An unmanned aerial vehicle route planning method based on air corridors

Abstract

A method for unmanned aerial vehicle (UAV) route planning based on airways includes the following steps: acquiring the UAV airway, the UAV's takeoff point, and its destination; determining the entry and exit layers based on the UAV's takeoff point, destination, and airway; determining the access segment flight trajectory based on the takeoff point, entry / exit layers, and UAV airway entry point information; and determining the exit segment flight trajectory based on the UAV airway exit point information, entry / exit layers, and destination. This application proposes a bidirectional connection mechanism that connects the destination and the UAV airway, as well as the takeoff point and the UAV airway, through the entry and exit layers, greatly improving the safety and smoothness of UAV airway switching.

Description

Technical Field

[0001] This application relates to a method for unmanned aerial vehicle (UAV) route planning based on air routes. Background Technology

[0002] With the development of the low-altitude economy, drone operation scenarios are becoming increasingly complex (such as power line inspection, emergency rescue, and logistics delivery). The future low-altitude airspace will exhibit a hybrid form of trunk-line structured operations and terminal-level decentralization. Existing path planning technologies face the following challenges when dealing with this hybrid airspace: 1. Disconnect between path planning and airspace structure: Existing path planning technologies (such as general) The RRT algorithm is primarily designed for unstructured free-flight environments and has not yet established a collaborative mechanism with structured facilities such as aerial drone corridors. This means that existing technologies cannot perceive or utilize pre-defined corridor resources. In practical applications, this leads to a disconnect between drone flight trajectories and airspace management rules, making it impossible to achieve an orderly transition from terminal free flight to controlled mainline cruise, and thus failing to meet the standardized operation requirements of future low-altitude transportation networks.

[0003] 2. Lack of business acumen and simplistic resource allocation: Most current path planning algorithms treat all drones as homogeneous objects, using only the shortest total flight distance or lowest overall energy consumption as the sole optimization objective. This simplistic approach fails to recognize the differentiated needs for timeliness and safety in various business scenarios. For example, high-priority medical emergency missions may be hindered by a lack of dedicated routes, while routine missions with low timeliness requirements may occupy critical airway resources, leading to inefficient airspace resource allocation.

[0004] 3. Separation between path geometric optimality and dynamic executability: Existing path planning algorithms generally focus on geometric distance metrics in their cost function design, lacking explicit modeling of dynamic performance constraints such as UAV climb rate and turn rate. While the resulting paths may be geometrically feasible in a grid or continuous space, they do not necessarily meet the attitude change requirements of the aircraft in actual operation. In complex airspace or high-density route environments, such paths often involve excessive altitude jumps or sharp turns, requiring subsequent trajectory smoothing or secondary optimization before execution. This creates a disconnect between the path planning results and the flight control layer, weakening the global planning's ability to constrain flight safety and stability. Summary of the Invention

[0005] To address the aforementioned problems, this application proposes a UAV route planning method based on air routes, comprising the following steps: Obtain the drone's flight path, as well as its takeoff point and destination; The entry and exit layers are determined based on the drone's takeoff point, destination, and flight path; The access segment flight path is determined based on the takeoff point, entry / exit layer, and UAV flight path; the exit segment flight path is determined based on the UAV flight path, entry / exit layer, and destination.

[0006] Preferably, the entry / exit layer is a transitional airspace between the ground and the UAV flight path. After takeoff, the aircraft flies into the entry / exit layer according to the navigation route, completes attitude adjustment and flight path alignment in the entry / exit layer, and then enters the UAV flight path.

[0007] Preferably, the guidance strategy between the entry / exit layer and the UAV flight path is implemented as follows: The access segment flight path includes entry point P. entry And the entry / exit point J in ; Get entry point P entry The up and down directions of the mission are determined based on the projection index relationship between the origin and destination on the channel, and a set of candidate entry points is formed in that direction. Select the point closest to the starting point as the entry point P. entry ; Get the ingress point J of the ingress / exgress layer in In determining P entry Then, within the entry and exit layer, use P entry For reference, trace back N distances D along the flight path of the entry / exit layer, determine the value of N by the specified UAV entry angle, and obtain the merging starting point J on the entry / exit layer. in This is used to ensure that attitude adjustment and heading alignment are completed within the entry / exit layer before entering the main channel, and to ensure that a reasonable tangential angle is maintained when entering the channel; Controlling tangential entry: Generating the ascent from the takeoff point into the entry / exit layer and reaching J. in The transition trajectory, then J in Diagonal cut into P entry And entered the main channel cruise section; due to J in It is for P entry With the reverse reserved point, the drone can complete the access with a small entry angle and in P entry The nearby lane merge was completed and aligned with the drone's flight path; The exit segment flight path includes exit point P. exit and the terminal point J of the inlet and outlet layers out ; Control output: Set of exit points The UAV flight path configuration is pre-defined, and the exit point P closest to the destination in the direction of travel is selected from the candidate set according to the up and down directions. exit The drone departs diagonally from the main channel, enters the entry / exit layer, and merges into its destination J. outThis allows the drone to first enter the access layer obliquely from the main channel, and then complete attitude adjustment along the access layer before descending to its destination.

[0008] Preferably, the entry and exit layers are located between adjacent UAV flight paths.

[0009] Preferably, the entry / exit layer includes a main layer and a branch layer connecting the main layer and the adjacent UAV flight path; the branch layer is obliquely connected between the main layer and the UAV flight path, and a vertical flight path is provided on the branch layer; the branch layer also provides avoidance space when UAVs may encounter each other in the UAV flight path or the entry / exit layer; the entrance point P entry Exit point P exit It is located in the sub-layer area.

[0010] Preferably, the entry / exit layer is also used for switching between adjacent UAV flight paths.

[0011] Preferably, the access segment flight track and the outgoing segment flight track are configured as follows: Define the horizontal grid and adjacency expansion rules for each candidate horizontal node. Calculate terrain elevation based on DEM data The necessary flight altitude is determined by combining safety redundancy and linear reference altitude. ; Based on hard constraints such as no-fly zones and maximum altitude limits, feasibility of candidate nodes is assessed, and nodes that meet the no-fly requirements are selected and included in the search space. Calculate and accumulate node costs: update based on horizontal distance between neighboring nodes and dynamic performance penalty term. Updated via spherical distance According to the total cost Sort and maintain the open list, execute Node expansion and closed list updates; If a candidate expansion node enters the influence range of an obstacle, the corresponding APF potential field influence radius is... This triggers the APF local obstacle avoidance module. After it generates the obstacle avoidance connection segment and injects back the intermediate node, it continues to execute subsequent expansion. After the search reaches the target node, a global path is generated by backtracking, based on each node. Reconstruct the 3D waypoint sequence and proceed to the post-processing stage.

