic dynamic data snapshot of airport operations. By acquiring flight information, autonomous shuttle vehicle status information, and airport road traffic information through data interfaces, and performing timestamp alignment and coordinate transformation, a dynamic association graph is constructed with flights, autonomous shuttle vehicles, and key location points as nodes and dynamic association relationships as edges. The dynamic association graph is input into a pre-trained graph neural network model to update the feature vector of each node, generate an enhanced feature vector set of global association information for flights, unmanned shuttle buses, and key location points, and encapsulate it in a structured manner to output a snapshot of dynamic panoramic data of airport operations.

[0008] As a preferred embodiment of the airport driverless shuttle bus scheduling method of the present invention, the method includes: calculating the dynamic passenger journey tracking index for each passenger to be served based on a snapshot of panoramic dynamic data of airport operations, and generating a service request queue sorted by the dynamic passenger journey tracking index, comprising the following steps: Extract the list of passengers to be served and the associated flight schedules, departure times, flight status, and real-time passenger location information from the airport's panoramic dynamic data snapshot; The flight urgency factor is calculated using flight schedule departure time and flight status; the passenger trip criticality factor is calculated using passenger trip attributes; and the spatial urgency factor is calculated using passenger real-time location. The flight urgency factor, passenger itinerary criticality factor, and spatial urgency factor are input into the fuzzy logic rule set for dynamic weight allocation, and the dynamic passenger itinerary urgency index is calculated. The list of passengers to be served is sorted in descending order based on the dynamic passenger trip tracking index, and a service request queue is generated.

[0009] As a preferred embodiment of the airport unmanned shuttle bus scheduling method of the present invention, wherein: the service request queue is used as the core input for multi-objective optimization calculation to generate a vehicle scheduling instruction set, including the following steps: By combining service request queues and real-time vehicle status information, a vehicle scheduling problem model with task set, vehicle set, and spatiotemporal and resource constraints is constructed. A multi-objective optimization function is defined for the vehicle scheduling problem model with the objectives of maximizing the satisfaction of the pursuit degree, minimizing the total energy consumption, and minimizing the global traffic congestion cost. An improved multi-objective ant colony algorithm is applied to solve the multi-objective optimization function, resulting in a set of Pareto optimal scheduling schemes that represent the trade-offs between different objectives. Based on a predefined decision-making strategy, a final scheduling scheme is selected from the Pareto optimal scheduling scheme set, and the final scheduling scheme is converted into a set of specific vehicle scheduling instructions that can be issued.

[0010] As a preferred embodiment of the airport driverless shuttle bus scheduling method of the present invention, the driverless shuttle bus executes a vehicle scheduling instruction set and plans a local route to the passenger pick-up point, including the following steps: The driverless shuttle bus receives and parses the vehicle dispatch instruction set through the vehicle network communication link to confirm the passenger boarding point location and target time window to be executed; The driverless shuttle bus uses its current location and the passenger boarding point location to call the on-board path planning algorithm, and combines high-precision map and real-time local obstacle information to calculate the optimal local driving path; The control unit of the driverless shuttle bus tracks and executes the optimal local driving path, controlling the vehicle to drive along the optimal local driving path to the passenger boarding point.

[0011] As a preferred embodiment of the airport driverless shuttle bus scheduling method of the present invention, the method includes the following steps: when the driverless shuttle bus is about to arrive, the service granularity of multiple passenger service requests is dynamically adjusted according to real-time capacity and demand to generate a task encapsulation instruction for a shared pick-up point: By determining that the driverless shuttle bus is about to arrive at the passenger pick-up point based on the vehicle's location information, a dynamic service granularity adjustment process is triggered to obtain the set of passengers waiting to be served in the service request queue and the real-time capacity distribution in the local area. Using the target boarding gate coordinates of each passenger in the set of passengers to be served as the clustering feature, the spatiotemporal clustering algorithm is used to divide passengers going to the same terminal area into different passenger request clusters, and the efficiency gain generated after merging with the current task of the triggering vehicle is calculated for each passenger request cluster. The efficiency gain is equal to the total time saved minus the total detour time caused by carpooling, and passenger request clusters that have an efficiency gain greater than a preset time threshold are selected for merging. Select the passenger request cluster with the highest efficiency gain from the mergeable passenger request clusters, calculate the shared boarding point coordinates that minimize the total walking distance for all passengers in the passenger request cluster with the highest efficiency gain, and generate a task encapsulation instruction that includes the shared boarding point coordinates and passenger request cluster information.

[0012] As a preferred embodiment of the airport unmanned shuttle bus scheduling method of the present invention, the method includes: receiving task completion information from the unmanned shuttle bus, and optimizing and updating the dynamic passenger journey tracking index and the strategy model for dynamic service granularity adjustment based on the task completion information, comprising the following steps: Receive task completion information from the driverless shuttle vehicle, and align the actual pick-up time and actual travel time in the task completion information with the estimated waiting time and estimated travel time recorded during the scheduling decision to form a task experience data record; Based on task experience data records, the waiting time deviation between the actual waiting time and the expected waiting time is calculated to obtain the efficiency gain deviation between the actual efficiency gain and the estimated efficiency gain. Using waiting time deviation as a loss signal, the weight parameters used in the dynamic passenger journey pursuit index are adjusted through an online stochastic gradient descent algorithm. The efficiency gain deviation is used as a reward signal, and the decision parameters of the policy model for dynamically adjusting service granularity are adjusted through a policy gradient algorithm.

