A method, apparatus, storage medium, and electronic equipment for aircraft route planning.

Abstract

This application relates to the field of flight control technology, specifically providing a method, apparatus, storage medium, and electronic device for aircraft route planning. The method may include: filtering candidate subgraphs that satisfy a flight mission from a pre-constructed route knowledge graph; wherein the candidate subgraph includes at least one corridor path; a corridor path represents a connected region; the route knowledge graph is constructed using multiple corridor paths in flight space as entities, along with the semantic relationships and entity feature vectors of the entities; calculating the relationship score of each corridor path; and determining the flight route for executing the flight mission from the at least one corridor path based on the relationship score of each corridor path. Embodiments of this application can enable efficient low-altitude route planning for aircraft.

Description

Technical Field

[0001] This application relates to the field of flight control technology, and more specifically, to a method, apparatus, storage medium, and electronic equipment for aircraft route planning. Background Technology

[0002] In recent years, aircraft (such as drones) have been increasingly used in urban inspection, logistics transportation, emergency rescue, and low-altitude mapping, significantly increasing the complexity and dynamism of the flight environment. To ensure the safe operation of aircraft in unknown or dynamic environments, route planning is of paramount importance.

[0003] Currently, traditional aircraft route planning methods typically use single-scale grid partitioning or graph search strategies to generate routes; however, single-scale grids are computationally expensive and lack detail in wide-area searches; and costly refinement is required in the detailed terminal region to plan a suitable route.

[0004] Therefore, how to provide a technical solution for an efficient aircraft route planning method has become an urgent technical problem to be solved. Summary of the Invention

[0005] The purpose of some embodiments of this application is to provide a method, apparatus, storage medium and electronic device for aircraft route planning. The technical solutions of the embodiments of this application can perform efficient route planning for aircraft with low planning cost and high accuracy.

[0006] In a first aspect, some embodiments of this application provide a method for aircraft route planning, comprising: filtering candidate subgraphs that satisfy a flight mission from a pre-constructed route knowledge graph; wherein the candidate subgraphs include at least one corridor path; a corridor path represents a connected region; the route knowledge graph is constructed using multiple corridor paths in the flight space as entities and the semantic relationships and entity feature vectors of the entities; calculating the relationship score of each corridor path in the at least one corridor path; and determining a flight route for performing the flight mission from the at least one corridor path based on the relationship score of each corridor path.

[0007] Some embodiments of this application select candidate subgraphs that meet the flight mission from a pre-constructed route knowledge graph in the flight space, score each corridor path in the candidate subgraphs to obtain a relationship score, and finally select the flight route based on the relationship score. The route knowledge graph in the embodiments of this application contains multiple corridor channels that the aircraft can choose from, which enables efficient route planning for the aircraft with low planning cost and high accuracy.

[0008] In some embodiments, the step of filtering candidate subgraphs that satisfy the flight mission from a pre-built route knowledge graph includes: obtaining the start and end points of the flight mission; when the start and end points exist in any corridor path in the route knowledge graph, using the any corridor path as the candidate subgraph; when the start and end points do not exist in any corridor path in the route knowledge graph, selecting the at least one corridor path based on the start and end points.

[0009] Some embodiments of this application determine at least one corridor path from the route knowledge graph by identifying the start and end points of the flight mission, providing an effective data foundation for the subsequent generation of flight routes.

[0010] In some embodiments, determining the flight route for performing the flight mission from the at least one corridor path based on the relationship score of each corridor path includes: confirming that the relationship score of any corridor path is greater than a score threshold, and then using the any corridor path as the flight route.

[0011] Some embodiments of this application achieve efficient flight route planning by using any corridor path whose relationship score is greater than a score threshold as a flight route.

[0012] In some embodiments, determining the flight route for performing the flight mission from the at least one corridor path based on the relationship score of each corridor path includes: selecting corridor paths with relationship scores greater than a score threshold as candidate corridor paths; determining a path selection strategy corresponding to the type of the flight mission; wherein the type includes inspection, delivery, or rescue; and selecting the flight route that matches the path selection strategy from the candidate corridor paths.

[0013] Some embodiments of this application select a flight path from at least one corridor path by selecting a candidate corridor path based on a path selection strategy, thereby ensuring the accuracy of the flight path while ensuring efficient planning.

[0014] In some embodiments, before selecting candidate subgraphs that meet the flight mission from the pre-constructed route knowledge graph, the method further includes: constructing a spatiotemporal risk field model corresponding to flight data in the flight space; wherein the flight data includes geographical environment data, meteorological data, and flight regulation data; the spatiotemporal risk field model characterizes the risk values ​​of different spatial points in the flight space; performing risk spatial grid layering on the spatiotemporal risk field model to obtain a multi-layer grid system; performing connectivity analysis on the grid cells in the multi-layer grid system to generate a candidate route set; wherein the candidate route set includes the multiple corridor paths; and constructing the route knowledge graph corresponding to the candidate route set.

[0015] Some embodiments of this application first construct a spatiotemporal risk field model, then layer it to obtain a multi-layer network system; then analyze the multi-layer network system to generate a candidate route set; finally, construct a route knowledge graph based on the candidate route set to provide effective data support for subsequent flight route planning.

[0016] In some embodiments, constructing a spatiotemporal risk field model corresponding to flight data in the flight space includes: acquiring the geographic environment risk, meteorological risk, and flight regulation risk corresponding to the geographic environment data, the meteorological data, and the flight regulation data, respectively; performing a weighted summation of the geographic environment risk, the meteorological risk, and the flight regulation risk to obtain a flight risk function; and normalizing the flight risk function to obtain the spatiotemporal risk field model.

