Limited space unmanned aerial vehicle autonomous homeward voyage method and device based on real-time mapping and path evaluation, and medium

Abstract

The invention belongs to the technical field of unmanned aerial vehicle path planning, and particularly relates to a limited space unmanned aerial vehicle autonomous homeward voyage method based on real-time mapping and path evaluation. The method comprises the steps that when the unmanned aerial vehicle flies in a limited space without a GPS, real-time positioning is carried out through a laser radar SLAM technology, a three-dimensional grid map is constructed synchronously, and meanwhile a flight path is recorded in real time in a key point mode. When the electric quantity is insufficient or communication is lost to trigger return flight, the system generates two candidate paths in parallel, one path is a historical path which is backtracked in a reverse order, and the other path is a new path which is searched on a real-time map by using an A * algorithm. Then, the two paths are evaluated through a comprehensive scoring function, and scoring standards comprise path smoothness, path safety and path length; and finally, selecting a path with a relatively high score as an execution path, and performing safe return by combining a local obstacle avoidance algorithm. According to the invention, the limitation of pure path backtracking is overcome, and safer, more efficient and more intelligent autonomous homeward voyage is realized.

Description

Technical Field

[0001] This invention belongs to the technical field of unmanned aerial vehicle (UAV) path planning, and more specifically, relates to a method, device, and medium for autonomous return of UAVs in confined space based on real-time mapping and path evaluation. Background Technology

[0002] With the rapid development of drone technology, its application scenarios have expanded from open outdoor environments to various complex confined spaces, such as indoor warehouse inspection, tunnel structure inspection, mine resource exploration, and bridge bottom inspection. These environments typically cannot receive Global Positioning System (GPS) signals or are subject to severe interference, making it impossible for drones to rely on traditional GPS for positioning and navigation.

[0003] Currently, in the absence of GPS, drones typically rely on technologies such as inertial measurement units (IMUs), lidar, visual sensors, and ultra-wideband (UWB) for autonomous positioning and obstacle avoidance. However, when drones face emergencies such as low battery or loss of connection with the remote controller and need to automatically return to base, existing technologies have the following limitations: Global path planning relies on prior maps: many solutions require precise mapping of the workspace beforehand to build a high-precision environmental map. This not only increases workload and cost, but also means that the map may become invalid when the environment changes dynamically, making the planned path infeasible.

[0004] Lack of effective utilization of flight history: The path taken by a drone into the operational area is usually a safe and passable path. Existing technology fails to efficiently utilize this known flight history information, instead resorting to time-consuming recalculations.

[0005] For example, Chinese patent document CN107291099A discloses a method for returning a drone to its home port, which includes obtaining the starting position information of the drone's starting position; obtaining map data of the current position; calculating the flight trajectory of the drone from the starting position to the current flight position when the drone is flying; calculating the flight direction and path of the drone when returning to its home port based on the flight trajectory; and guiding the drone to return to its home port according to the flight direction and path.

[0006] Therefore, a return-to-base method is needed that can make intelligent decisions by utilizing historical flight information and integrating real-time environmental awareness. Summary of the Invention

[0007] The present invention aims to overcome at least one of the defects of the prior art and provide a method for autonomous return of UAVs in confined space based on real-time mapping and path evaluation. The UAV uses real-time mapping and combined with flight history paths to perform multi-criteria evaluation to select the optimal return path, thereby overcoming the limitations of a single decision-making method.

[0008] The detailed technical solution of this invention is as follows: A method for autonomous return-to-home of a UAV in a confined space based on real-time mapping and path evaluation, the method comprising: S1. When the drone is flying in a limited space without GPS, it can perceive the surrounding environment through airborne sensors, perform autonomous positioning, and build a three-dimensional grid map in real time. S2. Starting from the takeoff point, the UAV records key waypoints in real time to form a historical flight path sequence; S3. Monitor the drone's battery level and communication status. When the preset return-to-home conditions are met, automatically trigger the return-to-home procedure and start hybrid path planning to generate two candidate paths: a historical backtracking path and a new optimal forward path. S4. Establish a comprehensive scoring function to evaluate the two candidate paths. The scoring criteria include path smoothness, path safety, and path length. Finally, select the path with the higher score as the final execution path. S5. Combine local obstacle avoidance algorithm for safe return. During the return process, the UAV continues to execute the process in step S2 and records the flight path. When an obstacle that the local planner cannot bypass appears on the final execution path, steps S3 and S4 are re-executed to re-plan and evaluate a new final execution path and continue execution.

[0009] According to a preferred embodiment of the present invention, step S1 specifically refers to: The drone is equipped with lidar and inertial measurement unit sensors. During flight, it uses SLAM technology to estimate the drone's pose in its own local coordinate system in real time. At the same time, it uses radar point cloud data to build and update a three-dimensional occupancy grid map of the environment in real time. This map identifies free space, obstacle space and unknown space.