[0013] Secondly, the present invention provides a scheduling system for unmanned shuttle buses at airports, including a fusion module that acquires flight information, status information of all unmanned shuttle buses and traffic information of airport roads, and fuses them to generate a dynamic snapshot of airport operation panorama. The evaluation module calculates the dynamic passenger journey tracking index for each passenger to be served based on a snapshot of dynamic data of the airport operation panorama, and generates a service request queue sorted by the dynamic passenger journey tracking index. The multi-objective collaborative scheduling module uses the service request queue as the core input to perform multi-objective optimization calculations, generate a vehicle scheduling instruction set, and the driverless shuttle bus executes the vehicle scheduling instruction set to plan a local route to the passenger boarding point. The adjustment module dynamically adjusts the service granularity of multiple passenger service requests based on real-time capacity and demand when the driverless shuttle bus is about to arrive, and generates task encapsulation instructions for the shared pick-up point. The optimization module receives task completion information from the driverless shuttle bus and optimizes and updates the dynamic passenger journey tracking index and the strategy model for dynamic service granularity adjustment based on the task completion information.

[0014] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the airport unmanned shuttle bus scheduling method as described in the first aspect of the present invention.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the airport driverless shuttle bus scheduling method as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By comprehensively considering multiple factors such as flight urgency, passenger itinerary criticality, and spatial urgency, and using fuzzy logic for dynamic weight fusion, a refined index quantifying passenger urgency is generated. This solves the core problem of existing technologies having coarse priority judgment standards and being unable to accurately respond to high-urgency demands. It achieves optimal allocation of capacity at critical moments. By establishing an online learning closed loop, the parameters of the urgency calculation model and service adjustment strategy are continuously corrected using actual operational data, enabling the system to learn from experience and improve itself. This overcomes the shortcomings of traditional static models, such as poor adaptability and long-term performance degradation. Together, these factors ensure that the scheduling system not only possesses initial decision-making intelligence but also can continuously evolve, ultimately comprehensively improving the efficiency, accuracy, and adaptability of airport shuttle services. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the scheduling method for driverless shuttle buses at airports.

[0019] Figure 2 This is a schematic diagram of the airport's driverless shuttle bus dispatch system.

[0020] Figure 3 This is a schematic diagram for calculating the dynamic passenger journey tracking index.

[0021] Figure 4 Flowchart for generating vehicle dispatch instruction sets for multi-objective optimization. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a method for scheduling unmanned airport shuttle buses, comprising the following steps: S1. Obtain flight information, status information of all driverless shuttle vehicles, and traffic information of airport roads, and integrate them to generate a dynamic snapshot of the airport operation panorama.

[0026] S1.1 Obtain flight information, autonomous shuttle vehicle status information, and airport road traffic information through data interfaces, and perform timestamp alignment and coordinate transformation to construct a dynamic association graph with flights, autonomous shuttle vehicles, and key location points as nodes and dynamic association relationships as edges.

[0027] Furthermore, data connections are established with the airport's airside operations database, vehicle network units, and roadside sensing devices. Flight schedules and real-time status, high-frequency positioning and chassis data of autonomous shuttle buses, and traffic flow parameters from road sensors are continuously acquired through streaming processing. After this multi-source, heterogeneous data flows in, timestamp alignment and coordinate transformation are performed. This involves synchronizing all data to a unified time base based on network time protocols and mapping all geographic information to the same coordinate system on the airport's high-precision map using coordinate transformation parameters. This ensures consistency of all elements in the spatiotemporal dimensions. The process of constructing a dynamic association graph involves abstracting each flight entity, each autonomous shuttle bus entity, and each key location point entity into an independent node in the graph, and assigning initial feature attributes to each node. Specifically, for example, flight nodes include planned departure time and actual status, vehicle nodes include real-time location and battery level, and key location nodes include their function type and accessibility. The construction of edges dynamically reflects the relationships between entities, such as connecting flight nodes with their boarding gate nodes, connecting autonomous shuttle bus nodes with their road nodes, and connecting currently closely related vehicle nodes with task nodes. The weight of the edges can be dynamically updated based on distance or correlation strength, thus forming a dynamic relational graph that truly reflects the overall operational status of the airport.

[0028] S1.2 Input the dynamic association graph into the pre-trained graph neural network model, update the feature vector of each node, generate an enhanced feature vector set of global association information for flights, unmanned shuttle buses, and key location points, and encapsulate it in a structured manner to output a snapshot of dynamic data of the airport operation panorama.

[0029] Furthermore, the dynamic association graph is input into a graph neural network model that has been pre-trained on the graph data. The graph neural network model performs operations on the dynamic association graph through its message passing mechanism. Each node aggregates the feature information of its neighboring nodes and combines it with its own features to update its own feature vector. For example, the updated feature vector of a flight node not only includes its own time urgency, but also incorporates the congestion situation in the boarding gate area where it is located and the status information of the vehicles serving the flight. Similarly, the updated feature vector of an autonomous shuttle bus contains the urgency of its current task and the complexity of the surrounding traffic environment.

[0030] Specifically, multi-layered iterative message passing ensures that each node's feature vector contains rich local and global contextual information, ultimately generating an enhanced feature vector for each node in the dynamic association graph. All these enhanced feature vectors, along with the entity identifiers they represent and the dynamic relationships between them, are encapsulated in a structured data snapshot object. This set of enhanced feature vectors containing global association information serves as the output of the airport operation panoramic dynamic data snapshot. This breaks through the limitations of traditional methods that can only process independent data points. By modeling the interdependencies between entities through a graph neural network model, it extracts deep features hidden in the data relationships. The generated airport operation panoramic dynamic data snapshot is no longer just a simple accumulation of data, but contains enhanced semantic information about the interactions and influences between entities.