[0017] Some embodiments of this application obtain a flight risk function by acquiring risk data corresponding to flight data and then performing a weighted summation on the data. After normalization, a spatiotemporal risk field model is obtained, thereby achieving unified mapping processing of multi-source data and providing support for subsequent grid partitioning.

[0018] In some embodiments, the step of performing risk space grid layering on the spatiotemporal risk field model to obtain a multi-layer grid system includes: after performing preliminary grid division on the spatiotemporal risk field model, obtaining the risk sampling value of each grid cell and the risk space gradient of adjacent grid cells; and determining the multi-layer grid system through the risk sampling value and the risk space gradient.

[0019] Some embodiments of this application first perform preliminary grid division on the spatiotemporal risk field model to obtain the risk sampling paper and risk spatial gradient of each grid cell, and then determine a multi-layer grid system based on the two, thereby achieving accurate grid division.

[0020] In some embodiments, the step of performing connectivity analysis on the grid cells in the multi-layer grid system to generate a candidate route set includes: selecting multiple grid cells with risk values ​​lower than a risk threshold from the multi-layer grid system; clustering the multiple grid cells to obtain multiple candidate connected regions; and performing corridor path planning based on the multiple candidate connected regions to obtain the candidate route set.

[0021] Some embodiments of this application obtain multiple candidate connected regions by performing low-risk screening and clustering of grid cells in a multi-layer grid system; finally, corridor path planning is performed based on these regions to obtain a set of candidate routes, thereby achieving accurate acquisition of effective routes.

[0022] In some embodiments, constructing the route knowledge graph corresponding to the candidate route set includes: obtaining the entity feature vector and semantic relationship of each corridor path in the candidate route set; wherein the semantic relationship includes geographical relationship, flight function relationship, altitude relationship, historical trajectory relationship and flight constraint relationship; using each corridor path as the entity, and combining the entity feature vector and the semantic relationship, the route knowledge graph is constructed.

[0023] Some embodiments of this application can accurately construct a route knowledge graph containing multiple types of relationships by using the entity feature vectors and semantic relationships of each corridor path in the candidate route set, thus providing support for efficient route planning in the future.

[0024] Secondly, some embodiments of this application provide an apparatus for aircraft route planning, comprising: a filtering module for filtering candidate subgraphs that satisfy a flight mission from a pre-constructed route knowledge graph; wherein the candidate subgraphs include at least one corridor path; a corridor path represents a connected region; the route knowledge graph is constructed using multiple corridor paths in the flight space as entities and the semantic relationships and entity feature vectors of the entities; a scoring module for calculating the relationship score of each corridor path in the at least one corridor path; and a planning module for determining a flight route for performing the flight mission from the at least one corridor path based on the relationship score of each corridor path.

[0025] Thirdly, some embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the method described in any embodiment of the first aspect.

[0026] Fourthly, some embodiments of this application provide an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, can implement the method as described in any embodiment of the first aspect.

[0027] Fifthly, some embodiments of this application provide a computer program product, the computer program product including a computer program, wherein the computer program, when executed by a processor, can implement the method described in any embodiment of the first aspect. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of some embodiments of this application, the accompanying drawings used in some embodiments of this application will be briefly described below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 Flowcharts of methods for constructing route knowledge graphs provided for some embodiments of this application;

[0030] Figure 2 A flowchart of a method for aircraft route planning is provided for some embodiments of this application;

[0031] Figure 3 A block diagram of an aircraft route planning device is provided for some embodiments of this application;

[0032] Figure 4 A schematic diagram of an electronic device provided for some embodiments of this application. Detailed Implementation

[0033] The technical solutions of some embodiments of this application will now be described with reference to the accompanying drawings.

[0034] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0035] In related technologies, with the large-scale operation of low-altitude aircraft, such as unmanned aerial vehicles (UAVs), in urban, rural, and mixed areas, the challenges of route planning are increasing. For example, complex regional risk environments (such as buildings, no-fly zones, and weather disturbances), diverse traffic flows, and frequent real-time events are common. Existing technologies using single-scale grid partitioning or graph search strategies to generate routes lack the support of a "knowledge structure" that integrates multimodal data (such as ground feature semantics, mission type, and wind field rheology), making it difficult to interpret and reuse routes during the planning process. Furthermore, existing methods involve significant computational costs during the search phase, requiring substantial time for route planning, compromising timeliness, and resulting in low efficiency.

[0036] In view of this, some embodiments of this application provide a method for aircraft route planning. This method can pre-construct a route knowledge graph for the flight space, and then select candidate subgraphs from it based on the flight mission. After scoring and filtering at least one corridor path in the candidate subgraphs, the flight route for executing the flight mission is obtained. The route knowledge graph provided in this application can be reused during route planning, reducing planning costs; it also reduces data search volume, improves the efficiency of route planning, and ensures planning accuracy.

[0037] As described above, the route knowledge graph forms the basis of route planning in this application; it can efficiently plan flight routes for aircraft missions. The following uses an unmanned aerial vehicle (UAV) as an example to illustrate the process of constructing the route knowledge graph and planning the aircraft's route. This process can be executed by the UAV's own processor or by the terminal through which the UAV communicates; specific limitations are not made here.

[0038] The following is in conjunction with the appendix Figure 1 The implementation process of constructing a route knowledge graph provided by some embodiments of this application is illustrated by way of example.

