A flight area vehicle group continuous time path planning method and system based on risk field perception and hard-soft mixed constraint

Abstract

The application provides a flight area vehicle group continuous time path planning method and system based on risk field perception and hard-soft mixed constraint, relates to the field of intelligent transportation, and comprises the following steps: obtaining a continuous time path according to a task set and directed graph data in a direction with minimum cost function; performing hard constraint and soft cost constraint in the search process; performing hard constraint according to a block time interval set, and adding soft cost to a cost function to perform soft cost constraint according to risk field data; performing conflict detection on the continuous time path, adding conflict constraint according to a conflict detection result, and re-searching according to the conflict constraint until an optimized continuous time path is obtained through conflict detection; calculating risk adjustment capacity according to vehicle path and risk field data; judging according to the risk adjustment capacity; and generating an executable continuous time path set of each vehicle according to a judgment result. The application can guide ground service vehicles to avoid high-risk space-time regions and reduce aircraft-vehicle conflict risk on the basis of maintaining aircraft hard safety constraint and vehicle conflict-free constraint.

Description

Technical Field

[0001] This invention relates to the fields of intelligent civil aviation operation support and intelligent transportation technology, and more specifically, to a method and system for continuous time path planning of vehicle groups in the flight area based on risk field perception and hard-soft hybrid constraints. Background Technology

[0002] Airport ground service vehicles need to perform high-frequency support tasks between garages, aircraft stands, support operation areas, and service roads. Vehicle routing planning not only needs to provide spatial routes but also needs to determine the continuous time sequence of vehicle entry nodes, occupancy edges, waiting, and departure. Because aircraft have high operational priority at taxiway and service road intersections, aircraft pushback areas, and apron entrances and exits, vehicle routing planning must avoid aircraft occupancy time windows and maintain safe temporal and spatial separation from other vehicles.

[0003] Existing multi-agent path planning methods typically employ conflict-first search, A* search, or their variants to address multi-vehicle path conflict problems. These methods can handle vertex and edge conflicts between vehicles, but they often eliminate conflict states through hard constraints and rarely utilize future aircraft-vehicle conflict risk field information. In airport flight area scenarios, different service road edges and nodes have different risk levels in the future. If path planning only guarantees that current hard conflicts do not occur, without considering probabilistic risk information for high-risk areas, it may choose to traverse nodes or road segments with higher future risks, preventing controlled vehicles from traveling safely. On the other hand, simply closing all high-risk areas can easily lead to path infeasibility or excessive waiting, affecting the efficiency of vehicle mission execution. Summary of the Invention

[0004] In view of this, the present invention proposes a continuous time path planning method and system for vehicle groups in the flight area based on risk field perception and hard-soft hybrid constraints, in order to solve the problems of existing flight area vehicle group path planning methods that usually only use hard conflict constraints, making it difficult to use edge / node probabilistic risk fields to guide path selection, and also lacking a quantitative relationship between risk fields and path capacity safety.

[0005] To achieve the above objectives, this invention proposes a continuous-time path planning method for vehicle groups in flight zones based on risk field perception and a hybrid hard-soft constraint, comprising: Obtain directed graph data of the task set and service road network of ground service vehicles; obtain the set of congestion time intervals and risk field data; the directed graph data includes road nodes and edges representing movement between nodes; A continuous time path is obtained by searching in the direction of minimizing the cost function based on the task set and directed graph data; the cost function is used to characterize the cost of the risk-aware path; hard constraints and soft cost constraints are applied during the search process; hard constraints are applied based on the set of blocking time intervals, and soft costs are calculated based on the risk field data and added to the cost function for soft cost constraints. Detect conflicts in continuous time paths, add conflict constraints based on the conflict detection results, and re-search based on the conflict constraints until the conflict detection is passed, thus obtaining an optimized continuous time path. Calculate the risk adjustment capacity based on vehicle path and risk field data; make a judgment based on the risk adjustment capacity; and generate an executable continuous time path set for each vehicle based on the judgment result.

[0006] Optionally, the cost function for:

[0007] in, Basic path cost, Indicates a path. and These are the edge risk weights and the node risk weights, respectively. For vehicle i in the time bucket The duration of time that edge e is occupied. For vehicle i in the time bucket The duration of time that node v is occupied.

[0008] Optionally, in the hard constraint, when the time when a vehicle reaches the first node falls within the blocking time interval of the first node, the node-time state in the vehicle path corresponding to the vehicle is determined to be unavailable, and a pruning operation is performed. The first node is the node through which the aircraft passes.

[0009] Optionally, in soft cost constraints, during the search process, when a vehicle occupies a node or edge within a time bucket, the risk value data corresponding to the time bucket and the node or edge is added to the cost function; the time bucket is obtained by dividing the path planning time into fixed time widths.

[0010] Optionally, conflict detection includes vertex conflict, same-direction edge conflict, and opposite-direction edge conflict, wherein vertex conflict is when two vehicles occupy the same node within an overlapping time window; same-direction edge conflict is when two vehicles occupy the same side on the same side and the safety interval is insufficient; and opposite-direction edge conflict is when two vehicles occupy the same side in opposite directions and the occupation time intervals overlap.

[0011] Optionally, risk adjustment capacity includes side risk adjustment capacity. and node risk adjustment capacity

[0012]

[0013] in, Let e ​​be the original capacity of edge e. Let v be the original capacity of node v. This indicates a risk field exceeding the limit. This indicates a risk field where nodes exceed limits.