[0012] Preferred, The algorithm is used for 3D trajectory planning in the access and exit segments of the terminal free airspace. Its core objective is to generate stable trajectories that meet the dynamic performance requirements of UAVs under complex terrain, static obstacles, and no-fly zones. The algorithm employs a hierarchical architecture combining global discrete programming and local continuous optimization. The algorithm is used for global horizontal path search, and the artificial potential field (APF) is used as a local obstacle avoidance tool to generate continuous connected segments. The responsibilities of the two are strictly separated, and their core logic does not interfere with each other, ensuring the stability and interpretability of the algorithm. The algorithm constructs a three-dimensional flightable space through a state determination layer and a decision guidance layer: the state determination layer defines the flight feasibility of nodes, and the decision guidance layer selects the optimal path among feasible nodes; APF and By working together under preset trigger conditions, a complete chain of "global planning - local completion" is formed; Required flight altitude As a physical feasibility constraint, the flight altitude of each horizontal node is uniquely determined, and the calculation formula is as follows: ; in: The linear interpolation height from the start point to the end point, and the ideal reference trajectory height; Provides the current location terrain elevation for DEM data; To ensure safety margins, uncertainties such as DEM errors and airflow disturbances are integrated to guarantee ground safety. Once determined, it will serve as the core criterion for judging the feasibility of flight altitude; The algorithm is responsible for determining the feasible path connection relationship in the global discrete grid space. Its core function is to search for the optimal path to the target in the flyable space constructed by the state determination layer based on dynamic performance cost, and to independently complete the global planning based on its own cost system. The algorithm adopts a classic evaluation function structure, focusing on two core elements: path length and dynamic performance. The expression is: ; in: Heuristic cost: The spherical distance calculated using the Haversine formula provides a clear target orientation for the search, ensuring the optimality and efficiency of the algorithm. The cumulative cost, based on the horizontal path distance and superimposed with a dynamic performance penalty term, reflects the cost of path length and flight attitude. The iterative update formula is as follows: , The horizontal distance between adjacent nodes. This is the distance weighting coefficient. For nodes Dynamic performance penalty items; : These are the weighting coefficients for cumulative cost and heuristic cost, respectively, which adjust the global search bias; Dynamic performance penalty It consists of two parts: a climb rate penalty, based on neighboring nodes. The difference calculation constrains height changes that exceed the range of smooth ascent, and the turning rate penalty is calculated based on the angle between the lines connecting adjacent nodes to constrain sharp turning actions, thus achieving a balance between path length and dynamic executability.

[0013] Preferably, the artificial potential field (APF) is applied to local obstacle avoidance scenarios. Its core function is to construct a horizontal guiding force, generate continuous obstacle avoidance connection segments, and compensate for... The local path smoothness gap in global discrete search serves as a dedicated local optimization tool to support local path optimization. Potential field definition and mathematical model: Let there be a three-dimensional position vector. Indicates the current location of the drone. For fixed height Focusing on horizontal potential field calculations. For a point representing an obstacle or no-fly zone, the three-dimensional Euclidean distance is defined as: ; in These are the horizontal coordinates of the local tangent plane after latitude and longitude conversion; Gravitational potential and gravity: The gravitational source is a local target point, providing directionality for the obstacle-avoiding path and preventing path drift; let the three-dimensional coordinates of the local target point be... The gravitational potential and gravity are respectively: ; ; In the formula The gravitational gain coefficient controls the intensity of convergence towards the local target point; the height component is 0 to ensure that gravity acts only in the horizontal direction. Repulsive potential and repulsive force: These describe the repulsive effect of an obstacle on a horizontal path, when the distance is less than the radius of influence. When it takes effect, the repulsive potential and the repulsive force are respectively: ; ; The gradient term for: ; In the formula The repulsive gain coefficient controls the intensity of the obstacle's repulsive force and affects the magnitude of the local obstacle bypass path deviation. Total Resultant Force and Tangential Guidance: The total horizontal resultant force is the superposition of the horizontal components of the gravitational force and the repulsive forces from all obstacles, expressed as: ; Projecting the total resultant force onto the horizontal plane yields the horizontal guiding resultant force. This is used to drive local iterative updates; to reduce the risk of local minima, a tangential guiding term is superimposed on the resultant force. The direction is orthogonal to the horizontal component of the repulsive force from the obstacle, causing the trajectory to slide along the obstacle boundary. The expression is: ; in This is a 90° rotation transformation. This is the tangential guidance gain coefficient. Represents the cross product of two-dimensional vectors Component value; this item in < When activated, among which ; Ensure the dominant position of local target point orientation; The obstacle avoidance process includes: APF local obstacle avoidance is an online optimization process triggered on demand, with the following triggering conditions: Horizontal distance between candidate expansion nodes and obstacles Alternatively, one can enter the edge of the no-fly zone, with the following specific steps: Local target point selection: based on Extended direction unit vector Set look-ahead distance The coordinates of the local target point are , The horizontal coordinates of the current node; validation. The feasibility of a no-fly zone, along Directional shortening Until a feasible target point is obtained; Potential field iterative update: Assume the current node height that triggers obstacle avoidance is fixed at 1. The three-dimensional coordinates of the iteration point are Update the horizontal position according to the normalized step formula: ; Feasibility verification: Verify after each iteration. The three core constraints are: being outside the no-fly zone, meeting the minimum safe distance from obstacles, and controlling the horizontal offset within the local window boundary. Within a given timeframe, the next iteration can only proceed after all constraints are satisfied. Iteration Termination and Node Backhearsal: The iteration termination condition is reaching the local target point, with distance error ≤ Reaching the step limit Or the iteration point does not meet the constraints; adjust the continuous iteration point sequence according to the planning resolution. Resampling is used as an intermediate node, and the cumulative cost is from Trigger node Start accumulating , back to Open lists enable seamless integration with global paths.