[0031] S2. Based on the dynamic snapshot of the airport operation panorama, calculate the dynamic passenger trip tracking index for each passenger to be served, and generate a service request queue sorted by the dynamic passenger trip tracking index.

[0032] S2.1 Extract the list of passengers to be served and the associated flight schedule departure time, flight status, and real-time passenger location information from the airport operation panoramic dynamic data snapshot.

[0033] Furthermore, this is accomplished by analyzing the node attributes and associated edge information in the panoramic dynamic data snapshot of airport operations. It identifies flight nodes with statuses of check-in or arrival, locates the list of all passengers to be served along the associated edges between flight nodes and passenger groups, and extracts the scheduled departure time and flight status from the attributes of the associated flight nodes for each passenger in the list. At the same time, it obtains the real-time location information of passengers by querying nodes representing the real-time location of passengers or by directly parsing their anonymized location data stream. It accurately extracts the core data elements directly related to the calculation of passenger itinerary closeness from the panoramic data snapshot rich in semantic relationships.

[0034] S2.2 Calculate the flight urgency factor using the flight schedule departure time and flight status, calculate the passenger trip criticality factor using passenger trip attributes, and calculate the spatial urgency factor using the passenger's real-time location.

[0035] Furthermore, by calculating three core influencing factors, the calculation of the flight urgency factor focuses on the timeliness pressure of the flight itself, the calculation of the passenger journey criticality factor focuses on the importance differences of individual passenger journey attributes, and the calculation of the spatial urgency factor focuses on the spatial difficulty for passengers to complete movement within the airport environment. The complex problem of passenger urgency assessment is decomposed into three relatively independent and quantifiable dimensions for specific analysis.

[0036] Specifically, the flight urgency factor is defined as consisting of two parts: a base value that decays over time and an excitation value triggered by the state. The basic expression for time decay is: ; in, This is the base value for time decay. Scaling factor The attenuation coefficient is... For the current time, This refers to the scheduled departure time of the flight.

[0037] The expression for the excitation value is: ; in, Current time Flight status To incentivize weights, As the incentive value, For the flight at the current time The coefficients of the real-time state discretization; For state excitation components.

[0038] The expression for the flight urgency factor is: ; in, For flight urgency factor; Specifically, the calculation of the flight urgency factor adopts a composite model. The base value of time decay uses an exponential function to simulate the natural law of nonlinear acceleration of the sense of urgency as departure time approaches. The excitation value of state triggering uses a judgment function to capture the step increase in urgency caused by flight state transitions, such as from boarding to last boarding. This more realistically simulates the dynamic evolution of flight urgency with time and key events, overcoming the limitations of simply relying on linear time progression or static state judgment, making the characterization of flight urgency more refined and realistic.

[0039] The expression for the passenger trip criticality factor is: in, As a critical factor in passenger journeys, For deblurring, For fuzzy rule engine, For blurring, For passenger attributes set.

[0040] Specifically, the fuzzifier converts clear passenger attribute sets, such as passenger type and transfer time, into membership degrees belonging to fuzzy linguistic values ​​such as important or urgent. Then, the fuzzy rule engine infers based on rules such as if the passenger type is an international transfer and the transfer time is extremely short, then the trip criticality is extremely high. The defuzzifier transforms the fuzzy inference result into a precise passenger trip criticality factor, handling the uncertainty brought about by categorical variables and subjective judgments in passenger attributes. It can simulate the decision-making thinking of human experts and flexibly evaluate the comprehensive impact of different attribute combinations on trip criticality.

[0041] The expression for the space urgency factor is: ; in, As a spatial urgency factor, Based on the level of urgency, It is an environmental regulating factor.

[0042] Specifically, the basic spatial pressure based on path planning is reflected by an exponential function with the ratio of effective walking time to critical time constant as a parameter. When the effective walking time approaches the critical time constant, the urgency increases sharply. The environmental adjustment factor quantifies the amplification effect of environmental resistance on basic spatial pressure by linearly superimposing the congestion index and the floor switching penalty as weights. By separating the passenger's own spatial movement cost from the additional resistance imposed by the external environment, it is possible to assess and synthesize the two different types of spatial pressure sources, making the measurement of spatial urgency more comprehensive.

[0043] The baseline urgency expression is: ; in, Based on the level of urgency, For effective walking time, This is the critical time constant.

[0044] The expression for the environmental regulation factor is: ; in, As an environmental regulator, The congestion index, Penalty for switching floors, This is the congestion weighting coefficient. This is the penalty weighting coefficient for switching floors.

[0045] S2.3 Input the flight urgency factor, passenger itinerary criticality factor, and spatial urgency factor into the fuzzy logic rule set for dynamic weight allocation, and calculate the dynamic passenger itinerary urgency index.

[0046] Furthermore, the precise values ​​of the flight urgency factor, passenger itinerary criticality factor, and spatial urgency factor are input into a specially configured fuzzy logic rule set for dynamic weight allocation. This rule set includes rules such as if the flight urgency factor is high, then the final catch-up index is mainly determined by it. Through fuzzy inference and defuzzification processes, a dynamic passenger itinerary catch-up index is output, solving the inflexibility problem of fixed weights when fusing multiple factors. The dynamic adaptive allocation of factor weights is achieved through fuzzy rules, so that the factors that play a dominant role in the final catch-up can be different in different contexts, thus achieving a more intelligent and context-aware catch-up assessment.