[0014] Optionally, the process of determining the risk-adjusted capacity includes: Does the set of vehicle paths satisfy the safety constraints? The safety constraints are:

[0015]

[0016] in, Let i be the capacity weight. For vehicle i in the time bucket The indicator variable for the inner occupancy edge e, For vehicle i in the time bucket The indicator variable for the internally occupied node v. Represents a set of nodes. Let T represent the set of edges, and T represent the planning time. Does the opportunity calibration safety guarantee meet the following requirements? All edge and node capacity constraints are satisfied.

[0017] Among them, total risk budget , , This represents the risk budget for edge-time buckets and node-time buckets.

[0018] Optionally, after generating the executable continuous time path set for each vehicle, the method further includes: evaluating the node risk exposure index of the executable continuous time path set.

[0019] On the other hand, the present invention provides a continuous time path planning system for a group of vehicles in a flight zone based on risk field perception and hard-soft hybrid constraints, for executing the above-mentioned method.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention introduces edge over-limit risk field and node over-limit risk field into continuous time path search, enabling vehicle path planning to proactively avoid service road edges and nodes with higher future risks; 2. This invention uses the aircraft occupancy time window as a hard safety constraint and the risk field as a soft cost, which can ensure basic safety while avoiding excessive waiting caused by simply closing high-risk areas; 3. This invention distinguishes the different safety implications of general traffic sections and aircraft-vehicle conflict points by using independent edge risk weights and node risk weights; 4. This invention transforms the probabilistic risk field into a capacity constraint by adjusting the risk capacity, providing a quantitative basis for capacity safety verification of path sets; 5. The continuous time path output by this invention includes node sequence, arrival time, waiting time, and security verification results, which facilitates integration with flight area vehicle scheduling systems or automatic driving control systems. Attached Figure Description

[0021] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. In the drawings: Figure 1 This is a flowchart illustrating the risk-aware continuous-time path planning method in an embodiment of the present invention. Figure 2 This is a schematic diagram of the hard-soft hybrid constraint architecture in an embodiment of the present invention; Figure 3 This is a schematic diagram of the high-level constraint tree and low-level time window search relationship in a continuous time conflict priority search embodiment of the present invention. Figure 4 This is a schematic diagram illustrating the relationship between risk adjustment capacity and capacity security verification in an embodiment of the present invention. Detailed Implementation

[0022] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] This invention relates to the fields of intelligent civil aviation operation support and intelligent transportation technology, specifically to a method and system for continuous time path planning of vehicle groups in the flight area based on risk field perception and a hybrid hard and soft constraint. The method acquires the task set of multiple ground service vehicles, the flight area service road network, aircraft occupancy time windows, and edge and node over-limit risk fields predicted by macroscopic conservation laws; constructs a risk-aware path cost function including basic passage cost, edge risk cost, and node risk cost; under a continuous time conflict-first search framework, the aircraft occupancy time window is used as an inviolable hard safety constraint, and the risk field is used as a soft cost to guide path selection, generating continuous time paths for single vehicles in the low-level time window search; vehicle-to-vehicle conflict detection is performed on multi-vehicle paths, and time window constraints are added through a high-level constraint tree for replanning; after no vehicle conflict, the risk adjustment capacity is calculated based on the edge / node risk field, and the capacity safety condition is verified. This invention can guide ground service vehicles to avoid high-risk spatiotemporal areas and reduce the exposure to aircraft-vehicle conflict risks while maintaining aircraft hard safety constraints and vehicle-to-vehicle conflict-free constraints.

[0024] Example 1 This invention proposes a continuous-time path planning method for vehicle groups in flight zones based on risk field perception and a hybrid hard and soft constraint: aircraft occupancy time windows and vehicle-to-vehicle conflicts are treated as hard constraints, edge / node over-limit risks output by the risk field prediction system are introduced into the path search as soft costs, and the planning results are verified for safety through risk adjustment capacity.

[0025] To achieve the above objectives, this invention provides a continuous-time path planning method for vehicle groups in flight zones based on risk field perception and a hybrid hard-soft constraint, such as... Figure 1 As shown, it includes the following steps: S1, Obtain task sets from multiple ground service vehicles And obtain the directed graph of the service roads. .

[0026] S2, obtain the set of congestion time intervals for aircraft at the conflict point, and obtain the edge over-limit risk field. and node over-limit risk field ; S3, Construct the risk perception path cost function; S4. Initialize the high-level constraint tree and perform a low-level time window search for each vehicle to generate a continuous time path that satisfies the vehicle task constraints and the aircraft occupancy time window hard constraints. S5, when the low-level search is in the unfolded state, the aircraft occupancy time window is used as a hard constraint for pruning, and the edge over-limit risk field and node over-limit risk field are added as soft costs to the path cost. S6, perform vehicle-to-vehicle conflict detection on the generated multi-vehicle paths. If there are vertex conflicts, same-direction edge conflicts, or opposite-direction edge conflicts, generate vehicle-position-time window constraints and add them to the higher-level constraint tree. Return to S4 to search for constrained vehicle paths again. S7. If there is no vehicle-to-vehicle conflict, calculate the risk adjustment capacity based on the edge over-limit risk field and the node over-limit risk field. S8, verify that the path set meets the capacity safety conditions; S9 outputs the set of executable continuous time paths for each vehicle.