[0039] like Figure 1 As shown, some embodiments of this application provide a flowchart of a method for constructing a route knowledge graph. This method for constructing a route knowledge graph may include:

[0040] S110, Construct a spatiotemporal risk field model corresponding to the flight data of the flight space; wherein, the flight data includes geographical environment data, meteorological data and flight regulations data; the spatiotemporal risk field model characterizes the risk value of different spatial points in the flight space.

[0041] For example, in a specific embodiment of this application, multi-source heterogeneous data from geographical environment, meteorology, regulatory rules (as a specific example of flight regulation data), historical flight paths, and mission requests are uniformly mapped into a standardized spatial field model to obtain a spatiotemporal risk field model. Here, regulatory rules refer to the clauses stipulated in relevant regulations that require drones to follow for low-altitude flight.

[0042] In some embodiments of this application, S110 may include:

[0043] S111, respectively acquire the geographical environment risk, meteorological risk and flight regulation risk corresponding to the geographical environment data, the meteorological data and the flight regulation data.

[0044] For example, in a specific embodiment of this application, it is assumed that the entire low-altitude operating space (as a specific example of flight space) is defined as a three-dimensional continuous domain: ;in, x and y Geographic plane coordinates z Represents flight altitude. The spatial boundaries are determined by the city boundaries, topographic elevation, and flight altitude restrictions.

[0045] Geographic environmental data is represented as follows: ;in, B ={ b i} represents the set of all buildings within the airspace, and each building b i Has attributes , These represent the building's height, type, and density, respectively. To assemble in the no-fly zone, It is a function of terrain height. This is a set of height-restricted zones; each location ( x , y (Corresponding to the maximum allowable height) z max ( x , y ).

[0046] Meteorological data are represented as vector fields: ;in, For the wind field at point ( x , y , z The data includes the wind velocity vector at time t, where T represents the temperature field. These field data can originate from weather radar, numerical weather prediction, or edge sensor grids, and are all spatiotemporally distributed data that are updated over time.

[0047] Geographic environmental risks are determined through geographic environmental data. f geo ( x , y , z , t It is characterized by the following formula:

[0048]

[0049] in, d (· , b i () represents the minimum distance from a point to the surface of a building. This is the attenuation coefficient.

[0050] Meteorological risk is determined through meteorological data. f meteo ( x , y , z It is characterized by the following formula:

[0051]

[0052] in, and These are the weighting coefficients.

[0053] Flight regulations risk f reg ( x , y , z This is used to consider no-fly zones or height-restricted zones, and it is characterized by the following formula:

[0054]

[0055] S112, the geographical environment risk, the meteorological risk, and the flight regulation risk are weighted and summed to obtain the flight risk function.

[0056] For example, in a specific embodiment of this application, after obtaining the aforementioned risk data, a risk function in a continuous space is output. f risk ( x , y , z (As a specific example of a flight risk function) is used to quantitatively describe the flight risk at any point in space, and its specific formula is:

[0057]

[0058] in, Let i be the weighting coefficients, i = 1, 2, 3, which satisfy the following conditions: .

[0059] S113, The flight risk function is normalized to obtain the spatiotemporal risk field model.

[0060] For example, in a specific embodiment of this application, the risk function is normalized to obtain a spatiotemporal risk field model. The spatiotemporal risk field model is characterized by the following formula:

[0061]

[0062] in, f max and f min These represent the maximum and minimum risk values, respectively.

[0063] Normally, the time interval t is set in minutes (e.g., 5 minutes) to provide data support for subsequent adaptive grid generation.

[0064] S120, Perform risk space grid layering on the spatiotemporal risk field model to obtain a multi-layer grid system.

[0065] For example, in a specific embodiment of this application, based on a known spatiotemporal risk field model, a multi-scale hierarchical network system that adaptively adjusts resolution according to spatial risk gradient and geographical complexity is constructed (as a specific example of a multi-layer network system): G={G (1) G (2) , ..., G (L)}. Among them, G (1) The grid representing the coarsest scale primarily constitutes the global airspace routes and corridors (i.e., corridor paths); G (L) The grid represents the finest scale and is used to guide local obstacle avoidance and terminal routes; L is the total number of grid levels. This step aims to minimize computational load while ensuring the fidelity of risk distribution, providing a balanced spatial structure for subsequent topology corridor extraction.

[0066] This process mainly consists of initial grid generation and risk sampling, spatial refinement based on risk gradient and complexity, and dynamic updating based on the time-varying risk field (time adaptive). The process of obtaining the multi-scale hierarchical network system is described in detail below.

[0067] In some embodiments of this application, S120 may include:

[0068] S121, After performing preliminary grid division on the spatiotemporal risk field model, obtain the risk sampling value of each grid cell and the risk spatial gradient of adjacent grid cells.

[0069] For example, in a specific embodiment of this application, an initial time t0 is set, and a regular cubic mesh G with an initial level of 1 is defined. (1) (That is, the coarsest grid):

[0070]

[0071] Each grid cell Corresponding space voxel :

[0072]

[0073] At the center point of each grid cell Calculate its risk sample value :

[0074]

[0075] Define the risk space gradient between two adjacent grid cell spaces. for:

[0076]

[0077] S122, the multi-layer grid system is determined by the risk sampling value and the risk spatial gradient.

[0078] For example, in a specific embodiment of this application, spatial grid refinement and hierarchical construction are performed based on risk sampling values ​​and risk spatial gradients. Specifically, whether the spatial grid needs refinement is determined by setting indicators.