[0014] Preferably, the post-processing stage includes track post-processing and three-dimensional trajectory reconstruction and altitude interpolation: The APF collaboration mechanism and the APF collaboration mechanism have a master-slave collaboration relationship: As the main body of global planning, it determines the overall path direction and node connection relationship; APF, as a subordinate local optimization, supplements continuous obstacle bypass segments in the obstacle neighborhood. Provides triggering conditions and local target pointing for APF; APF is... Optimize the smoothness of local paths to ensure a balance between global optimality and local executability; The post-processing of the track is as follows: After the global path is generated, post-processing steps are performed to optimize the trajectory morphology: through continuous horizontal correction, trajectory smoothing and equidistant resampling, the impact of the inflection point of the polyline on the dynamic control of the UAV is reduced, and a three-dimensional trajectory point sequence that meets the flight control requirements is output; this process optimizes the trajectory geometry while ensuring the feasibility of the path and the stability of the core cost features. : The weighting coefficients for cumulative cost and heuristic cost in the algorithm; : APF related gain coefficient, attraction, repulsion, tangential guidance; The radius of influence of the obstacle's potential field defines the local area of ​​influence around the obstacle. : APF iteration parameters, including maximum number of steps, step size, and maximum step speed; : APF local window parameters, look-ahead distance, maximum offset; 3D trajectory reconstruction and height interpolation are as follows: exist In the discrete paths obtained through the search, let the first... The latitude and longitude of each path point are: The corresponding cumulative mileage along the path is For each path point, calculate the linear reference height. Safety height relative to terrain The maximum value of the two is taken as the three-dimensional flight altitude of that point. ; In engineering implementation, this step adds a height dimension to the two-dimensional path point. The height calculation relies on the terrain elevation and safety redundancy at the point location to form a three-dimensional planning trajectory that meets safety constraints. Represent the path points as This will yield the final three-dimensional planning trajectory; Horizontal continuous correction, trajectory smoothing, and resampling; This step is in Execute after the search ends and the discrete path is obtained by backtracking; The output path is a discrete polyline, which is difficult to directly match the continuity constraints of the UAV's speed and turning. Therefore, geometrical continuity smoothing and equidistant resampling are performed on the polyline path to output an executable track point sequence with consistent point spacing. Horizontal continuous correction and trajectory smoothing focus on horizontal coordinate optimization, reducing inflection point angle changes and increasing obstacle boundary margins; maintaining the calculated safe height throughout the entire process. The feasibility assessment result for the no-fly zone remains unchanged; to ensure the established safety margin, an upper bound constraint is set on the smoothed horizontal offset. ; During the resampling stage, the smoothed path is sampled at fixed intervals to obtain a path point sequence with uniform point spacing. The corresponding mileage parameters and altitude information are retained or interpolated for each sampling point and used for subsequent distribution and spatiotemporal consistency control.

[0015] It also includes collaborative optimization strategies, such as dynamic matching of optimal entrances and exits, full-trajectory spatiotemporal resampling, collaboration between the potential field module and path search and trajectory optimization, and task attribute-driven potential field parameter mapping. Optimal dynamic matching of entrances and exits includes: We introduce a KD-Tree, a k-dimensional tree, a spatial index structure, to preprocess the discrete points along the entire waterway.

[0016] Using the nearest neighbor search algorithm of KD-Tree, to... It can quickly locate the candidate access / exit point closest to the start / end point with a time complexity of 0.

[0017] Combined with channel direction constraints, such as one-way traffic, it automatically locks the closest and legal optimal entry / exit pair.

[0018] Scalability support: This mechanism inherently supports batch concurrent processing of multiple target points. For complex tasks involving multiple consecutive work points, the algorithm can use KD-Tree to quickly match the optimal entry and exit ports for each target point, and seamlessly connect multiple access-cruise-exit sub-processes through dynamic programming to achieve multi-target coverage in a single sortie, significantly improving operational efficiency.

[0019] Full-trajectory spatiotemporal resampling includes: After the planning is completed, the path equal-interval resampling algorithm is used to resample the entire path from access to cruise to outgoing at equal intervals.

[0020] Effect: It eliminates the differences in point density generated at different planning stages, provides the flight control system with temporally and spatially consistent, smooth and continuous trajectory commands, and avoids flight oscillations caused by command jumps.

[0021] The synergy between the potential field module and path search and trajectory optimization includes: The APF-Map module provides basic terrain and obstacle data support for the state determination layer with a unified potential field model, provides local obstacle avoidance calculation basis for APF-Optimize, and provides risk assessment of continuous space in the post-processing stage.

[0022] The APF-Optimize module uses the same approach as... Synchronous local obstacle avoidance mechanism: If a candidate expansion direction is found to enter the neighborhood of a high-risk obstacle during the expansion process, a finite number of potential field-driven correction steps are performed within a local window, starting from the current search point and ending at a feasible look-ahead point in the original target direction, to generate a bypass connection segment that can be quickly deployed. This connection segment is then used as a new expansion candidate to continue participating in the process. search.

[0023] Task attribute-driven potential parameter mapping includes: Map task attributes to APF-related parameter combinations, including , , , , The terrain sensitivity coefficient and obstacle risk weight, among others, make different task types differ in terms of safety redundancy, detour tendency, and trajectory smoothness.

[0024] High-priority tasks can choose higher obstacle risk weights and stricter ground altitude redundancy while ensuring safety, while general tasks can use a more balanced combination of parameters to achieve a graded trade-off between timeliness and safety.

[0025] This application can bring the following beneficial effects: 1. This application proposes a two-way connection mechanism based on virtual tangential guidance points, which connects the destination and UAV flight path, as well as the take-off point and UAV flight path, through the entry and exit layers, greatly improving the safety and smoothness of UAV flight path switching.

[0026] 2. This application constructs virtual ramp buffer points and access offsets at the channel entrances and exits, decomposing the complex entry and exit maneuvers into two stages: obstacle avoidance and pathfinding, and tangential alignment. This effectively reduces abrupt changes in the heading angle at the merging point, achieves smooth tangential vector alignment between the flight trajectory and the channel centerline, ensures dynamic continuity during entry and exit, and greatly improves the safety and stability of UAV channel switching.

[0027] 3. This application addresses low-altitude complex terrain environments by constructing a two-layer potential field mechanism consisting of a state determination layer and a decision guidance layer. This mechanism is embedded into the APF-Optimize trajectory optimization process using the APF-Map module, while simultaneously providing data support for the state determination layer. The state determination layer is a three-dimensional flyable state mapping layer. Through safe altitude mapping, maximum ceiling constraints, and no-fly zone determination, it constructs a three-dimensional flyable space that varies with terrain undulations, strictly defining the flight feasibility of nodes and determining their existence. The APF-Map provides unified terrain and obstacle data mapping support for this layer. The decision guidance layer is a risk-guided decision-making layer. Only within the flyable space, it guides the APF-Optimize trajectory optimization and path post-processing process using the potential field gradient information provided by the APF-Map. The path selection and cost calculation are based on the UAV's dynamic performance (climb rate, turn rate) to complete the global search. This mechanism focuses the role of the potential field on path geometry optimization and online local correction, providing core support for APF-Optimize. While ensuring computational efficiency, it significantly enhances the UAV's robustness against airflow disturbances and positioning errors in complex terrain.