[0027] In one implementation, in step S1, the task set includes a vehicle index, vehicle start point, end point, release time, and deadline; the directed graph of the service road includes a set of service road nodes. Edge set The set of nodes includes node coordinates, edge lengths, and edge travel times, where the service road node set V contains different representative road nodes v, and the edge set... The edge e contains different representations of the movement paths and directions between road nodes.

[0028] In one implementation, step S2 further includes obtaining a time bucket with a fixed time width under the planning time; the edge over-limit risk field and the node over-limit risk field are output by the risk field prediction system.

[0029] The risk field is predicted and obtained as follows: First, based on the aircraft operation plan, real-time monitoring data, and historical disturbance distribution, the aircraft pushback, taxiing, waiting, and crossing processes within the future planning time window are sampled to obtain multiple aircraft process samples; then, based on the time and location of the aircraft passing through the conflict interface, the capacity limit on the service road edge and node passage capacity is calculated; then, the ground vehicle density conservation equation and the node reservoir queuing equation are solved on the directed graph of the service road to obtain the edge density trajectory and node queuing trajectory for each sample; finally, within a fixed time bucket, the probability of the edge density exceeding the critical density threshold and the probability of the node queuing vehicle mass exceeding the critical value are statistically analyzed to obtain the edge over-limit risk field and the node over-limit risk field, respectively.

[0030] The equations include the ground support vehicle density conservation equation and the nodal reservoir queuing equation. The vehicle density conservation equation uses the Lighthill–Whitham–Richards macroscopic traffic flow conservation law (LWR model); the nodal reservoir queuing equation uses the nodal mass conservation relationship (inflow flux minus outflow flux equals the change in reservoir mass). Specifically: The queuing equation for the node reservoir is:

[0031] In the formula, Let v be the state variable (cumulative number of vehicles). This represents the rate of change of the state over time; and These represent the sets of all edges flowing into and out of the node, respectively. Let be the flux function on edge e, representing the number of vehicles transported through that boundary per unit time, with its independent variable t being the current time. This represents the boundary value at the upstream end of the edge (the side entering the node). This represents the boundary value at the downstream end of the edge (away from the node). For density or state distribution variables on the edge; For source terms or state transition terms within a node, used to describe the state changes of the node itself due to processes such as physicochemical reactions, emissions, consumption, or service switching. Its parameter A is the characteristic area (or cross-sectional area) of the node. For the velocity vector, Input the flow intensity at the node. This refers to the density or concentration parameter at the node.

[0032] Specifically, the risk field exceeding the limit can be represented as:

[0033] The node over-limit risk field can be represented as:

[0034] in, For the first Service roadside under individual aircraft process samples Vehicle density, Let V be the mass of vehicles in the reservoir or the number of vehicles in the queue at node v under the nth sample. and These are the edge critical density threshold and the node critical vehicle mass threshold, respectively. N is the total number of Monte Carlo samples for the aircraft process, and n=1,2,…,N is the sample index; The physical length of the service road edge e; Δτ is the width of the time bucket, t τ and t τ+1 Let τ be the start and end times of the time bucket; 1{·} is the indicator function. The aforementioned risk field serves as input to the path planning module, used for subsequent calculation of soft risk costs and verification of risk adjustment capacity.

[0035] In one implementation, in step S3, the risk perception path cost function is:

[0036] in, Basic path cost, Indicates a path. and These are the edge risk weights and the node risk weights, respectively. For vehicle i in the time bucket The duration of time that edge e is occupied. For vehicle i in the time bucket The duration of time that node v is occupied.

[0037] In one implementation, in step S4, the continuous time path of vehicle i Represented as:

[0038] in, Let k be the k-th node of vehicle i. The time when the vehicle arrives at this node. This represents the total number of nodes in the path.

[0039] In one implementation, in step S5, the basic path cost... Including passage time and waiting time, it is as follows:

[0040] in, For path The cumulative waiting time Weighted by waiting time.

[0041] In one implementation, in step S4, the aircraft occupancy time window is a hard safety constraint; if the time t when the vehicle arrives at node v falls within the aircraft blocking time interval associated with that node, then the low-level time window search will determine the corresponding state (v,t) as unavailable.

[0042] The edge over-limit risk field and node over-limit risk field are soft routing costs; when a vehicle occupies an edge or node, the low-level time window search increases the path cost according to the risk field value in the corresponding time bucket, but does not directly exclude the edge or node.

[0043] The "increased cost" refers to the weighted summation of risk field values ​​based on the actual vehicle occupancy time when calculating the cost of candidate paths or actions in the low-level time window search. Specifically, when a vehicle occupies edge e, the system calculates the intersection time of that occupancy time interval with each risk time bucket. and will Incorporate path cost; when a vehicle occupies node v, the system calculates the intersection duration of the node occupancy time and each risk time bucket. and will Add path cost. The higher the risk value and the longer the time occupied, the greater the added cost. This risk cost only changes the priority of path search and does not directly make the edge or node impassable; aircraft occupancy time windows and vehicle conflict constraints are still treated as hard constraints for pruning. Specifically: In the process of increasing path cost: When evaluating a vehicle path or a search action in low-level time window search, an additional risk cost is added on top of the original costs such as travel time and waiting time. The risk cost is calculated as "risk field value × vehicle occupancy time × risk weight". Vehicles passing through edges or nodes with higher risk will not be directly prohibited, but the total cost of the path will increase, thus reducing its priority in A* or time window search, causing the algorithm to prioritize detour paths or waiting solutions with lower risk.