[0079] Risk gradient index S1: ;like This indicates that the risk has changed drastically and needs to be further refined. δ s For a predefined threshold.

[0080] Risk variance index S2: ;like This indicates that localized risk fluctuations are significant and require further analysis. For a predefined threshold, l This indicates the current network level.

[0081] The final refinement condition is: Refine if or .

[0082] If the mesh satisfies this refinement condition, then the mesh cell will be in the next level G. (l+1)It is subdivided into 2×2×2 sub-mesh elements. Then, based on the refinement conditions defined above, an iterative refinement process is performed on the initially divided mesh elements: G (l+1) =Refine(G (l) |criteria), until for any mesh cell All satisfied , Or the grid level has reached the preset maximum level threshold. L max Once the mesh refinement process has been iterated, each parent cell corresponds to 8 (2×2×2) child mesh cells; and the parent-child mapping table Φ is maintained in real time. (l→l+1) It is used to support route extraction and backtracking.

[0083] Since the risk field used in the flight space changes over time, the grid cells in the above multi-layer grid system need to be locally updated when the risk changes significantly (i.e., further divided into adaptive refinement and adaptive merging).

[0084] The criterion for adaptive refinement is the same as the refinement conditions mentioned above, namely, satisfying... or When the conditions are met, refinement is performed, and the number of layers in the refined mesh must not exceed [a certain threshold]. L max .

[0085] Adaptive merging is applied to 8 sub-mesh elements that originally belonged to the same parent element. If any sub-mesh element p satisfies the following condition, then the 8 sub-elements will be merged into one parent element:

[0086]

[0087] in, and And is the merging threshold, and , This ensures that units are dynamically merged when the risk is stable, achieving a two-way balance between spatial resolution and computing units.

[0088] At the same time, saving each grid cell It also records its attributes. :

[0089]

[0090] in, bbox represents the risk value at the center of the grid cell. i,j,k This represents the bounding box of the grid cell space. Meanwhile, at the () lWithin a layer of mesh, a mesh adjacency graph N is constructed based on the mesh adjacency relationships (i.e., two meshes are considered adjacent if they share a face or an edge). (l) The node in this diagram is the ( ) l The mesh of the layer is such that if the meshes of two corresponding nodes are spatially adjacent, then there is an edge between the two nodes.

[0091] S130, perform connectivity analysis on the grid cells in the multi-layer grid system to generate a candidate route set; wherein, the candidate route set includes the multiple corridor paths.

[0092] For example, in a specific embodiment of this application, after obtaining the aforementioned adaptive multi-scale hierarchical network system G and the time-risk field model, low-risk, well-connected airway corridors (i.e., corridor paths, or simply corridors) are extracted, and based on this, flight routes can be planned for subsequent actual flight missions. The corridor network C(t) output by this step (as a specific example of a subsequent route set) will provide a structured spatial channel for subsequent route planning.

[0093] In some embodiments of this application, S130 may include:

[0094] S131, Select multiple grid cells from the multi-layer grid system whose risk values ​​are lower than the risk threshold.

[0095] For example, in a specific embodiment of this application, a risk threshold is set. λ cand Select several grid cells with lower risk from a multi-scale hierarchical network architecture. The selected formula is:

[0096]

[0097] S132, cluster the multiple grid cells to obtain multiple candidate connected regions.

[0098] For example, in a specific embodiment of this application, through the adjacency graph N (l) Examine the adjacency relationships among the selected grid cells, cluster adjacent candidate grid cells to obtain several connected regions. R cand (t) (as a specific example of multiple candidate connected regions): R cand (t)={R1,R2,...,R K These regions represent potential "low-risk corridor candidate regions" at the current time t.

[0099] S133, based on the multiple candidate connected regions, perform corridor path planning to obtain the candidate route set.

[0100] For example, in a specific embodiment of this application, in order to make the corridor approximate a real and feasible flight area, it is necessary to introduce a small amount of historical flight track data H, where H={ π j ( t )| j =1,..., N hist}; among them, each historical trajectory π j ( t (a) is a continuous function (containing multiple trajectory points): .

[0101] Using historical trajectory points in H ( t j,k , x j,k , y j,k , z j,k Constructing the trajectory spatiotemporal density field h (( x , y , z ), t This represents the probability density of a location being used by a shipping route in history.

[0102]

[0103] Where M represents the total number of historical trajectories, K j Indicates the first j The number of trajectory points in the trajectory. j , k Indicates the first j The first of the trajectories k A trajectory point, and These represent the spatial and temporal smoothing coefficients, respectively.

[0104] Then, in each candidate connected region R k Internal local graph structure is established: G R =( V R , E R ); where, node V R The center and edge of each grid cell E RIt consists of adjacent units.

[0105] Define the corridor generation cost function for edges. c ab :

[0106]

[0107] in, Represents a node a The risk value, h ( a , t (This refers to historical density.) d ( a , t ) represents the distance between nodes. α , β , η Predefined weights for users are used to control the impact of risk, historical preferences, and distance on costs.

[0108] To ensure the generated corridor conforms to flight habits, from V R Select a set of "high-density core points" as the starting point for corridor generation: S k ={ v i ∈ V R | h ( v i , t )};in, v i for h Local maxima and h ( v i , t ) ≥ λ h , λ h The range ∈[0,1] is used to control the sparsity of the selected starting points (for example, only the top 5% of nodes with the highest density can be selected). These points represent high-frequency traffic seed regions and will be used for subsequent corridor extraction.