[0028] 4. This application proposes a rule engine that maps business attributes (job type, priority) to physical airspace resources (altitude layer, lateral position of airway). This enables refined management of airspace resources, ensuring rapid passage for high-priority tasks (such as emergencies) while meeting the operational needs of routine tasks (such as inspections), and resolving airspace conflicts and resource contention issues in multi-aircraft, multi-tasking scenarios.

[0029] 5. This application constructs a dynamic programming mechanism based on KD-Tree spatial indexing and a modular design with a three-segment architecture, supporting continuous operation of multiple target points in a single sortie. This mechanism breaks through the limitations of traditional point-to-point planning and can quickly generate optimal serial paths covering multiple operation points (such as continuous tower inspection and multi-point logistics delivery) with extremely low computational cost (O(KlogN)), significantly improving the efficiency of UAV operations and airway utilization in complex task scenarios. Attached Figure Description

[0030] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the process of this application.

[0031] Figure 2 This is a schematic diagram of the waterway for this application. Detailed Implementation

[0032] To clearly illustrate the technical features of this solution, the following detailed description, in conjunction with the accompanying drawings, will be provided.

[0033] A method for unmanned aerial vehicle (UAV) route planning based on airways, such as Figure 1 As shown, it includes the following steps: S1 obtains the drone's flight path, takeoff point, and destination; S2 determines the entry and exit layers based on the drone's takeoff point, destination, and flight path; like Figure 2 As shown, the physical airspace is modeled as a pipeline-like structure attached to the ground road network. To achieve secure isolation under multi-task concurrency, a rule engine is introduced to dynamically allocate airspace resources.

[0034] 1. Vertical Layer Mapping: The rule engine assigns drones to specific altitude layers based on task attributes (job type, drone speed), achieving vertical isolation of business logic. –Inspection: Assigned to the inspection operation layer (W-class airspace), usually located at the bottom layer, to facilitate ground observation.

[0035] –Logistics: Assigned to the logistics delivery layer (Class G airspace), located in the middle and upper levels, leveraging altitude advantage to ensure long-distance cruise efficiency.

[0036] –Emergency mission: Assigned to a dedicated emergency layer (W / G airspace) based on takeoff weight, enjoying the highest priority independent airspace to ensure rapid arrival without interference.

[0037] 2. Naming of airspace bearings, dual air routes, and up / down rules: Within the same altitude layer, to avoid opposing or intersecting conflicts, the algorithm uses the road centerline as a reference for spatial segmentation and flow management, and uniformly uses east, west, south, north, and up / down representations: – Navigation area naming: First, estimate the overall azimuth of the road centerline, then map the two sides of the centerline as the east and west navigation areas (when the overall direction is closer to north-south), or the north and south navigation areas (when the overall direction is closer to east-west). Lateral determination can be completed based on the cross product sign of the target position relative to a local line segment of the centerline, and the final output is unified as east-west and south-north navigation areas.

[0038] – Dual-route configuration: Two parallel routes are set up at the same altitude level in the navigation area on one side of the road, serving as the inner route and the outer route respectively, to enable two-way traffic on the same side.

[0039] – Upward and downward flow management: The rule engine binds the direction attributes of internal routes (closer to the road) and external routes on the same side of the road, so that opposite traffic in the same height layer runs on different routes, thereby avoiding head-on conflicts within a single layer.

[0040] This rules engine transforms complex mixed airspace missions into a layered, lane-based operational mode. The configuration can be arranged as described above in the altitude direction. The entry / exit layer serves as a transitional airspace between the ground and the UAV's flight path. After takeoff, the UAV flies into the entry / exit layer according to its pathfinding route, completes attitude adjustments and flight path alignment there, and then enters the UAV's flight path.

[0041] The entry and exit layers are located between adjacent UAV flight paths.

[0042] The entry / exit layer includes a main layer and a branch layer connecting the main layer and the adjacent UAV flight path; the branch layer is obliquely connected between the main layer and the UAV flight path, and a vertical flight path is provided on the branch layer. The branch layer also provides avoidance space when UAVs may encounter each other in the UAV flight path or the entry / exit layer; the entrance point P entry Exit point P exit It is located in the sub-layer area.

[0043] The entry and exit layers are also used for switching between adjacent UAV flight paths.

[0044] S3 determines the access segment flight path based on the takeoff point, entry / exit layer, and UAV flight path, and determines the exit segment flight path based on the UAV flight path, entry / exit layer, and destination.

[0045] The guidance strategy between the entry / exit layer and the UAV flight path shall be implemented as follows: The access segment flight track mentioned in S31 includes the entry point P. entry And the entry / exit point J in ; Get entry point P entry The up and down directions of the mission are determined based on the projection index relationship between the origin and destination on the channel, and a set of candidate entry points is formed in that direction. Select the point closest to the starting point as the entry point P. entry ; Get the ingress point J of the ingress / exgress layer in In determining P entry Then, within the entry and exit layer, use P entry For reference, trace back N distances D along the flight path of the entry / exit layer, determine the value of N by the specified UAV entry angle, and obtain the merging starting point J on the entry / exit layer. in This is used to ensure that attitude adjustment and heading alignment are completed within the entry / exit layer before entering the main channel, and to ensure that a reasonable tangential angle is maintained when entering the channel; Controlling tangential entry: Generating the ascent from the takeoff point into the entry / exit layer and reaching J. in The transition trajectory, then J in Diagonal cut into P entry And entered the main channel cruise section; due to J in It is for P entry With the reverse reserved point, the drone can complete the access with a small entry angle and in P entry The nearby lane merge was completed and aligned with the drone's flight path; The exit segment flight path described in S32 includes exit point P. exit and the terminal point J of the inlet and outlet layers out ; Control output: Set of exit points The UAV flight path configuration is pre-defined, and the exit point P closest to the destination in the direction of travel is selected from the candidate set according to the up and down directions. exit The drone departs diagonally from the main channel, enters the entry / exit layer, and merges into its destination J. out This allows the drone to first enter the access layer obliquely from the main channel, and then complete attitude adjustment along the access layer before descending to its destination.