[0044] In one implementation, the path cost of low-level time window search includes basic path cost and risk path cost. For vehicles... Candidate paths The total cost is:

[0045] in, The basic path cost can include vehicle travel time, waiting time, turning costs, or service delay costs. To serve the roadside e in the time bucket The risk value of exceeding the limit within the boundary; For node v in the time bucket The risk value of node over-limit within the range; For vehicle i in the time bucket The duration of time that edge e is occupied; For vehicle i in the time bucket The duration of time node v is occupied; and These are edge risk weights and node risk weights, used to adjust the degree of influence of risk cost on the total path cost.

[0046] The time interval for a vehicle to occupy an edge may span multiple time buckets. If vehicle i starts from time... Enter edge e, and at time e If it leaves edge e, then it is in the time bucket. The time taken for the inner edge e is:

[0047] Similarly, if vehicle i starts from time... Enter or arrive at node v, and at time If a node leaves node v, it is in the time bucket. The duration of occupancy of the inner node v is:

[0048] When expanding a candidate action, the low-level time window search queries the risk field value within the corresponding time bucket based on the edge occupancy interval or node occupancy interval corresponding to the action, and adds the risk increment to the action cost. For the action of vehicle i traveling on edge e from time t_a to time t_b, its edge risk increment is:

[0049] For vehicle i's action of waiting or occupying at node v from time t_c to time t_d, the node risk increment is:

[0050] Therefore, the cost of an action in a low-level time window search can be expressed as:

[0051] in, This refers to the passage time or duration of the action. The waiting time generated by this action. This represents the waiting weight. If the action does not occupy a node or generate a wait, the corresponding value can be 0.

[0052] The aforementioned risk cost is incorporated as a soft cost into the evaluation function of the low-level time window search. For example, in A* search, the evaluation function of node state s can be expressed as: f(s) = g(s) + h(s) Where g(s) is the cumulative cost from the starting point to state s, which includes the basic travel cost, waiting cost, and the risk cost of the aforementioned edges / nodes; h(s) is the heuristically estimated cost from state s to the target point. Since high-risk edges or high-risk nodes only increase g(s) and do not directly make the state infeasible, this mechanism belongs to soft constraint or soft routing cost.

[0053] The aircraft occupancy time window is still treated as a hard constraint. That is, if the time when a vehicle enters a node or edge conflicts with the aircraft occupancy time window, the corresponding action will be pruned directly; if there is no hard conflict, but the edge or node is of high risk in the corresponding time bucket, the action is still feasible, but at a higher cost.

[0054] Example demonstration: For example, the risk time bucket width is 30 seconds. Vehicle i occupies service roadside e_4 from 08:30:10 to 08:30:50, and this occupancy spans two time buckets: 1. From 08:30:00 to 08:30:30, 20 seconds will be used; 2. From 08:30:30 to 08:31:00, 20 seconds are occupied.

[0055] like , And edge risk weight =1, then the risk increment of this edge is:

[0056] If the travel time of another bypass edge is slightly longer, but the corresponding time bucket risk value is lower, the search algorithm will weigh the "increased travel time" against the "reduced risk cost". In this way, high-risk edges or nodes will not be directly eliminated, but their probability of being selected will decrease due to the increased total cost.

[0057] In one implementation, in step S6, vehicle-to-vehicle conflicts include vertex conflicts, same-direction edge conflicts, and opposite-direction edge conflicts. A vertex conflict indicates that two vehicles occupy the same node within an overlapping time window; a same-direction edge conflict indicates that a following vehicle catches up with a preceding vehicle on the same side, i.e., occupying the same side and the safety interval is insufficient; an opposite-direction edge conflict indicates that two vehicles are traveling towards each other on the same side and their occupation time intervals overlap, i.e., two vehicles are traveling towards each other and occupying the same side with overlapping occupation time intervals.

[0058] In one implementation, in step S7, the risk-adjusted capacity is... and node risk adjustment capacity satisfy:

[0059] in, Let e ​​be the original capacity of edge e. This represents the original capacity of node v.

[0060] In one implementation, in step S8, the path set satisfies the following capacity safety constraints, which include edge capacity safety constraints and node capacity safety constraints:

[0061]

[0062] in, Let i be the capacity weight of vehicle i; For vehicle i in the time bucket The binary indicator variable for edge e, which indicates the time interval and time bucket of edge occupancy for vehicle i. When there is an intersection, ,otherwise ; For vehicle i in the time bucket The binary indicator variable for the occupancy of node v, when the node occupancy time interval and time bucket of vehicle i. When there is an intersection, ,otherwise ; This represents the set of time buckets obtained by dividing the planning time window.

[0063] Furthermore, the opportunity calibration safety guarantee is expressed as: Path set All edge and node capacity constraints Ce and Cv are satisfied.

[0064] Among them, total risk budget , , This represents the risk budget for edge-time buckets and node-time buckets.

[0065] and The capacity violation risk budget allocated to edge-time buckets and node-time buckets represents the permissible probability of "actual available capacity being lower than planned capacity". These can be obtained in three ways: one is by using the total risk budget... The allocation methods are: 1) uniform distribution across all edges, nodes, and time buckets; 2) weighted allocation based on edge / node importance, risk field value, or expected occupancy; and 3) calculation using Monte Carlo aircraft process samples, i.e., the actual available capacity in the statistical sample. or Below planned capacity or The proportion. Summing the above smaller budgets yields the total risk budget. This ensures that the probability of the path set satisfying all capacity constraints is no less than [a certain value]. . Specifically, if and Let e ​​and v be the actual available capacities of the lower edge e and node v of the nth sample, respectively. and It can be estimated based on the proportion of samples where the actual available capacity is lower than the planned capacity.