[0109] For each selected starting point v s ∈ S k Perform the following operations:

[0110] S1, Initialize the path set Pv s =[ v s], and set the current node as v cur = v s ;

[0111] S2, in v cur Adjacent node set NB ( v cur Select the node with the smallest cost increment from the given information.

[0112]

[0113] in, This represents a penalty for directional deviation (used to maintain a smooth path and prevent frequent turns):

[0114]

[0115] in, d prev Indicates the direction of the previous movement. d cur-j Indicates the current direction of movement. γ This is the directional smoothing coefficient.

[0116] S3, will v next Add path, Pv s ← Pv s ∪ v next ;

[0117] S4, Update v cur ← v next ;

[0118] The loop executes S2 to S4 repeatedly until any of the following termination conditions are met, at which point the loop ends:

[0119] 1) The risk cost exceeds the threshold: ;

[0120] 2) Low historical density: h ( v cur , t )< λ min ;

[0121] 3) No better neighbors (i.e., local convergence).

[0122] In this way, the path starts from a high-density point and "grows" in a direction with lower local risk and higher historical density until it naturally decays or encounters an obstacle.

[0123] Finally, for each starting point v s ∈ S k After executing S1~S4 above, a set of corridors can be obtained. P k : P k ={ Pv s | v s ∈ S k At this point, the region can be output. R k corridor collection P k Then, for each region k, the set of all region corridors P is obtained using the same principle described above: This forms multiple travel skeleton lines under non-task constraints. These skeletons are the basis of low-risk corridors at the current time t, and the set of all regional corridors constitutes multiple corridor paths in the candidate route set.

[0124] S140, construct the route knowledge graph corresponding to the candidate route set.

[0125] For example, in a specific embodiment of this application, a route knowledge graph is constructed using corridors from all the aforementioned regional corridor sets.

[0126] In some embodiments of this application, S140 may include:

[0127] S141, obtain the entity feature vector and semantic relationship of each corridor path in the candidate route set; wherein, the semantic relationship includes geographical relationship, flight function relationship, altitude relationship, historical trajectory relationship and flight constraint relationship.

[0128] For example, in a specific embodiment of this application, each corridor in the generated set of all regional corridors... P i As an entity in the route knowledge graph The entity feature vector of each corridor att i for:

[0129]

[0130] in, c i =(x i , y i , z i () is the center point of the corridor; This represents the risk value at that point. h This represents the historical flight density at that point; Embedding semantic attributes; p alt ( z i )= z i / z max It is highly normalized.

[0131] The set of semantic attributes of an entity is as follows:

[0132] in, Used to encode the type of corridor (e.g., "urban patrol corridor"); Used to encode areas of the corridor (e.g., "Northern Corridor"); Used to indicate whether the corridor is in a height-restricted area (e.g., "outside the height-restricted area").

[0133] In addition, the route knowledge graph also has the following semantic relationships (i.e., semantic relations):

[0134] R={ r geo , r alt , r func , r hist , r rule},in, r geo Geographic proximity (spatial distance within a threshold, i.e., geographic relationship); r alt The functions of the corridors are relatively consistent, such as "emergency takeoff" and "emergency cruise"; r func This refers to a height continuity relationship (two corridors are at the same or similar height levels, i.e., a height relationship). r hist This is a historical co-occurrence relationship (the two corridors have a high probability of being used consecutively in historical flight paths, i.e., a historical trajectory relationship). r rule For flight constraints (e.g., "mutual restricted zone boundaries", "adjacent no-fly zones").

[0135] S142, using each corridor path as the entity, and combining the entity feature vector and the semantic relationship, the route knowledge graph is constructed.

[0136] For example, in a specific embodiment of this application, a route knowledge graph is constructed based on the aforementioned entities, entity feature vectors, and semantic-level relationships. ,in, Att ={ att i} represents the set of entity feature vectors.

[0137] To ensure the effectiveness and timeliness of the route knowledge graph, this embodiment of the application can also use a relational graph convolutional neural network (R-GCN) to implement feature propagation and updating based on semantic relationships. Specifically, a 3-layer R-GCN is used to update the entity feature vectors of entities. l The entity embedding calculation for layer +1 is as follows:

[0138]

[0139] in, For the first l The entity embedding vector (i.e., the entity feature vector) of the layer. N r ( i ) indicates a relationship between entity r and r. E i Related neighboring entities, and These are the weight matrices, These are the normalization coefficients; σ ( () is the ReLU activation function. Initial layer input: .

[0140] To plan more accurate flight routes for subsequent flight missions, this application also proposes a mechanism for scoring the feasibility of passage between entities. Specifically, the feasibility score of passage between two entities i and j under each relationship can be calculated. :

[0141]

[0142] in, This indicates that "under the current environmental risks, mission semantics, and historical constraints, the two corridors..." E i and E j The probability of a safe and reasonable join under relation r; and FC( ) is a fully connected neural network, and the ReLU activation function is used to make the output value between 0 and 1.

[0143] For example, suppose there are two corridors, E i and E j . E i Designated as an "urban patrol corridor," with a height of 120m. E j This is the "airport perimeter corridor," at a height of 150m. Therefore, there are 5 possible relationships between them: r geo , r alt , r func , r hist , r rule After learning, R-GCN yielded the results shown in Table 1:

[0144] Table 1

[0145]

[0146] Based on the aforementioned relationship scores, path optimization can be performed for actual flight missions. The specific optimization process includes: defining drivability relationships. NR ( i , j , r )for:( i , j , r Passable ≥ λ r ;in, λ r A pre-defined threshold related to the relationship (e.g., 0.6). That is, ≥ λ r And if the geometric reachability of both is true, then for the entity E i and E j Under relation r, it is traversable. In other words, this "relation score" is used to measure "whether two corridors can be safely and reasonably connected into a route under a specific semantic relation". It is not a geometric "can be connected", but a semantic and environmental "should be connected" or "is it worth connecting".