[0046] The access segment flight track and the outgoing segment flight track are configured as follows: S301 sets the horizontal grid and adjacency expansion rules for each candidate horizontal node. Calculate terrain elevation based on DEM data The necessary flight altitude is determined by combining safety redundancy and linear reference altitude. ; Based on hard constraints such as no-fly zones and maximum altitude, S302 assesses the feasibility of candidate nodes, selects nodes that meet the no-fly requirements, and includes them in the search space. S303 calculates and accumulates node costs: updates based on horizontal distance between neighboring nodes and dynamic performance penalty terms. Updated via spherical distance According to the total cost Sort and maintain the open list, execute Node expansion and closed list updates; S304 If a candidate expansion node enters the influence range of an obstacle, the corresponding APF potential field influence radius is... This triggers the APF local obstacle avoidance module. After it generates the obstacle avoidance connection segment and injects back the intermediate node, it continues to execute subsequent expansion. After S305 reaches the target node, it backtracks to generate a global path, according to each node. Reconstruct the 3D waypoint sequence and proceed to the post-processing stage.

[0047] The algorithm is used for 3D trajectory planning in the access and exit segments of the terminal free airspace. Its core objective is to generate stable trajectories that meet the dynamic performance requirements of UAVs under complex terrain, static obstacles, and no-fly zones. The algorithm employs a hierarchical architecture combining global discrete programming and local continuous optimization. The algorithm is used for global horizontal path search, and the artificial potential field (APF) is used as a local obstacle avoidance tool to generate continuous connected segments. The responsibilities of the two are strictly separated, and their core logic does not interfere with each other, ensuring the stability and interpretability of the algorithm. The algorithm constructs a three-dimensional flightable space through a state determination layer and a decision guidance layer: the state determination layer defines the flight feasibility of nodes, and the decision guidance layer selects the optimal path among feasible nodes; APF and By working together under preset trigger conditions, a complete chain of "global planning - local completion" is formed; Required flight altitude As a physical feasibility constraint, the flight altitude of each horizontal node is uniquely determined, and the calculation formula is as follows: ; in: The linear interpolation height from the start point to the end point, and the ideal reference trajectory height; Provides the current location terrain elevation for DEM data; To ensure safety margins, uncertainties such as DEM errors and airflow disturbances are integrated to guarantee ground safety. Once determined, it will serve as the core criterion for judging the feasibility of flight altitude; The algorithm is responsible for determining the feasible path connection relationship in the global discrete grid space. Its core function is to search for the optimal path to the target in the flyable space constructed by the state determination layer based on dynamic performance cost, and to independently complete the global planning based on its own cost system. The algorithm adopts a classic evaluation function structure, focusing on two core elements: path length and dynamic performance. The expression is: ; in: Heuristic cost: The spherical distance calculated using the Haversine formula provides a clear target orientation for the search, ensuring the optimality and efficiency of the algorithm. The cumulative cost, based on the horizontal path distance and superimposed with a dynamic performance penalty term, reflects the cost of path length and flight attitude. The iterative update formula is as follows: , The horizontal distance between adjacent nodes. This is the distance weighting coefficient. For nodes Dynamic performance penalty items; : These are the weighting coefficients for cumulative cost and heuristic cost, respectively, which adjust the global search bias; Dynamic performance penalty It consists of two parts: a climb rate penalty, based on neighboring nodes. The difference calculation constrains height changes that exceed the range of smooth ascent, and the turning rate penalty is calculated based on the angle between the lines connecting adjacent nodes to constrain sharp turning actions, thus achieving a balance between path length and dynamic executability.

[0048] Artificial potential field (APF) is applied in local obstacle avoidance scenarios. Its core function is to construct a horizontal guiding force, generate continuous obstacle avoidance connection segments, and compensate for... The local path smoothness gap in global discrete search serves as a dedicated local optimization tool to support local path optimization. Potential field definition and mathematical model: Let there be a three-dimensional position vector. Indicates the current location of the drone. For fixed height Focusing on horizontal potential field calculations. For a point representing an obstacle or no-fly zone, the three-dimensional Euclidean distance is defined as: ; in These are the horizontal coordinates of the local tangent plane after latitude and longitude conversion; Gravitational potential and gravity: The gravitational source is a local target point, providing directionality for the obstacle-avoiding path and preventing path drift; let the three-dimensional coordinates of the local target point be... The gravitational potential and gravity are respectively: ; ; In the formula The gravitational gain coefficient controls the intensity of convergence towards the local target point; the height component is 0 to ensure that gravity acts only in the horizontal direction. Repulsive potential and repulsive force: These describe the repulsive effect of an obstacle on a horizontal path, when the distance is less than the radius of influence. When it takes effect, the repulsive potential and the repulsive force are respectively: ; ; The gradient term for: ; In the formula The repulsive gain coefficient controls the intensity of the obstacle's repulsive force and affects the magnitude of the local obstacle bypass path deviation. Total Resultant Force and Tangential Guidance: The total horizontal resultant force is the superposition of the horizontal components of the gravitational force and the repulsive forces from all obstacles, expressed as: ; Projecting the total resultant force onto the horizontal plane yields the horizontal guiding resultant force. This is used to drive local iterative updates; to reduce the risk of local minima, a tangential guiding term is superimposed on the resultant force. The direction is orthogonal to the horizontal component of the repulsive force from the obstacle, causing the trajectory to slide along the obstacle boundary. The expression is: ; in This is a 90° rotation transformation. This is the tangential guidance gain coefficient. Represents the cross product of two-dimensional vectors Component value; this item in < When activated, among which A preset small constant threshold is used to ensure the dominant position of the local target point's orientation. The obstacle avoidance process includes: APF local obstacle avoidance is an online optimization process triggered on demand, with the following triggering conditions: Horizontal distance between candidate expansion nodes and obstacles Alternatively, one can enter the edge of the no-fly zone, with the following specific steps: S41. Local target point selection: based on Extended direction unit vector Set look-ahead distance The coordinates of the local target point are , The horizontal coordinates of the current node; validation. The feasibility of a no-fly zone, along Directional shortening Until a feasible target point is obtained; S42. Potential Field Iterative Update: Assume the current node height that triggers obstacle avoidance is fixed at . The three-dimensional coordinates of the iteration point are Update the horizontal position according to the normalized step formula: ; S43. Feasibility Verification: Verify feasibility after each iteration. The three core constraints are: being outside the no-fly zone, meeting the minimum safe distance from obstacles, and controlling the horizontal offset within the local window boundary. Within a given timeframe, the next iteration can only proceed after all constraints are satisfied. S44. Iteration Termination and Node Backhearsal: The iteration termination condition is reaching the local target point, with distance error ≤ Reaching the step limit Or the iteration point does not meet the constraints; adjust the continuous iteration point sequence according to the planning resolution. Resampling is used as an intermediate node, and the cumulative cost is from Trigger node Start accumulating , back to Open lists enable seamless integration with global paths.