[0066]

[0067]

[0068] In one implementation, step S9 further includes evaluating the node risk exposure index of the executable continuous time path set, the node risk exposure index... for:

[0069] In one implementation, in step S9, the executable continuous time path set output by the path output module includes the node access sequence, edge passage sequence, departure time, arrival time, waiting time, vehicle-to-vehicle conflict detection result, and capacity safety verification result for each vehicle.

[0070] Example 2 Building upon the technology of Embodiment 1, this invention provides a continuous-time path planning method for vehicle groups in flight zones based on risk field perception and a hybrid hard-soft constraint. This method takes the dynamic road network of the service road and the risk field prediction results as input, and incorporates risk cost and capacity safety verification into the continuous-time conflict-first search framework, enabling path planning to take into account aircraft hard safety constraints, multi-vehicle conflict-free constraints, and risk avoidance objectives.

[0071] 1. Input data The system inputs include vehicle task sets, a directed graph of service roads, aircraft occupancy time windows, edge over-limit risk fields, and node over-limit risk fields. The vehicle task set includes vehicle number, vehicle type, origin, destination, release time, and deadline. The directed graph of service roads includes 30 service road nodes, 67 service road edges, node coordinates, edge lengths, and travel times.

[0072] The aircraft occupancy time window is determined by the aircraft taxiing plan, conflict interface location, and operating rules. Edge over-limit risk fields and node over-limit risk fields are output by the risk field prediction system, and the risk field time bucket width can be set to 30 seconds.

[0073] In practical implementation, vehicle task set This information can be provided by the airport vehicle dispatching system, support operation plan, or real-time dispatching system. Each task must include at least the vehicle number, vehicle type, origin, destination, and release time; optionally, it may also include a deadline, priority, task service duration, allowed waiting area, maximum vehicle speed, and vehicle capacity weight. And attributes such as whether detours are allowed. If a vehicle needs to perform multiple support tasks consecutively, the system can break it down into multiple path planning sub-tasks with sequential constraints, or treat multiple task points as consecutive access nodes in the same vehicle's path.

[0074] The directed graph of the service roads can be provided by a dynamic road network system. In addition to the static topology of nodes and edges, the system also reads edge length, direction of travel, baseline travel time, speed limit, available stopping nodes, prohibited vehicle types, and road capacity. Attributes such as these are also considered. For two-way roads, the system treats each direction as a directed edge and additionally records the corresponding physical road during vehicle-to-vehicle conflict detection to identify conflicts on the opposite edges.

[0075] Aircraft occupancy time windows can be organized by node, edge, or conflict interface. For service road nodes adjacent to conflict interfaces, the system can form node blocking time windows; for scenarios where aircraft impact extends to service road edges, the system can form edge blocking time windows or edge capacity reduction time windows. Blocking time windows include start time, end time, associated aircraft identifier, and blocking object identifier. The path planning module treats these as inviolable hard safety constraints, distinct from the soft costs of risk fields.

[0076] In one embodiment, the input task includes 10 ground service vehicles, where small service vehicles have a capacity weight of 1.0 and large support vehicles have a capacity weight of 1.5 to 2.0. The planning time window is the next 15 minutes, and the risk field time bucket width is 30 seconds. The system reads the risk value of each edge and each node over the next 30 time buckets and uses it in low-level search and capacity safety verification.

[0077] 2. Construction of Risk Perception Cost The system first constructs a basic path cost, which includes vehicle travel time and waiting time. Then, edge risk cost and node risk cost are superimposed on the basic path cost. The edge risk cost reflects the density over-limit risk faced by vehicles occupying the edges of general service roads; the node risk cost reflects the node over-limit risk faced by vehicles occupying conflict points, intersections, waiting areas, or parking spaces.

[0078] In one embodiment, edge risk weight Node risk weight Waiting time weight The risk weight of nodes can also be increased based on airport operation strategies to enhance the tendency to avoid aircraft-vehicle conflict points.

[0079] In practical implementation, the system calculates the intersection of the continuous time interval during which a vehicle occupies an edge with the time bucket of the risk field, and accumulates the edge risk cost based on the intersection duration. If the vehicle starts from time... to If edge e is occupied, and the occupied interval spans multiple risk time buckets, then the system calculates the occupied duration for each time bucket separately. and will Add path cost. Node risk cost is similarly calculated and can be accumulated when a vehicle arrives, waits, turns, or a service operation occupies a node.

[0080] For example, vehicle i occupies the service roadside from 08:30:10 to 08:30:50. The risk time buckets are 08:30:00 to 08:30:30 and 08:30:30 to 08:31:00. If the edge risks of the two time buckets are 0.2 and 0.5 respectively, then the cost of the edge risk is... This calculation method correlates the risk cost with the vehicle's actual exposure time, rather than simply imposing a rough penalty based on whether or not the vehicle passes through a high-risk edge.

[0081] To address node risk, the system can assign higher weights to nodes adjacent to conflicting interfaces. For example, when node v is an adjacent node in an aircraft-vehicle conflicting interface, a higher weight is assigned. When the node is a regular intersection or a parking garage node, take... In this way, even if the risks are the same, vehicles will prioritize minimizing their time spent near aircraft conflict points. Optionally, the system can also reduce the risk weight or waiting weight for vehicles with higher mission priority to meet the needs of emergency support missions.