[0147] After determining the traversability relationship, the comprehensive cost is defined as:

[0148]

[0149] Among them, weight 1, 2, 3. Adjustments can be made based on the actual task (e.g., adding risk weights in emergency tasks). 1).

[0150] After constructing the route knowledge graph as described above, flight mission path searches can be performed based on it. The following section, in conjunction with the appendix... Figure 2 The implementation process of aircraft route planning provided by some embodiments of this application is illustrated by way of example.

[0151] Please see the appendix Figure 2 , Figure 2 A flowchart of a method for aircraft route planning is provided for some embodiments of this application.

[0152] The above process is illustrated below by example.

[0153] S210, Filter candidate subgraphs that meet the flight mission from the pre-constructed route knowledge graph; wherein, the candidate subgraph includes at least one corridor path; a corridor path represents a connected region; the route knowledge graph is constructed using multiple corridor paths in the flight space as entities, as well as the semantic relationships and entity feature vectors of the entities.

[0154] For example, in a specific embodiment of this application, after obtaining the flight mission, a search is performed on the route knowledge graph constructed above to obtain candidate subgraphs.

[0155] In some embodiments of this application, S210 may include:

[0156] S211, Obtain the start and end points of the flight mission.

[0157] For example, in a specific embodiment of this application, a flight mission can be defined as:

[0158]

[0159] in, Q start =( x s , y s , z s () represents the starting point of the task; Q end =( x g , y g ,z g () represents the end point of the task; τ ∈ This represents the mission tag type (i.e., the type of flight mission, such as: inspection, delivery, rescue, etc.).

[0160] S212, when the starting point and the ending point exist in any corridor path in the route knowledge graph, the any corridor path is used as the candidate subgraph.

[0161] For example, in a specific embodiment of this application, if the starting point and the ending point are located in any corridor, they are directly mapped to the corresponding entity E. start and E end E start and E end Any corridor corresponding to it is a candidate subgraph.

[0162] S213, when the starting point and the ending point do not exist in any corridor path in the route knowledge graph, select the at least one corridor path based on the starting point and the ending point.

[0163] For example, in a specific embodiment of this application, if not in a corridor, then within a certain space... d anchor Select several nearest corridors (e.g.) E i and E j And add virtual entities. , And the corresponding connections link these corridors. This step ensures that the geographical location of the flight mission can be mapped to graph structure nodes on the semantic graph. Then, relation triples that satisfy the traversability criteria are selected from the route knowledge graph. i , j , r ), forming candidate subgraphs G can : G can =(V, E can )={( i , j , r ) |( i , j , r The candidate subgraph contains a series of connected corridors that are semantically reasonable, geometrically reachable, and permitted by regulations.

[0164] S220, calculate the relationship score for each corridor path in the at least one corridor path.

[0165] For example, in a specific embodiment of this application, the relationship score of each corridor is obtained through the above relationship score calculation formula (as a specific example of relationship score).

[0166] S230, based on the relationship score of each corridor path, determine the flight route for performing the flight mission from the at least one corridor path.

[0167] For example, in a specific embodiment of this application, the final flight route is selected from the relationship scores calculated above. If only one corridor satisfies both the start and end points of the flight mission, it can also be directly used as the flight route, i.e., any of the corridors mentioned above.

[0168] In some embodiments of this application, S230 may include: confirming that the relationship score of any corridor path is greater than a score threshold, and then using any corridor path as the flight route.

[0169] For example, in a specific embodiment of this application, if the relationship score ≥ λ r , λ r If the score threshold is used, then it is used as the flight path.

[0170] In some embodiments of this application, S230 may include: selecting corridor paths with relationship scores greater than a score threshold among the at least one corridor path as candidate corridor paths; determining a path selection strategy corresponding to the type of the flight mission; wherein the type includes inspection, delivery, or rescue; and selecting the flight route that matches the path selection strategy from the candidate corridor paths.

[0171] For example, in a specific embodiment of this application, after filtering candidate corridor paths by relation scores, it is necessary to determine the priority relation types based on the type of flight mission. r geo , r alt , r func , r hist , r rule Which relationships (as a specific example of a path selection strategy) are used to finally match the flight route? It's understandable that the correspondence between flight mission types and path selection strategies is pre-defined. That is, when multiple candidate corridors exist, the final flight route needs to be determined by comprehensively considering the relationship scores and the priority of the path selection strategy.

[0172] In other words, a semantic priority mechanism is introduced when searching on candidate subgraphs. The search algorithm will prioritize searching along high-confidence semantic relations (e.g., The path is expanded by considering the order of larger, more semantically relevant r values, thereby generating semantically coherent flight path links. Specifically, different relation types embody different semantic logics, as shown in Table 2:

[0173] Table 2

[0174]

[0175] During the actual search process, the system will determine the priority of each type of relationship (e.g., emergency tasks should be given priority). r func and r hist (As a concrete example of a path selection strategy); when expanding from each entity, first try high-scoring relationships (e.g. >0.8 and r takes precedence, i.e. λ r If these relationships are all infeasible, then the search is expanded along the suboptimal relationships. For example, the final search target is a path (i.e., a flight path), and π is:

[0176]

[0177] This path must ensure that semantic rationality, spatial feasibility, and regulatory constraints are met while minimizing risk and cost.