[0049] The post-processing stage includes track post-processing and 3D trajectory reconstruction and altitude interpolation: The APF collaboration mechanism and the APF collaboration mechanism have a master-slave collaboration relationship: As the main body of global planning, it determines the overall path direction and node connection relationship; APF, as a subordinate local optimization, supplements continuous obstacle bypass segments in the obstacle neighborhood. Provides triggering conditions and local target pointing for APF; APF is... Optimize the smoothness of local paths to ensure a balance between global optimality and local executability; The method for track post-processing is as follows: After the global path is generated, post-processing steps are performed to optimize the trajectory morphology: through continuous horizontal correction, trajectory smoothing and equidistant resampling, the impact of the inflection point of the polyline on the dynamic control of the UAV is reduced, and a three-dimensional trajectory point sequence that meets the flight control requirements is output; this process optimizes the trajectory geometry while ensuring the feasibility of the path and the stability of the core cost features. : The weighting coefficients for cumulative cost and heuristic cost in the algorithm; : APF related gain coefficient, attraction, repulsion, tangential guidance; The radius of influence of the obstacle's potential field defines the local area of ​​influence around the obstacle. : APF iteration parameters, including maximum number of steps, step size, and maximum step speed; : APF local window parameters, look-ahead distance, maximum offset; The method for 3D trajectory reconstruction and height interpolation is as follows: exist In the discrete paths obtained through the search, let the first... The latitude and longitude of each path point are: The corresponding cumulative mileage along the path is For each path point, calculate the linear reference height. Safety height relative to terrain The maximum value of the two is taken as the three-dimensional flight altitude of that point. ; In engineering implementation, this step adds a height dimension to the two-dimensional path point. The height calculation relies on the terrain elevation and safety redundancy at the point location to form a three-dimensional planning trajectory that meets safety constraints. Represent the path points as This will give you the final three-dimensional planning trajectory.

[0050] Horizontal continuous correction, trajectory smoothing and resampling This step is in Execute after the search ends and the discrete path is obtained by backtracking; The output path is a discrete polyline, which is difficult to directly match the continuity constraints of the UAV's speed and turning. Therefore, geometrical continuity smoothing and equidistant resampling are performed on the polyline path to output an executable track point sequence with consistent point spacing. Horizontal continuous correction and trajectory smoothing focus on horizontal coordinate optimization, reducing inflection point angle changes and increasing obstacle boundary margins; maintaining the calculated safe height throughout the entire process. The feasibility assessment result for the no-fly zone remains unchanged; to ensure the established safety margin, an upper bound constraint is set on the smoothed horizontal offset. ; During the resampling stage, the smoothed path is sampled at fixed intervals to obtain a path point sequence with uniform point spacing. The corresponding mileage parameters and altitude information are retained or interpolated for each sampling point for subsequent distribution and spatiotemporal consistency control. It also includes collaborative optimization strategies, such as dynamic matching of optimal entrances and exits, full-trajectory spatiotemporal resampling, collaboration between the potential field module and path search and trajectory optimization, and task attribute-driven potential field parameter mapping. The optimal dynamic matching of entrances and exits includes: introducing KD-Tree, k-dimensional tree, spatial index structure, and preprocessing discrete points along the entire waterway; Using the nearest neighbor search algorithm of KD-Tree, to... With reduced time complexity, it quickly locates the nearest candidate access / exit point to the start / end point; Combined with channel direction constraints, such as one-way direction, automatically lock the closest and legal optimal entry / exit pair; Scalability support: The mechanism naturally supports batch concurrent processing of multiple target points; for complex tasks involving multiple continuous operation points, the algorithm can use KD-Tree to quickly match the optimal entry and exit ports for each target point, and seamlessly connect multiple access-cruise-exit sub-processes through dynamic programming to achieve multi-target coverage in a single sortie, significantly improving operational efficiency. Full-trajectory spatiotemporal resampling includes: After the planning is completed, the path equal-interval resampling algorithm is used to resample the entire path from access to cruise to outgoing at equal intervals; Effect: It eliminates the differences in point density caused by different planning stages, provides the flight control system with temporally and spatially consistent, smooth and continuous trajectory commands, and avoids flight oscillations caused by command jumps; The synergy between the potential field module and path search and trajectory optimization includes: The APF-Map module provides basic terrain and obstacle data support for the state determination layer with a unified potential field model, provides local obstacle avoidance calculation basis for APF-Optimize, and provides continuous space risk assessment in the post-processing stage. The APF-Optimize module uses the same approach as... Synchronous local obstacle avoidance mechanism: If a candidate expansion direction is found to enter the neighborhood of a high-risk obstacle during the expansion process, a finite number of potential field-driven correction steps are performed within a local window, starting from the current search point and ending at a feasible look-ahead point in the original target direction, to generate a bypass connection segment that can be quickly deployed. This connection segment is then used as a new expansion candidate to continue participating in the process. search; Task attribute-driven potential parameter mapping includes: Map task attributes to APF-related parameter combinations, including , , , , Terrain sensitivity coefficient and obstacle risk weight make different task types show differences in safety redundancy, detour tendency and trajectory smoothness; High-priority tasks can choose higher obstacle risk weights and stricter ground altitude redundancy while ensuring safety, while general tasks can use a more balanced combination of parameters to achieve a graded trade-off between timeliness and safety.

[0051] The above are merely embodiments of this application and are not intended to limit this application.

[0052] 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 the claims of this application.

Claims

1. A method for unmanned aerial vehicle (UAV) route planning based on air routes, characterized in that: Includes the following steps: Obtain the drone's flight path, as well as its takeoff point and destination; The entry and exit layers are determined based on the drone's takeoff point, destination, and flight path; The access segment flight path is determined based on the takeoff point, entry / exit layer, and UAV flight path; the exit segment flight path is determined based on the UAV flight path, entry / exit layer, and destination.

2. The UAV route planning method based on airways according to claim 1, characterized in that: The entry / exit layer is a transitional airspace between the ground and the UAV flight path. After takeoff, the aircraft flies into the entry / exit layer according to the pathfinding route, completes attitude adjustment and flight path alignment in the entry / exit layer, and then enters the UAV flight path.