[0082] The waiting time in the basic path cost can be distinguished into active waiting and passive waiting. Active waiting refers to vehicles waiting in safe waiting areas or garage nodes to avoid aircraft hard blockages or high-risk periods; passive waiting refers to waiting caused by vehicle-to-vehicle conflicts, insufficient node capacity, or no feasible passage time window. The system can assign different weights to the two types of waiting according to airport operation strategies, thereby encouraging vehicles to wait in advance in safe areas instead of queuing after entering the vicinity of conflict areas.

[0083] 3. Low-level continuous-time search like Figure 2 As shown, the continuous-time conflict-priority search module performs a time window search on a single vehicle at a low level. For vehicle i, the system expands candidate states on the service road graph starting from the origin and release time. Each state includes the current node, current time, cumulative cost, and path predecessor.

[0084] When a node state is expanded, the system first checks the aircraft occupancy time window. If the vehicle arrival time falls within the aircraft blocking time interval associated with that node, the state is pruned; otherwise, the node risk cost and subsequent edge risk cost are calculated. Thus, the aircraft occupancy time window serves as a hard safety constraint, while the risk field, as a soft cost, influences the search priority.

[0085] In practical implementation, the low-level search can employ continuous-time A-satellite search, SIPP safe time interval search, or other time window search methods. The system first constructs a set of safe time intervals for each node, which is the available time interval obtained after deleting the aircraft blocking time window and the vehicle constraint time window passed from the high-level constraint tree from the planned time window. The low-level search state can be represented as (v, I, t, g, h), where v is the current node, I is the available time interval for the current node, t is the vehicle arrival time, g is the cumulative cost, and h is the heuristically estimated cost.

[0086] When a vehicle travels from node v to node w along edge e=(v,w), the system estimates the arrival time based on edge length, vehicle type, speed limit, and risk-adjusted travel time. If the arrival time falls within the safe time interval of w, a successor state is generated; if it is earlier than the start time of the safe time interval, the vehicle can wait at v or a permitted waiting position until an entry time is reached; if the arrival time falls within the aircraft hard blocking interval, the successor state is discarded. For edge blocking time windows, the system also checks whether the time interval during which the vehicle occupies the edge overlaps with the blocking time window.

[0087] In one embodiment, when a vehicle travels from the garage node to the parking space node, the shortest path requires passing through conflict nodes. .like If an aircraft congestion window exists between 08:31:00 and 08:31:40, then low-level searches will not generate data for aircraft arriving or occupying space within that window. The state of the alternative path. If the alternative path has no hard blockage but is riskier and more expensive, the search will compare the total cost of waiting to take the original path with the total cost of taking the alternative path, and output the feasible continuous-time path with the lower cost.

[0088] The low-level search can also record edge occupancy intervals and node occupancy intervals for high-level vehicle-to-vehicle conflict detection and subsequent capacity safety verification. For each path segment, the system saves the vehicle's entry time, exit time, entry time, and exit time, as well as the corresponding risk time bucket set. This record allows high-level modules to identify conflicts without re-parsing the path.

[0089] 4. High-level conflict detection and constraint addition like Figure 3 As shown, after all vehicles have generated their initial paths, the system detects conflicts between vehicles. If two vehicles occupy the same node within an overlapping time window, a vertex conflict occurs; if two vehicles travel in the same direction on the same service road and the safety interval is insufficient, a same-direction edge conflict occurs; if two vehicles travel towards each other on the same service road and their time intervals overlap, a reverse edge conflict occurs.

[0090] For any conflict, the system generates two sub-constraints, restricting one of the conflicting vehicles from passing through the corresponding location and time window, and adds the sub-constraints to the higher-level constraint tree. The system then re-searches the lower-level paths of the constrained vehicles until there are no vehicle-to-vehicle conflicts, or determines that the current task set is infeasible under the given constraints.

[0091] In practice, each node in the high-level constraint tree contains a set of vehicle constraints, the corresponding multi-vehicle path set, the total cost, and the conflict detection result. The root node does not contain vehicle-to-vehicle constraints. The system performs a low-level search for each vehicle to obtain the initial path. If there is a conflict in the initial path set, the system selects a conflict as a branch; for conflicting vehicles i and j, constraints "vehicle i must not occupy the conflict position within the conflict time window" and "vehicle j must not occupy the conflict position within the conflict time window" are generated respectively, forming two child nodes.

[0092] The constraint for vertex conflicts is the node and its time window. For example, vehicle i and vehicle j occupy node i from 08:32:05 to 08:32:20 and from 08:32:12 to 08:32:25, respectively. If the two vehicles occupy the same time period, the system will generate a vertex conflict time window from 08:32:12 to 08:32:20, and will prohibit one of the vehicles from occupying the same time window for each of the two child nodes. Conflicts along the same direction can be determined based on the time interval between vehicles entering the same side and the overtaking relationship on the edge. The constraints are the edge, direction, and time window. Conflicts along opposite sides are constrained by two opposite directed edges on the same physical road and their overlapping time windows.

[0093] For large vehicles or tractor-trailers, the system can extend the conflict detection time window or spatial safety interval. For example, the node occupancy time for large support vehicles can be extended by several seconds before and after the actual arrival and departure times, and reverse edge conflict detection can require a longer clearing time. This extension is achieved through vehicle type parameters without changing the basic structure of the high-level constraint tree.