[0047] The expression for the dynamic passenger journey tracking index is: ; in, This is a dynamic passenger journey tracking index.

[0048] S2.4 Sort the list of passengers to be served in descending order according to the dynamic passenger trip tracking index, and generate a service request queue.

[0049] Furthermore, the system iterates through the list of passengers awaiting service, reads the dynamic passenger trip closeness index value for each passenger, and then uses an efficient sorting algorithm such as quicksort to sort all passengers in descending order according to the dynamic passenger trip closeness index. Finally, an ordered service request queue is generated, which transforms the closeness metric into a clear scheduling priority order, providing a clear and unambiguous execution basis for downstream scheduling decisions and ensuring that passenger requests with high closeness can be responded to first.

[0050] S3 uses the service request queue as the core input to perform multi-objective optimization calculations and generate a vehicle dispatch instruction set.

[0051] S3.1. Combining service request queues and real-time vehicle status information, construct a vehicle scheduling problem model with task set, vehicle set, and spatiotemporal and resource constraints. Define a multi-objective optimization function for the vehicle scheduling problem model with the objectives of maximizing the satisfaction of the pursuit degree, minimizing the total energy consumption, and minimizing the global traffic congestion cost.

[0052] Furthermore, each passenger request in the service request queue is defined as a task and added to the task set. At the same time, the real-time status of all available autonomous shuttle vehicles, such as location, battery level, and current load, is included in the vehicle set. A set of constraints is constructed, including capacity limits for each vehicle, battery range constraints, time constraints that the task must be completed within a time window, and road capacity constraints, thus forming a complete vehicle scheduling problem model. A multi-objective optimization function is defined for the vehicle scheduling problem model. The objective of maximizing the pursuit degree satisfaction is achieved by accumulating the dynamic passenger journey pursuit degree index of each served passenger. The objective of minimizing total energy consumption is achieved by accumulating the estimated energy consumption of all vehicles due to task execution. The objective of minimizing global traffic congestion cost is achieved by predicting the vehicle density on the critical path and calculating its congestion cost.

[0053] S3.2. The improved multi-objective ant colony algorithm is applied to solve the multi-objective optimization function, and a set of Pareto optimal scheduling schemes representing the trade-offs between different objectives is obtained.

[0054] Furthermore, an improved multi-objective ant colony algorithm is applied to solve the multi-objective optimization function, with each ant solving using a probability formula. Gradually construct a complete scheduling scheme, using probability formulas. medium pheromone concentration It encodes the experience and heuristic information of excellent solutions from historical iterations. It incorporates immediate costs such as the distance from the vehicle to the passenger, and the pheromone heuristic factor and the expectation heuristic factor control the trade-off between historical experience and current greed, respectively. After multiple rounds of iterative search, the ant colony selects the Pareto optimal scheduling scheme set that is not dominated by any other scheme from all the schemes constructed by all ants through non-dominated sorting. It uses swarm intelligence to efficiently find high-quality compromise solutions in a huge solution space. The improved multi-objective ant colony algorithm can effectively explore the solution space through the positive feedback of pheromones and the guidance of heuristic information. The obtained Pareto optimal scheduling scheme set provides decision-makers with a rich and high-quality choice space.

[0055] The probability expression for the ant constructing a solution is: ; in, For the tasks currently pending assignment Assigned to specific vehicles The probability, For the task Assigned to specific vehicles The degree to which the decision has proven to be excellent. For the task Assigned to specific vehicles The immediate appeal As the expected heuristic factor, For feasible vehicles. For all those capable of performing the task A collection of driverless shuttle buses For feasible vehicles With the task The pheromone concentration on the corresponding edge For feasible vehicles The task The heuristic expectation assigned to it, As a pheromone-inspired factor, For a specific vehicle, Tasks to be assigned These are feasible candidate vehicles.

[0056] S3.3. Select the final scheduling scheme from the Pareto optimal scheduling scheme set according to the predefined decision strategy, and convert the final scheduling scheme into a set of specific vehicle scheduling instructions that can be issued.

[0057] Furthermore, a final scheduling scheme is selected from the Pareto optimal scheduling scheme set according to a predefined decision strategy. The decision strategy can be a simple pursuit degree satisfaction maximization strategy or a more complex utility function maximization strategy based on weighted sum. Once the final scheduling scheme is selected, it will be decoded into a set of specific vehicle scheduling instructions that can be issued to each driverless shuttle vehicle. The instructions include the specified vehicle ID, the list of passenger IDs to be picked up, the precise boarding point coordinates, and the expected time window.

[0058] The expression for the Pareto optimal scheduling scheme set is: ; in, This is the set of optimal scheduling schemes for Pareto. These are candidate solutions to be evaluated. For use with Another candidate solution for comparison, For feasible solution set S4. The driverless shuttle bus executes the vehicle dispatch instruction set and plans a local route to drive to the passenger boarding point.

[0059] S4.1 The driverless shuttle vehicle receives and parses the vehicle dispatch instruction set through the vehicle network communication link to confirm the passenger boarding point location and target time window to be executed.

[0060] Furthermore, by continuously monitoring the communication channel from the central dispatch unit through the vehicle-to-everything (V2X) communication link, upon detecting and receiving a data packet destined for the vehicle's identifier, the onboard computing unit's protocol stack is invoked to parse the data packet payload, extract the effective payload of the vehicle dispatch instruction set, and then parse the coordinates of the target passenger boarding point and the expected arrival time window from the vehicle dispatch instruction set. This information is then matched and verified with the onboard high-precision map to ensure the validity and accessibility of the coordinates, ensuring that the driverless shuttle can accurately receive and understand the dispatch intentions from the central dispatch unit. Through standardized communication and parsing processes, abstract dispatch decisions are transformed into concrete geographical goals and time constraints that the vehicle can execute.