3. The UAV route planning method based on airways according to claim 2, characterized in that: The guidance strategy between the entry / exit layer and the UAV flight path shall be implemented as follows: The access segment flight path includes entry point P. entry And the entry / exit point J in ; Get entry point P entry The up and down directions of the mission are determined based on the projection index relationship between the origin and destination on the channel, and a set of candidate entry points is formed in that direction. Select the point closest to the starting point as the entry point P. entry ; Get the ingress point J of the ingress / exgress layer in In determining P entry Then, within the entry and exit layer, use P entry For reference, trace back N distances D along the flight path of the entry / exit layer, determine the value of N by the specified UAV entry angle, and obtain the merging starting point J on the entry / exit layer. in This is used to ensure that attitude adjustment and heading alignment are completed within the entry / exit layer before entering the main channel, and to ensure that a reasonable tangential angle is maintained when entering the channel; Controlling tangential entry: Generating the ascent from the takeoff point into the entry / exit layer and reaching J. in The transition trajectory, then J in Diagonal cut into P entry And entered the main channel cruise section; due to J in It is for P entry With the reverse reserved point, the drone can complete the access with a small entry angle and in P entry The nearby lane merge was completed and aligned with the drone's flight path; The exit segment flight path includes exit point P. exit and the terminal point J of the inlet and outlet layers out ; Control output: Set of exit points The UAV flight path configuration is pre-defined, and the exit point P closest to the destination in the direction of travel is selected from the candidate set according to the up and down directions. exit The drone departs diagonally from the main channel, enters the entry / exit layer, and merges into its destination J. out This allows the drone to first enter the access layer obliquely from the main channel, and then complete attitude adjustment along the access layer before descending to its destination.

4. The UAV route planning method based on airways according to claim 3, characterized in that: The entry and exit layers are located between adjacent UAV flight paths.

5. The UAV route planning method based on airways according to claim 4, characterized in that: The entry / exit layer includes a main layer and a branch layer connecting the main layer and the adjacent UAV flight path; the branch layer is obliquely connected between the main layer and the UAV flight path, and a vertical flight path is provided on the branch layer. The branch layer also provides avoidance space when UAVs may encounter each other in the UAV flight path or the entry / exit layer; the entrance point P entry Exit point P exit It is located in the sub-layer area.

6. The UAV route planning method based on airways according to claim 5, characterized in that: The entry and exit layers are also used for switching between adjacent UAV flight paths.

7. The UAV route planning method based on airways according to claim 1, characterized in that: The access segment flight track and the outgoing segment flight track are configured as follows: Define the horizontal grid and adjacency expansion rules for each candidate horizontal node. Calculate terrain elevation based on DEM data The necessary flight altitude is determined by combining safety redundancy and linear reference altitude. ; Based on hard constraints, such as no-fly zones and maximum altitude limits, feasibility of candidate nodes is assessed, and nodes that meet the no-fly requirements are selected and included in the search space. Calculate and accumulate node costs: update based on horizontal distance between neighboring nodes and dynamic performance penalty term. Updated via spherical distance According to the total cost Sort and maintain the open list, execute Node expansion and closed list updates; If a candidate expansion node enters the influence range of an obstacle, the corresponding APF potential field influence radius is... This triggers the APF local obstacle avoidance module. After it generates the obstacle avoidance connection segment and injects back the intermediate node, it continues to execute subsequent expansion. After the search reaches the target node, a global path is generated by backtracking, based on each node. Reconstruct the 3D waypoint sequence and proceed to the post-processing stage.

8. The UAV route planning method based on airways according to claim 7, characterized in that: The algorithm is used for 3D trajectory planning in the access and exit segments of the terminal free airspace. It constructs a 3D flyable space through a state determination layer and a decision guidance layer: the state determination layer defines the flight feasibility of nodes, and the decision guidance layer selects the optimal path among feasible nodes; APF and By working together under preset trigger conditions, a complete "global planning - local completion" chain is formed; Required flight altitude As a physical feasibility constraint, the flight altitude of each horizontal node is uniquely determined, and the calculation formula is as follows: ； in: The linear interpolation height from the start point to the end point, and the ideal reference trajectory height; Provides the current location terrain elevation for DEM data; To ensure safety margins, DEM errors and airflow disturbance uncertainties are integrated to guarantee ground safety. Once determined, it will serve as the core criterion for judging the feasibility of flight altitude; The algorithm is responsible for determining the feasible path connection relationship in the global discrete grid space. Its core function is to search for the optimal path to the target in the flyable space constructed by the state determination layer based on dynamic performance cost, and to independently complete the global planning based on its own cost system. The algorithm expression is: ; in: Heuristic cost: The spherical distance calculated using the Haversine formula provides a clear target orientation for the search, ensuring the optimality and efficiency of the algorithm. The cumulative cost, based on the horizontal path distance and superimposed with a dynamic performance penalty term, reflects the cost of path length and flight attitude. The iterative update formula is as follows: , The horizontal distance between adjacent nodes. This is the distance weighting coefficient. For nodes Dynamic performance penalty items; : These are the weighting coefficients for cumulative cost and heuristic cost, respectively, which adjust the global search bias; Dynamic performance penalty It consists of two parts: a climb rate penalty, based on neighboring nodes. The difference calculation constrains height changes that exceed the range of smooth ascent, and the turning rate penalty is calculated based on the angle between the lines connecting adjacent nodes to constrain sharp turning actions, thus achieving a balance between path length and dynamic executability.

9. The UAV route planning method based on airways according to claim 8, characterized in that: Artificial potential field (APF) is applied in local obstacle avoidance scenarios. Its core function is to construct a horizontal guiding force, generate continuous obstacle avoidance connection segments, and compensate for... The local path smoothness gap in global discrete search serves as a dedicated local optimization tool to support local path optimization. Potential field definition and mathematical model: Let there be a three-dimensional position vector. Indicates the current location of the drone. For fixed height Focusing on horizontal potential field calculations. For a point representing an obstacle or no-fly zone, the three-dimensional Euclidean distance is defined as: ; in These are the horizontal coordinates of the local tangent plane after latitude and longitude conversion; Gravitational potential and gravity: The gravitational source is a local target point, providing directionality for the obstacle-avoiding path and preventing path drift; let the three-dimensional coordinates of the local target point be... The gravitational potential and gravity are respectively: ; ; In the formula This is the gravitational gain coefficient, which controls the intensity of convergence towards the local target point; The height component is 0, ensuring that gravity acts only in the horizontal direction; Repulsive potential and repulsive force: These describe the repulsive effect of an obstacle on a horizontal path, when the distance is less than the radius of influence. When it takes effect, the repulsive potential and the repulsive force are respectively: ; ; The gradient term for: ; In the formula The repulsive gain coefficient controls the intensity of the obstacle's repulsive force and affects the magnitude of the local obstacle bypass path deviation. Total Resultant Force and Tangential Guidance: The total horizontal resultant force is the superposition of the horizontal components of the gravitational force and the repulsive forces from all obstacles, expressed as: ； Projecting the total resultant force onto the horizontal plane yields the horizontal guiding resultant force. This is used to drive local iterative updates; to reduce the risk of local minima, a tangential guiding term is superimposed on the resultant force. The direction is orthogonal to the horizontal component of the repulsive force from the obstacle, causing the trajectory to slide along the obstacle boundary. The expression is: ； in This is a 90° rotation transformation. This is the tangential guidance gain coefficient. Represents the cross product of two-dimensional vectors Component value; this item in < When activated, among which A preset small constant threshold is used to ensure the dominant position of the local target point's orientation. The obstacle avoidance process includes: APF local obstacle avoidance is an online optimization process triggered on demand, with the following triggering conditions: Horizontal distance between candidate expansion nodes and obstacles Alternatively, one can enter the edge of the no-fly zone, with the following specific steps: Local target point selection: based on Extended direction unit vector Set look-ahead distance The coordinates of the local target point are , The horizontal coordinates of the current node; validation. The feasibility of a no-fly zone, along Directional shortening Until a feasible target point is obtained; Potential field iterative update: Assume the current node height that triggers obstacle avoidance is fixed at 1. The three-dimensional coordinates of the iteration point are Update the horizontal position according to the normalized step formula: ； Feasibility verification: Verify after each iteration. The three core constraints are: being outside the no-fly zone, meeting the minimum safe distance from obstacles, and controlling the horizontal offset within the local window boundary. Within a given timeframe, the next iteration can only proceed after all constraints are satisfied. Iteration Termination and Node Backhearsal: The iteration termination condition is reaching the local target point, with distance error ≤ Reaching the step limit Or the iteration point does not meet the constraints; adjust the continuous iteration point sequence according to the planning resolution. Resampling is used as an intermediate node, and the cumulative cost is from Trigger node Start accumulating , back to Open lists enable seamless integration with global paths.