[0094] In one embodiment, the high-level search prioritizes expanding the constraint tree node with the lowest total cost. The total cost can be the sum of the costs of all vehicle risk-aware paths, or it can include task lateness penalties or capacity verification penalties. If a constrained vehicle in a child node cannot find a low-level path that satisfies both hard safety constraints and vehicle constraints, that child node is pruned; if a new path is found, the system updates the path set and re-detects conflicts.

[0095] 5. Risk Adjustment Capacity and Security Validation like Figure 4 As shown, after obtaining the set of paths without vehicle conflicts, the system calculates the edge risk adjustment capacity and node risk adjustment capacity based on the risk field. For time buckets with higher risks, the available capacity of edges or nodes is reduced accordingly. The system calculates the sum of the vehicle capacity weights occupying a certain edge or node in each time bucket and verifies that it does not exceed the risk adjustment capacity.

[0096] In one embodiment, vehicle capacity weighting The threshold can be determined based on the vehicle type: 1.0 for small service vehicles and 1.5 to 2.0 for large support vehicles. The risk closure threshold can be set to 0.9; when the risk of a certain side or node exceeds this threshold, the system can mark that side or node as unavailable in the corresponding time bucket as an extreme risk protection mechanism in addition to the soft cost mechanism.

[0097] In practical implementation, the system first calculates the risk adjustment capacity of each edge and each node within each time bucket based on the risk field. and The higher the risk value, the lower the available capacity. Subsequently, the system iterates through the multi-vehicle path set, counts the set of vehicles occupying edge e and the set of vehicles occupying node v within each time bucket, and sums the vehicle capacity weights. If the sum of the weights exceeds the corresponding risk adjustment capacity, it indicates that although the path set has no hard vehicle-to-vehicle conflicts, it may still be over-concentrated at the macro-capacity level and requires further adjustment.

[0098] For example, on a certain side Original capacity If the edge risk within the time bucket from 08:33:00 to 08:33:30 is 0.5, then the risk adjustment capacity is... If two small service vehicles and one large support vehicle occupy an edge within a time bucket in the path set, their combined capacity weight is 1.0 + 1.0 + 1.5 = 3.5, exceeding the risk adjustment capacity of 2, then the capacity safety verification fails. The system can convert this capacity violation into a new time window constraint or cost penalty, causing some vehicles to detour or wait.

[0099] Capacity safety verification can be performed in two modes. The first is the post-hoc verification mode, which checks capacity safety after obtaining a set of conflict-free paths; if it is not satisfied, it returns to a higher-level search to add constraints. The second is the in-search verification mode, which pre-estimates the capacity occupancy of paths during low-level search or high-level node expansion, and prunes candidate paths that would obviously lead to capacity overruns as early as possible. For online planning scenarios, post-hoc verification can be prioritized to reduce search complexity; for high-density vehicle groups, in-search verification can be used to reduce the number of repeated replanning iterations.

[0100] Opportunity-calibrated safety verification is used to link the probabilistic implications of a risk field with capacity constraints. The system can assign risk budgets to different edges and nodes. and and based on the total risk budget Check if the path set meets the confidence requirements of the capacity constraint. If some high-risk time buckets consume too much risk budget, the system can increase the risk weight of the corresponding edge or node, or temporarily add it to the hard constraint set.

[0101] 6. Output Path and Evaluation Metrics The system outputs the continuous time path of each vehicle, including node access sequence, edge passage sequence, departure time, arrival time, waiting time, detour result, vehicle-to-vehicle conflict detection result, and capacity safety verification result.

[0102] The system can also output a node risk exposure index, NRE(p), to evaluate the risk-weighted dwell time of the path set at the aircraft-vehicle conflict point. In one embodiment, by using an over-limit risk field as a soft cost, the path set can reduce high-risk node occupancy and node risk exposure without significantly increasing the total passage cost.

[0103] In practical implementation, the path output module can generate an executable path table for each vehicle. This path table includes information such as path segment number, current location node, next node, used edges, entry edge time, departure edge time, node waiting time, estimated arrival time at the destination, and whether it passes through high-risk time buckets. For autonomous vehicles or dispatch terminals, the system can also output edge-level passage instructions and waiting instructions; for manual dispatch systems, the system can output suggestions on where vehicles should wait, when to release them, and whether detours are necessary.

[0104] The path set level output may include total travel time, total waiting time, total risk cost, node risk exposure index, edge risk exposure index, number of vehicle-to-vehicle conflicts, capacity verification results, and reasons for infeasibility. If planning fails, the system may output the hard blocking time window, vehicle constraints, or capacity overrun locations that cause infeasibility, so that dispatchers can adjust task release time, number of vehicles, or road control strategies.

[0105] In one embodiment, the node risk exposure index can be compared before and after path planning. Without using risk field soft cost, vehicles may choose the shortest path and wait near high-risk conflict nodes; with risk field soft cost, vehicles may wait in a safe waiting area in advance or choose a low-risk detour, thereby reducing NRE(p). This evaluation index can serve as a quantitative basis for airport operations management systems to assess the safety of path planning.

[0106] Optionally, the system can also perform rolling route planning. At the current moment, it issues the most recently executable route segment and retains subsequent routes as planned routes. When new aircraft occupancy time windows, risk field prediction results, or vehicle mission statuses change, the system replans the remaining routes using the vehicle's current position and time as a new starting point. Through rolling updates, the system can adapt to operational disturbances such as aircraft pushback delays, taxiing waits, vehicle mission changes, and temporary road closures.