[0178] As a concrete example, suppose the flight mission is "urban emergency delivery," starting from the origin... Q start (Hospital) to the destination Q end (Accident location). In the route knowledge graph, the starting point is mapped to the corridor. E 2. Destination mapping to E 8; The results of calculating the relation score (i.e., relation value) using the R-GCN method described above:

[0179] Historically, it has often been used consecutively. Both belong to the emergency corridor; A series of results, including spatial proximity, were obtained. Then, considering that the flight mission was an emergency mission, the algorithm prioritized following... r func , r hist The relationships are expanded, and the risk costs are considered; this yields the semantically coherent path (i.e., flight route) with the minimum risk cost: Q start → E 2→E 3→ E 4→ E 8→ Q end The path is semantically represented as: "Following the historical main route → Entering the emergency function channel → A branch line approaching the spatial endpoint → The target point." The risk cost is derived according to the aforementioned comprehensive cost formula.

[0180] As can be seen from the embodiments described above, the flight route planning method based on route knowledge graph proposed in this application aims to achieve dynamic, safe, and semantically driven flight route generation in complex low-altitude environments. This method constructs a unified spatiotemporal risk field model by integrating multi-source data such as geographic terrain, meteorological elements, and flight rules to characterize the dynamic changes in environmental risks. Then, through a multi-scale adaptive grid partitioning mechanism, the spatial resolution is automatically adjusted according to the risk distribution, refining the model in high-risk or complex areas to achieve hierarchical modeling and efficient updating of the risk field. Subsequently, a track density field is constructed using the risk distribution and historical flight data, from which low-risk, high-accessibility potential corridor skeletons are extracted to form a candidate route set. Finally, a knowledge graph structure is introduced, abstracting corridors into semantic entities. A semantic route graph (i.e., a route knowledge graph) is constructed using multiple types of relationships, such as geographical adjacency, functional association, historical co-occurrence, and regulatory constraints. A graph neural network is then used to predict the semantic traversability between corridors. Combining task semantics and risk costs (i.e., comprehensive costs), a semantically prioritized intelligent search is performed to generate optimal routes that meet safety constraints, are semantically consistent, and are interpretable. This achieves a shift from geometric navigation to knowledge-driven intelligent route planning. The embodiments of this application can improve the timeliness and accuracy of flight route planning, have high practicality, and realize efficient, semantically rich, and interpretable low-altitude route planning.

[0181] Understandably, this application models the UAV corridor structure in the form of a knowledge graph, semantically expressing various relationships such as geographical adjacency, functional association, historical co-occurrence, and regulatory constraints. Through joint modeling of entity attributes and relationship types, it establishes an interpretable and scalable "semantic route knowledge layer," breaking through the traditional single-dimensional representation of routes that relies solely on geometric connections. It innovatively introduces a "semantic-first search" mechanism, integrating relationship scores, task semantics, and risk costs into a joint search criterion. This ensures that routes not only meet the requirements of risk minimization and geometric feasibility but also maintain semantic continuity and task consistency, thereby generating optimal routes that are "semantically coherent, regulatory compliant, and task-oriented."

[0182] Please refer to Figure 3 , Figure 3The diagram illustrates the composition of an aircraft route planning apparatus according to some embodiments of this application. It should be understood that this aircraft route planning apparatus corresponds to the method embodiments described above and is capable of performing the various steps involved in the method embodiments. The specific functions of this aircraft route planning apparatus can be found in the description above; detailed descriptions are omitted here to avoid repetition.

[0183] Figure 3 The aircraft route planning device includes at least one software functional module that can be stored in a memory or embedded in the aircraft route planning device in the form of software or firmware. The aircraft route planning device includes: a screening module 310, used to screen candidate subgraphs that meet the flight mission from a pre-built route knowledge graph; wherein, the candidate subgraph includes at least one corridor path; a corridor path represents a connected region; the route knowledge graph is constructed with multiple corridor paths in the flight space as entities and the semantic relationships and entity feature vectors of the entities; a scoring module 320, used to calculate the relationship score of each corridor path in the at least one corridor path; and a planning module 330, used to determine the flight route for performing the flight mission from the at least one corridor path according to the relationship score of each corridor path.

[0184] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the aforementioned method, and will not be elaborated further here.

[0185] Some embodiments of this application also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can perform the operation of any of the methods corresponding to the methods provided in the above embodiments.

[0186] Some embodiments of this application also provide a computer program product, which includes a computer program, wherein when the computer program is executed by a processor, it can perform the operation of any of the methods corresponding to the above embodiments provided in the above embodiments.

[0187] like Figure 4 As shown, some embodiments of this application provide an electronic device 400, which includes a memory 410, a processor 420, and a computer program stored in the memory 410 and executable on the processor 420. When the processor 420 reads the program from the memory 410 via a bus 430 and executes the program, it can implement the methods of any of the above embodiments.

[0188] Processor 420 can process digital signals and may include various computing architectures. For example, it may be a complex instruction set computer architecture, a reduced instruction set computer architecture, or an architecture that implements multiple instruction set combinations. In some examples, processor 420 may be a microprocessor.

[0189] Memory 410 can be used to store instructions executed by processor 420 or data related to the execution of instructions. These instructions and / or data may include code for implementing some or all of the functions of one or more modules described in the embodiments of this application. The processor 420 of this disclosure embodiment can be used to execute instructions in memory 410 to implement the methods shown above. Memory 410 includes dynamic random access memory, static random access memory, flash memory, optical memory, or other memories well known to those skilled in the art.