10. The UAV route planning method based on airways according to claim 7, characterized in that: The post-processing stage includes track post-processing and 3D trajectory reconstruction and altitude interpolation: The APF collaboration mechanism and the APF collaboration mechanism have a master-slave collaboration relationship: As the main body of global planning, it determines the overall path direction and node connection relationship; APF, as a subordinate local optimization, supplements continuous obstacle bypass segments in the obstacle neighborhood. Provides triggering conditions and local target pointing for APF; APF is... Optimize the smoothness of local paths to ensure a balance between global optimality and local executability; Track post-processing includes the following steps: After the global path is generated, post-processing steps are performed to optimize the trajectory morphology: through continuous horizontal correction, trajectory smoothing and equidistant resampling, the impact of the inflection point of the polyline on the dynamic control of the UAV is reduced, and a three-dimensional trajectory point sequence that meets the flight control requirements is output; this process optimizes the trajectory geometry while ensuring the feasibility of the path and the stability of the core cost features. 3D trajectory reconstruction and height interpolation include the following steps: exist In the discrete paths obtained through the search, let the first... The latitude and longitude of each path point are: The corresponding cumulative mileage along the path is For each path point, calculate the linear reference height. Safety height relative to terrain The maximum value of the two is taken as the three-dimensional flight altitude of that point. ; In engineering implementation, this step adds a height dimension to the two-dimensional path point. The height calculation relies on the terrain elevation and safety redundancy at the point location to form a three-dimensional planning trajectory that meets safety constraints. Represent the path points as This will yield the final three-dimensional planning trajectory; Horizontal continuous correction, trajectory smoothing, and resampling; This step is in Execute after the search ends and the discrete path is obtained by backtracking; The output path is a discrete polyline, which is difficult to directly match the continuity constraints of the UAV's speed and turning. Therefore, geometrical continuity smoothing and equidistant resampling are performed on the polyline path to output an executable track point sequence with consistent point spacing. Horizontal continuous correction and trajectory smoothing focus on horizontal coordinate optimization, reducing inflection point angle changes and increasing obstacle boundary margins; maintaining the calculated safe height throughout the entire process. The results of the feasibility assessment for the no-fly zone remain unchanged; To ensure a predetermined safety margin, an upper bound constraint is set on the smoothed horizontal offset. ; During the resampling stage, the smoothed path is sampled at fixed intervals to obtain a path point sequence with uniform point spacing. The corresponding mileage parameters and altitude information are retained or interpolated for each sampling point for subsequent distribution and spatiotemporal consistency control. It also includes collaborative optimization strategies, such as dynamic matching of optimal entrances and exits, full-trajectory spatiotemporal resampling, collaboration between the potential field module and path search and trajectory optimization, and task attribute-driven potential field parameter mapping. Optimal dynamic matching of entrances and exits includes: A KD-Tree spatial indexing structure is introduced to preprocess the discrete points along the entire waterway; Using the nearest neighbor search algorithm of KD-Tree, With reduced time complexity, it quickly locates the nearest candidate access / exit point to the start / end point; Combined with channel direction constraints, such as one-way direction, automatically lock the closest and legal optimal entry / exit pair; Scalability support: This mechanism naturally supports batch concurrent processing of multiple target points; for complex tasks involving multiple continuous operation points, KD-Tree can be used to quickly match the optimal entry and exit ports for each target point, and multiple access-cruise-exit sub-processes can be seamlessly connected through dynamic planning to achieve multi-target coverage in a single sortie, significantly improving operational efficiency. Full-trajectory spatiotemporal resampling includes: After the planning is completed, the path equal-interval resampling algorithm is used to resample the entire path from access to cruise to outgoing at equal intervals; Effect: It eliminates the differences in point density caused by different planning stages, provides the flight control system with temporally and spatially consistent, smooth and continuous trajectory commands, and avoids flight oscillations caused by command jumps; The synergy between the potential field module and path search and trajectory optimization includes: The APF-Map module provides basic terrain and obstacle data support for the state determination layer with a unified potential field model, provides local obstacle avoidance calculation basis for APF-Optimize, and provides continuous space risk assessment in the post-processing stage. The APF-Optimize module uses the same approach as... Synchronous local obstacle avoidance mechanism: If a candidate expansion direction is found to enter the neighborhood of a high-risk obstacle during the expansion process, a finite number of potential field-driven correction steps are performed within a local window, starting from the current search point and ending at a feasible look-ahead point in the original target direction, to generate a bypass connection segment that can be quickly deployed. This connection segment is then used as a new expansion candidate to continue participating in the process. search; Task attribute-driven potential parameter mapping includes: Map task attributes to APF-related parameter combinations, including , , , , Terrain sensitivity coefficient and obstacle risk weight make different task types show differences in safety redundancy, detour tendency and trajectory smoothness; High-priority tasks can choose higher obstacle risk weights and stricter ground altitude redundancy while ensuring safety, while general tasks can use a more balanced combination of parameters to achieve a graded trade-off between timeliness and safety.

Images