[0107] Example 3 This invention provides a continuous-time path planning system for vehicle groups in an airfield based on risk field perception and hybrid hard and soft constraints. The system includes a vehicle task set input module, a service road network input module, an aircraft occupancy time window module, a spatiotemporal risk field input module, a risk cost construction module, a continuous-time conflict priority search module, a risk adjustment capacity calculation module, an opportunity calibration safety verification module, and a path output module. The system performs the following steps: Obtain task sets from multiple ground service vehicles through the vehicle task set input module. The task set includes vehicle index, vehicle origin, destination, release time, and deadline, and obtains a directed graph of service roads through the service road network input module. The directed graph of the service road includes a set of service road nodes, a set of edges, node coordinates, edge lengths, and edge travel times.

[0108] The aircraft occupancy time window module obtains the set of congestion time intervals of the aircraft at the conflict point, and the spatiotemporal risk field input module obtains the edge over-limit risk field. and node over-limit risk field ; Construct a risk perception path cost function using a risk cost construction module; The high-level constraint tree is initialized by the continuous time conflict priority search module, and a low-level time window search is performed for each vehicle to generate a continuous time path that satisfies the vehicle task constraints and the aircraft occupancy time window hard constraints. When the low-level search is in the unfolded state, the aircraft occupancy time window is used as a hard constraint for pruning, and the edge over-limit risk field and node over-limit risk field are added as soft costs to the path cost. Vehicle-to-vehicle conflict detection is performed on the generated multi-vehicle paths. If there are vertex conflicts, same-direction edge conflicts, or opposite-direction edge conflicts, vehicle-position-time window constraints are generated and added to the high-level constraint tree. Then, return to S4 to search for constrained vehicle paths again. If there is no vehicle-to-vehicle conflict, the risk adjustment capacity is calculated by the risk adjustment capacity calculation module based on the edge over-limit risk field and the node over-limit risk field. The path set is verified to meet capacity safety conditions by using the opportunity calibration safety verification module. The path output module outputs the set of executable continuous time paths for each vehicle.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A continuous-time path planning method for vehicle groups in a flight zone based on risk field perception and hybrid hard and soft constraints, characterized in that, include: Obtain directed graph data of the task set and service road network of ground support vehicles; Obtain the set of blocking time intervals and risk field data; the directed graph data includes road nodes and edges representing movement between nodes; A continuous time path is obtained by searching in the direction of minimizing the cost function based on the task set and directed graph data; the cost function is used to characterize the cost of the risk-aware path; hard constraints and soft cost constraints are applied during the search process; hard constraints are applied based on the set of blocking time intervals, and soft costs are calculated based on the risk field data and added to the cost function for soft cost constraints. Detect conflicts in continuous time paths, add conflict constraints based on the conflict detection results, and re-search based on the conflict constraints until the conflict detection is passed, thus obtaining an optimized continuous time path. Calculate the risk adjustment capacity based on vehicle path and risk field data; make a judgment based on the risk adjustment capacity; and generate an executable continuous time path set for each vehicle based on the judgment result.

2. The method according to claim 1, characterized in that, The cost function for: ， in, Basic path cost, Indicates a path. and These are the edge risk weights and the node risk weights, respectively. For vehicle i in the time bucket The duration of time that edge e is occupied. For vehicle i in the time bucket The duration of time that node v is occupied.

3. The method according to claim 1, characterized in that, In the hard constraint, when the time when a vehicle reaches the first node falls within the blocking time interval of the first node, the node-time state in the vehicle path corresponding to the vehicle is determined to be unavailable, and a pruning operation is performed. The first node is the node through which the aircraft passes.

4. The method according to claim 1, characterized in that, In soft cost constraints, during the search process, when a vehicle occupies a node or edge within a time bucket, the risk value data corresponding to the time bucket and the node or edge is added to the cost function; the time bucket is obtained by dividing the path planning time into fixed time widths.

5. The method according to claim 1, characterized in that, Conflict detection includes vertex conflict, same-direction edge conflict, and opposite-direction edge conflict. Vertex conflict occurs when two vehicles occupy the same node within an overlapping time window. Same-direction edge conflict occurs when two vehicles occupy the same side with insufficient safety distance. Opposite-direction edge conflict occurs when two vehicles occupy the same side in opposite directions with overlapping time intervals.

6. The method according to claim 1, characterized in that, Risk adjustment capacity includes side risk adjustment capacity and node risk adjustment capacity ,in: ， in, Let e ​​be the original capacity of edge e. Let v be the original capacity of node v. This indicates a risk field exceeding the limit. This indicates a risk field where nodes exceed limits.

7. The method according to claim 6, characterized in that, The process of determining the risk adjustment capacity includes: Does the set of vehicle paths satisfy the safety constraints? The safety constraints are: ， ， in, Let i be the capacity weight. For vehicle i in the time bucket The indicator variable for the inner occupancy edge e, For vehicle i in the time bucket The indicator variable for the internally occupied node v. Represents a set of nodes. Let T represent the set of edges, and T represent the planning time. Does the opportunity calibration safety guarantee meet the following requirements? All edge and node capacity constraints are satisfied. , Among them, total risk budget , , This represents the risk budget for edge-time buckets and node-time buckets.

8. The method according to claim 1, characterized in that, After generating the executable continuous time path set for each vehicle, the process also includes: evaluating the node risk exposure index of the executable continuous time path set.

9. A continuous-time path planning system for vehicle groups in a flight zone based on risk field perception and hybrid hard and soft constraints, characterized in that, Used to perform the method described in any one of claims 1-8.

Images