[0190] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0191] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0192] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for aircraft route planning, characterized in that, include: Candidate subgraphs that meet the flight mission requirements are selected from a pre-constructed route knowledge graph. Each candidate subgraph includes at least one corridor path; a corridor path represents a connected region. The route knowledge graph is constructed using multiple corridor paths in the flight space as entities, along with the semantic relationships and feature vectors of these entities. These multiple corridor paths are generated through connectivity analysis of grid cells in a multi-layered grid system obtained after layering the risk space grid of a spatiotemporal risk scenario model. The spatiotemporal risk field model represents the risk values ​​of different spatial points in the flight space. The multi-layered grid system is obtained as follows: after preliminary grid division of the spatiotemporal risk field model, the risk sampling value of each grid cell and the risk space gradient of adjacent grid cells are obtained; the multi-layered grid system is determined using the risk sampling value and the risk space gradient. Calculate the relationship score for each corridor path in the at least one corridor path; Based on the relationship score of each corridor path, a flight route for performing the flight mission is determined from the at least one corridor path.

2. The method as described in claim 1, characterized in that, The process of selecting candidate subgraphs that meet the flight mission requirements from a pre-built route knowledge graph includes: Obtain the start and end points of the flight mission; When the starting point and the ending point exist in any corridor path in the route knowledge graph, the any corridor path is used as the candidate subgraph; When the starting point and the ending point do not exist in any corridor path in the route knowledge graph, at least one corridor path is selected based on the starting point and the ending point.

3. The method as described in claim 2, characterized in that, Determining the flight route for the flight mission from the at least one corridor path based on the relationship score of each corridor path includes: If the relationship score of any corridor path is confirmed to be greater than the score threshold, then any corridor path is taken as the flight route.

4. The method as described in claim 1 or 2, characterized in that, Determining the flight route for the flight mission from the at least one corridor path based on the relationship score of each corridor path includes: The corridor paths whose relationship scores are greater than a score threshold in the at least one corridor path are selected as candidate corridor paths. Determine the path selection strategy corresponding to the type of flight mission; wherein, the type includes inspection, delivery, or rescue; Select the flight path that matches the path selection strategy from the candidate corridor paths.

5. The method according to any one of claims 1-3, characterized in that, Before filtering candidate subgraphs that satisfy the flight mission from the pre-built route knowledge graph, the method further includes: A spatiotemporal risk field model is constructed corresponding to the flight data of the flight space; wherein, the flight data includes geographical environment data, meteorological data, and flight regulation data; the spatiotemporal risk field model characterizes the risk value of different spatial points in the flight space; The spatiotemporal risk field model is subjected to risk space grid layering to obtain a multi-layer grid system; Connectivity analysis is performed on the grid cells in the multi-layered grid system to generate a candidate route set; wherein, the candidate route set includes the multiple corridor paths; Construct the route knowledge graph corresponding to the candidate route set.

6. The method as described in claim 5, characterized in that, The construction of the spatiotemporal risk field model corresponding to the flight data in the flight space includes: The geographical environment risk, meteorological risk, and flight regulation risk corresponding to the geographical environment data, the meteorological data, and the flight regulation data are respectively obtained; The flight risk function is obtained by weighted summation of the geographical environment risk, the meteorological risk, and the flight regulation risk. The flight risk function is normalized to obtain the spatiotemporal risk field model.

7. The method as described in claim 6, characterized in that, The process of performing connectivity analysis on the grid cells in the multi-layered grid system to generate a candidate route set includes: Select multiple grid cells with risk values ​​lower than the risk threshold from the multi-layer grid system; Clustering the multiple grid cells yields multiple candidate connected regions; Corridor path planning is performed based on the multiple candidate connected regions to obtain the candidate route set.

8. The method as described in claim 6, characterized in that, The construction of the route knowledge graph corresponding to the candidate route set includes: Obtain the entity feature vector and semantic relationship for each corridor path in the candidate route set; wherein, the semantic relationship includes geographical relationship, flight function relationship, altitude relationship, historical trajectory relationship and flight constraint relationship; Using each corridor path as the entity, and combining the entity feature vector and the semantic relationship, the route knowledge graph is constructed.

9. A device for aircraft route planning, characterized in that, include: A filtering module is used to filter candidate subgraphs that meet the flight mission requirements from a pre-constructed route knowledge graph. Each candidate subgraph includes at least one corridor path; a corridor path represents a connected region. The route knowledge graph is constructed using multiple corridor paths in the flight space as entities, along with the semantic relationships and feature vectors of these entities. These multiple corridor paths are generated through connectivity analysis of grid cells in a multi-layered grid system obtained after layering the spatiotemporal risk scenario model into a risk space grid. The spatiotemporal risk field model represents the risk values ​​of different spatial points in the flight space. The filtering module is further used to: after preliminary grid division of the spatiotemporal risk field model, obtain the risk sampling value of each grid cell and the risk space gradient of adjacent grid cells; and determine the multi-layered grid system using the risk sampling value and the risk space gradient. The scoring module is used to calculate the relationship score for each corridor path in the at least one corridor path; A planning module is used to determine the flight route for performing the flight mission from the at least one corridor path based on the relationship score of each corridor path.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program is executed by a processor to perform the method as described in any one of claims 1-8.

11. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and running on the processor, wherein the computer program is executed by the processor to perform the method as described in any one of claims 1-8.

Images