[0061] S4.2 The driverless shuttle bus uses its current location and the passenger boarding point location to call the on-board path planning algorithm, and combines high-precision map and real-time local obstacle information to calculate the optimal local driving path.

[0062] Furthermore, after confirming the target, the driverless shuttle vehicle uses its current location provided by its own GPS / INS integrated navigation system as the starting point of the path and the parsed passenger boarding point coordinates as the ending point of the path. It calls the path planning algorithm integrated in the on-board computing unit, uses the lane-level road network provided by the high-precision map as the search map, and uses the dynamic obstacle information generated by the real-time perception fusion of LiDAR, millimeter-wave radar and camera as part of the cost map. The calculation of the total cost of dynamic obstacles is that each path node n is superimposed with the repulsion field contribution of all dynamic obstacles around it. The contribution value of each obstacle decreases exponentially with the increase of distance, and the weight coefficient controls the repulsion strength. Specifically, the search algorithm ultimately generates an optimal local driving path that comprehensively considers static path length and dynamic obstacle avoidance. This transforms the macro-level scheduling objective into a micro-level driving trajectory that the vehicle can execute safely and efficiently. By incorporating dynamic obstacle information into the path search in the form of a continuous cost field, the planned optimal local driving path is not only shorter but also actively and smoothly avoids real-time obstacles, thus improving driving safety and comfort.

[0063] The optimal local driving path expression is: ; in, Path node The total cost of dynamic obstacles, A collection of dynamic obstacles. For a single dynamic obstacle, The obstacle cost weighting coefficient. Path node To a single Dynamic obstacles Euclidean distance, For path nodes.

[0064] S4.3 The control unit of the driverless shuttle bus tracks and executes the optimal local driving path, and controls the vehicle to drive along the optimal local driving path to the passenger boarding point.

[0065] Furthermore, the control unit of the driverless shuttle receives the optimal local driving path sequence output by the path planning module. This sequence consists of a series of ordered waypoints, each containing position, heading, and desired speed information. The control unit generates corresponding steering, throttle, and braking control commands by comparing the deviation between the vehicle's current pose and the target pose of the next waypoint. These commands are then executed through the drive-by-wire system, steering-by-wire system, and braking-by-wire system, ensuring that the vehicle's actual trajectory tracks the optimal local driving path until it reaches the passenger boarding point. The planned path is then transformed into the vehicle's actual motion control. Closed-loop feedback control corrects the deviation between the vehicle and the predetermined path in real time, ensuring that dispatch commands are executed accurately and that the vehicle can safely and smoothly travel along the optimal local driving path to the target.

[0066] S5. When the driverless shuttle bus is about to arrive, it dynamically adjusts the service granularity of multiple passenger service requests based on real-time capacity and demand, and generates task encapsulation instructions for shared pick-up points.

[0067] S5.1. Based on the vehicle positioning information, determine that the driverless shuttle bus is about to arrive at the passenger boarding point, trigger the dynamic service granularity adjustment process, and obtain the set of passengers waiting to be served and the real-time capacity distribution in the service request queue in the local area.

[0068] Furthermore, by continuously receiving high-frequency positioning data periodically reported by the autonomous shuttle vehicles and calculating the Euclidean distance between the vehicle's current location and the coordinates of the passenger boarding point, when this distance is less than a preset proximity threshold, it is determined that the autonomous shuttle vehicle is about to arrive, and a dynamic service granularity adjustment process is triggered. After the process is triggered, the service request queue is immediately queried, and all passengers waiting to be served within a circular area centered on the triggering vehicle's current location are selected according to a preset geographical radius to form a local area set of passengers waiting to be served. The real-time location and task status of other autonomous shuttle vehicles in the area are scanned to form a real-time capacity distribution snapshot, accurately capturing the moment when the vehicle is about to be idle and quickly perceiving the local supply and demand situation.

[0069] S5.2 Using the target boarding gate coordinates of each passenger in the set of passengers to be served as the clustering feature, the spatiotemporal clustering algorithm is used to divide passengers going to the same terminal area into different passenger request clusters, and the efficiency gain generated after merging with the current task of the triggering vehicle is calculated for each passenger request cluster.

[0070] Furthermore, the target boarding gate coordinates of each passenger in the local area's waiting passenger set are extracted as feature vectors. A density-based spatiotemporal clustering algorithm, such as DBSCAN, is used to cluster these coordinate points, grouping geographically adjacent passengers belonging to the same terminal sub-area into the same passenger request cluster. For each generated passenger request cluster, the scheduling scenario after merging it with the triggering vehicle's current task is simulated, and the efficiency gain is calculated. This is obtained by estimating the total task time required to dispatch a single vehicle to each passenger in the service passenger request cluster. The time difference between when the triggered vehicle returns to the standby point empty after completing the current task and when it immediately executes the ride-sharing task is calculated. The difference in total travel time between a carpooling route and a separate pick-up / drop-off route is then quantified by simulating the route planning algorithm.

[0071] Efficiency gain formula: ; in, For net time gain, The time saved by carpooling for the entire mission in another vehicle. To trigger the reduction in idle time caused by vehicles immediately performing ride-sharing tasks, This refers to the total additional travel time incurred by all relevant passengers due to ride-sharing. For a passenger request cluster, This refers to the total additional travel time incurred by all relevant passengers due to ridesharing.

[0072] S5.3 The efficiency gain is equal to the total time saved minus the total detour time caused by carpooling, and passenger request clusters that can be merged are selected if the efficiency gain is greater than the preset time threshold.

[0073] Furthermore, the efficiency gain of each passenger request cluster is compared with a preset positive time threshold. All passenger request clusters with efficiency gains greater than the time threshold are selected to form a set of passenger request clusters that can be merged. The purpose of this preset time threshold is to ensure that only those ride-sharing operations that can bring significant positive net benefits are considered, avoiding uneconomical merging that saves limited time but has high detour costs. Setting an economic threshold for ride-sharing feasibility filters out ride-sharing options with insignificant benefits, ensuring that dynamic service granularity adjustments always move in the direction of improving the overall operating efficiency of the system, and avoiding unnecessary increases in scheduling complexity.

[0074] S5.4 Select the passenger request cluster with the highest efficiency gain from the mergeable passenger request clusters, calculate the shared boarding point coordinates that minimize the total walking distance for all passengers in the passenger request cluster with the highest efficiency gain, and generate a task encapsulation instruction for the shared boarding point coordinates and passenger request cluster information.

[0075] Furthermore, from the set of mergeable passenger request clusters, the passenger request cluster with the largest efficiency gain is selected as the optimal cluster. An optimization problem is constructed with the objective function [function name missing]. It minimizes the distance all passengers travel from their current locations. Walk to Path planning distance The weighted sum, weight It can reflect the special needs and constraints of passengers. Ensure that the shared pick-up point is located in a feasible area of ​​the airport where vehicle parking is permitted. Within this process, the shared boarding point coordinates are obtained by solving this optimization problem. Finally, the shared boarding point coordinates are compared with... The passenger list and other information of the cluster are encapsulated into task encapsulation instructions to determine the most convenient meeting point for all members of the carpooling group. By minimizing the total walking distance, the fairness of the carpooling service and the passenger experience are improved. The generated standardized task encapsulation instructions enable carpooling scheduling to be executed accurately.

[0076] The coordinates of the shared pick-up point are expressed as follows: ; ; in, To minimize the total weighted walk cost, For the optimal cluster The first in One passenger, For passenger weight, To select shared pick-up points To passengers Current location Path planning distance, As candidate shared pick-up points, For the set of feasible solutions, For passengers.

[0077] S6. Receive task completion information from the driverless shuttle bus, and optimize and update the dynamic passenger journey tracking index and the strategy model for dynamic service granularity adjustment based on the task completion information.

[0078] S6.1 Receive the task completion information fed back by the unmanned shuttle vehicle, and align the actual pick-up time and actual travel time in the task completion information with the estimated waiting time and estimated travel time recorded during the scheduling decision to form a task experience data record.

[0079] Furthermore, the system continuously monitors the data receiving port. When it receives a data packet containing task completion information sent back by the autonomous shuttle vehicle through the vehicle-to-everything (V2X) communication link, it parses the data packet to extract fields such as the actual pick-up timestamp and actual travel time. Based on the task's unique identifier, it retrieves the estimated waiting time and estimated travel time recorded at the time of the decision from the scheduling decision history log. It then precisely associates the actual and estimated values ​​of the same task according to the time sequence and task identifier, and supplements the task context information to form a complete task experience data record, which is stored in a dedicated database. This transforms discrete, one-off scheduling and execution feedback into continuous, traceable, and structured experience data.

[0080] S6.2. Based on task experience data records, calculate the waiting time deviation between the actual waiting time and the expected waiting time to obtain the efficiency gain deviation between the actual efficiency gain and the estimated efficiency gain.

[0081] Furthermore, the historical task experience data record library calculates the actual waiting time for each record by subtracting the passenger request time from the actual pick-up time, and the estimated waiting time by subtracting the passenger request time from the estimated pick-up time. The absolute value of the difference between the two is taken as the waiting time deviation. For task records involving dynamic service granularity adjustments, the actual time savings brought by the ride-sharing operation are recalculated based on the actual pick-up time and actual travel route as the actual efficiency gain, and compared with the efficiency gain estimated during scheduling decisions to obtain the efficiency gain deviation. This quantifies specific operational results into calculable performance indicator deviations, transforming complex scheduling effect evaluations into clear, optimizable numerical signals, providing a direct optimization direction for parameter adjustments.

[0082] S6.3. Using the waiting time deviation as a loss signal, the weight parameters used in the dynamic passenger journey pursuit index are adjusted through an online stochastic gradient descent algorithm.

[0083] Furthermore, the waiting time deviation is used as the value of the loss function to characterize the prediction accuracy of the dynamic passenger trip pursuit index. The online stochastic gradient descent algorithm calculates the gradient of the loss function with respect to the weight parameters used in the calculation of the dynamic passenger trip pursuit index (such as the fusion weights of flight urgency factor, passenger trip criticality factor, and space urgency factor) based on the loss value. These weight parameters are updated in a small step along the gradient in the opposite direction, so that the adjustment of the weight parameters is in the direction of reducing the waiting time deviation in future predictions.

[0084] Specifically, a closed-loop optimization path is established from the feedback of scheduling results to the parameters of the priority evaluation model, so that the calculation model of the dynamic passenger trip urgency index can adapt to the actual operating environment and continuously fine-tune its internal weight allocation with the accumulation of experience, thereby continuously improving its accuracy in predicting the true urgency of passengers.

[0085] S6.4. Using efficiency gain deviation as a reward signal, the decision parameters of the strategy model for dynamically adjusting service granularity are adjusted through the strategy gradient algorithm.

[0086] Furthermore, by treating the efficiency gain deviation as a scalar reward signal, the policy gradient algorithm uses this reward signal to evaluate the quality of historical decisions and calculates the policy gradient to update the decision parameters of the policy model for dynamic service granularity adjustment (such as parameter preferences of clustering algorithms or efficiency gain judgment thresholds). If the efficiency gain deviation is positive, it indicates that the historical decision has generated unexpected positive returns. In this case, the policy gradient algorithm will adjust the decision parameters to increase the probability of making similar decisions in similar situations in the future. Conversely, it will reduce the probability of such decisions, allowing the dynamic service granularity adjustment strategy to learn from the actual returns of decisions. This makes the multiplication decision no longer dependent on fixed rules, but evolves into an adaptive strategy that can continuously optimize itself based on the historical multiplication effect, thereby continuously improving the effectiveness and economy of dynamic service granularity adjustment decisions.

[0087] This embodiment also provides a scheduling system for unmanned shuttle buses at airports, including: a fusion module that acquires flight information, status information of all unmanned shuttle buses and traffic information of airport roads, and fuses them to generate a snapshot of dynamic panoramic data of airport operations; The evaluation module calculates the dynamic passenger journey tracking index for each passenger to be served based on a snapshot of dynamic data of the airport operation panorama, and generates a service request queue sorted by the dynamic passenger journey tracking index. The multi-objective collaborative scheduling module uses the service request queue as the core input to perform multi-objective optimization calculations, generate a vehicle scheduling instruction set, and the driverless shuttle bus executes the vehicle scheduling instruction set to plan a local route to the passenger boarding point. The adjustment module dynamically adjusts the service granularity of multiple passenger service requests based on real-time capacity and demand when the driverless shuttle bus is about to arrive, and generates task encapsulation instructions for the shared pick-up point. The optimization module receives task completion information from the driverless shuttle bus and optimizes and updates the dynamic passenger journey tracking index and the strategy model for dynamic service granularity adjustment based on the task completion information.

[0088] This embodiment also provides a computer device applicable to the scheduling method of unmanned airport shuttle buses, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the scheduling method of unmanned airport shuttle buses as proposed in the above embodiment.

[0089] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0090] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the scheduling method for unmanned airport shuttle buses as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0091] In summary, this invention generates a refined index that quantifies passenger urgency by integrating multiple factors such as flight urgency, passenger itinerary criticality, and spatial urgency, and using fuzzy logic for dynamic weight fusion. This solves the core problem of existing technologies, which have coarse priority judgment standards and cannot accurately respond to high-urgency demands. It achieves optimal allocation of capacity at critical moments. By establishing an online learning loop, the system continuously corrects the parameters of the urgency calculation model and service adjustment strategy using actual operational data, enabling the system to learn from experience and improve itself. This overcomes the shortcomings of traditional static models, such as poor adaptability and long-term performance degradation. Together, these factors ensure that the scheduling system not only has initial decision-making intelligence but also can continuously evolve, ultimately comprehensively improving the efficiency, accuracy, and adaptability of airport shuttle services.

[0092] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for dispatching an unmanned airport shuttle, the method comprising: This includes acquiring flight information, status information of all driverless shuttle vehicles, and traffic information of airport roads, and integrating them to generate a snapshot of dynamic panoramic data of airport operations; ​ Based on the panoramic dynamic data snapshot of airport operations, the dynamic passenger trip tracking index of each passenger to be served is calculated, and a service request queue sorted by the dynamic passenger trip tracking index is generated. The service request queue serves as the core input for multi-objective optimization calculations, generating a vehicle dispatch instruction set. The driverless shuttle bus executes the vehicle dispatch instruction set and plans a local route to the passenger boarding point. When the driverless shuttle bus is about to arrive, the service granularity of multiple passenger service requests is dynamically adjusted according to real-time capacity and demand, and a task encapsulation instruction for the shared pick-up point is generated. The system receives task completion information from the driverless shuttle bus and optimizes and updates the dynamic passenger journey tracking index and the strategy model for dynamic service granularity adjustment based on the task completion information.

2. The method of claim 1, wherein: Acquire flight information, status information of all driverless shuttle vehicles, and traffic information of airport roads, and integrate them to generate a dynamic snapshot of the airport's overall operations. This includes the following steps: By acquiring flight information, autonomous shuttle vehicle status information, and airport road traffic information through data interfaces, and performing timestamp alignment and coordinate transformation, a dynamic association graph is constructed with flights, autonomous shuttle vehicles, and key location points as nodes and dynamic association relationships as edges. The dynamic association graph is input into a pre-trained graph neural network model to update the feature vector of each node, generate an enhanced feature vector set of global association information for flights, unmanned shuttle buses, and key location points, and encapsulate it in a structured manner to output a snapshot of dynamic panoramic data of airport operations.

3. The method of claim 2, wherein: Based on a snapshot of dynamic airport operation data, a dynamic passenger journey tracking index is calculated for each passenger awaiting service. A service request queue is then generated, sorted by the dynamic passenger journey tracking index, including the following steps: Extract the list of passengers to be served and the associated flight schedules, departure times, flight status, and real-time passenger location information from the airport's panoramic dynamic data snapshot; The flight urgency factor is calculated using flight schedule departure time and flight status; the passenger trip criticality factor is calculated using passenger trip attributes; and the spatial urgency factor is calculated using passenger real-time location. The flight urgency factor, passenger itinerary criticality factor, and spatial urgency factor are input into the fuzzy logic rule set for dynamic weight allocation, and the dynamic passenger itinerary urgency index is calculated. The list of passengers to be served is sorted in descending order based on the dynamic passenger trip tracking index, and a service request queue is generated.

4. The method of claim 3, wherein: The service request queue serves as the core input for multi-objective optimization calculations to generate a vehicle dispatch instruction set, including the following steps: By combining service request queues and real-time vehicle status information, a vehicle scheduling problem model with task set, vehicle set, and spatiotemporal and resource constraints is constructed. A multi-objective optimization function is defined for the vehicle scheduling problem model with the objectives of maximizing the satisfaction of the pursuit degree, minimizing the total energy consumption, and minimizing the global traffic congestion cost. The improved multi-objective ant colony algorithm is applied to solve the multi-objective optimization function to obtain a set of Pareto optimal scheduling scheme sets representing the trade-off between different objectives; According to the predefined decision strategy, a final scheduling scheme is selected from the set of Pareto optimal scheduling schemes, and the final scheduling scheme is converted into a specific vehicle scheduling instruction set that can be issued.

5. The method of claim 4, wherein: The unmanned shuttle vehicle executes the vehicle scheduling instruction set and plans a local path to the passenger pickup point, including the following steps: The unmanned shuttle vehicle receives and parses the vehicle scheduling instruction set through the vehicle networking communication link, and confirms the passenger pickup point position and target time window to be executed; Based on the current position of the unmanned shuttle vehicle and the passenger pickup point position, the vehicle calls the on-board path planning algorithm, combines the high-precision map and real-time local obstacle information, and calculates the optimal local driving path; The control unit of the unmanned shuttle vehicle tracks and executes the optimal local driving path to control the vehicle to drive along the optimal local driving path to the passenger pickup point.

6. The method of claim 5, wherein: When the unmanned shuttle vehicle is about to arrive, the service granularity of multiple passenger service requests is dynamically adjusted according to the real-time capacity and demand, and task packaging instructions for shared pickup points are generated, including the following steps: Determine whether the unmanned shuttle vehicle is about to arrive at the passenger pickup point through vehicle positioning information, trigger the dynamic service granularity adjustment process, and obtain the set of passengers to be served in the service request queue in the local area and the real-time capacity distribution; Use the target boarding gate coordinates of each passenger in the set of passengers to be served as clustering features, and use a spatiotemporal clustering algorithm to divide the passengers to be served going to the same terminal area into different passenger request clusters, and calculate the efficiency gain produced by merging the current task of the vehicle for each passenger request cluster; The efficiency gain is equal to the total time saved minus the total detour time caused by carpooling, and the passenger request clusters with efficiency gain greater than the preset time threshold are selected; Select the passenger request cluster with the highest efficiency gain from the mergable passenger request clusters, calculate the shared pickup point coordinates that minimize the total walking distance for all passengers in the passenger request cluster with the highest efficiency gain, and generate task packaging instructions for the shared pickup point coordinates and passenger request cluster information.

7. The airport driverless shuttle bus scheduling method as described in claim 6, characterized in that: Receive the task completion information fed back by the unmanned shuttle vehicle, and optimize and update the dynamic passenger journey tightness index and the strategy model for dynamic service granularity adjustment based on the task completion information, including the following steps: Receive the task completion information fed back by the unmanned shuttle vehicle, align the actual pickup time and actual driving time in the task completion information with the predicted waiting time and predicted driving time recorded at the time of scheduling decision, and form a task experience data record; Calculate the waiting time deviation between the actual waiting time and the predicted waiting time based on the task experience data record to obtain the efficiency gain deviation between the actual efficiency gain and the estimated efficiency gain; Use the waiting time deviation as a loss signal to adjust the weight parameters used by the dynamic passenger journey tightness index through an online stochastic gradient descent algorithm; Use the efficiency gain deviation as a reward signal to adjust the decision parameters of the strategy model for dynamic service granularity adjustment through a policy gradient algorithm.

8. A dispatch system for an airport unmanned people mover based on the dispatch method for an airport unmanned people mover according to any one of claims 1 to 7, characterized in that: The method comprises: a fusion module, which acquires flight information, state information of all unmanned shuttles and traffic information of airport roads, and generates a snapshot of airport operation panoramic dynamic data through fusion; an evaluation module, which calculates a dynamic passenger travel pursuit index of each passenger to be served based on the snapshot of airport operation panoramic dynamic data, and generates a service request queue sorted according to the dynamic passenger travel pursuit index; a multi-target coordinated scheduling module, which performs multi-target optimization calculation on the service request queue as a core input, generates a vehicle scheduling instruction set, and drives the unmanned shuttles to the pickup points of passengers along local paths according to the vehicle scheduling instruction set; an adjustment module, which performs dynamic service granularity adjustment on multiple passenger service requests according to real-time capacity and demand when the unmanned shuttles are about to arrive, and generates task packaging instructions for shared pickup points; an optimization module, which receives task completion information fed back by the unmanned shuttles, and optimizes and updates the dynamic passenger travel pursuit index and the strategy model of dynamic service granularity adjustment based on the task completion information. 9.A computer device, comprising a memory and a processor, the memory storing a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the scheduling method of the airport unmanned shuttle according to any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the scheduling method of the airport unmanned shuttle according to any one of claims 1-7.

Images