[0010] According to a preferred embodiment of the present invention, the specific recording strategy in step S2 is as follows: S21. Straight Flight Judgment: Continuously monitor the rate of change of the drone's heading angle. When the rate of change is lower than a preset threshold, the drone is judged to be in a state of near-straight flight. S22. Straight line segment recording: When in a straight flight state, only the starting coordinates and ending coordinates of the straight line segment are recorded. The straight line segment itself represents this flight path. S23. Curve segment recording: When the rate of change of heading angle exceeds the threshold, the UAV enters a non-straight flight state, stops the current straight segment recording, and starts the distance interval recording mode, that is, every preset fixed distance D, a waypoint coordinate is recorded until the flight tends to be straight again. S24. Pathpoint Deduplication Optimization: While recording new waypoints, the consistency between the new waypoints and the already stored pathpoint sequences is checked in real time. The checking methods include: Distance judgment: Calculate the Euclidean distance between the new waypoint and the nearest stored waypoint. If the distance is less than a set threshold, it is judged as overlapping or too close, and the new waypoint is discarded. Path overlap detection: When the positioning information determines that the current flight path of the drone overlaps with a recorded path segment, all intermediate waypoints between the two overlapping waypoints are automatically removed from the recording sequence. When returning home, the drone will fly directly from the end waypoint of the overlapping path to the starting waypoint, thereby shortening the return path and avoiding repeated flights. S25. Data Storage: All optimized key waypoints, including the start and end points of straight segments, waypoints of curved segments, and deduplicated pathpoints, are stored in chronological order into a circular queue or list within the flight control system, forming an optimized return pathpoint sequence. ,in The takeoff point is used as the reference point, and the minimum radius R of the free space between the location of each waypoint detected by the lidar and the plane perpendicular to the heading is recorded.

[0011] According to a preferred embodiment of the present invention, in step S3, the historical backtracking path refers to: directly calling the path point sequence stored in step S2. Reverse the order of the data to generate a historical backtracking path from the current position to the takeoff point. ; The new optimal forward path refers to: taking the current location of the UAV as the starting point, and the takeoff point... With the destination as the starting point, on the currently constructed 3D raster map, use the A* search algorithm or its variants for global path planning to generate a new optimal forward path. .

[0012] According to a preferred embodiment of the present invention, in step S4, the path smoothness refers to the number and angle of turning points in the path. The turning angles between all segments of the path are calculated, and a function penalizes paths with a high number of turns and large turning angles, with its score component... It is inversely proportional to the total angle of change and the number of changes; The path safety refers to the spaciousness of the space traversed by the path. It analyzes the grids and their neighborhoods along the path, and calculates the spatial scale cost of historically traced paths using the minimum radius R of the recorded free space. Paths traversing narrow spaces are penalized, and their score components are used to determine the safety factor. It is proportional to the minimum radius of the average free space of the area traversed by the path or the width of the passage; The scoring component of the path length It is inversely proportional to the total path length L, that is =1 / L; The total score is the weighted sum of all components: Score = w1 + w2 + w3

[0013] Where w1, w2, and w3 are weighting coefficients, which can be adjusted according to task requirements; Finally, the scores of the two paths are compared, and the path with the higher score is selected as the final return path. .

[0014] According to a preferred embodiment of the present invention, the specific method for calculating the path smoothness is as follows: For a given path, it is defined as a series of ordered waypoints. Composition, in which As the starting point of the path, The path endpoint is the path segment vector formed by adjacent waypoints. For each internal waypoint in the path Calculate the vectors of the two path segments before and after it. and The angle between As a turning point, The value range is [0°, 180°]; Turning angle The larger, the greater the cost The higher, For the entire path:

[0015] Where N is the number of internal waypoints where the path turns, i.e. The number of points, or approximately. > Quantity, A threshold close to 0 is used to ignore minute jitter, set according to the actual scenario. α is a fixed cost weighting coefficient that penalizes each inflection point, which can be set according to the specific drone platform. The item is a direct penalty for the number of transitions; The specific method for calculating path safety is as follows: calculate the distance from each grid along the path to its nearest surrounding obstacle, and for those distances less than a preset threshold... All grids, whose distance from the nearest obstacle is... ,but , where i is 1 to n.

[0016] According to a preferred embodiment of the present invention, the safe return to base using the combined local obstacle avoidance algorithm means that the UAV returns to base along the finally selected path. During flight, the radar continuously detects obstacles and integrates a local planner, employing Timed ElasticBand or model predictive control algorithms to handle unforeseen dynamic obstacles on the global path, ensuring flight safety.

[0017] In another aspect of the invention, an apparatus is also provided for implementing an autonomous return-to-home method for unmanned aerial vehicles in confined spaces based on real-time mapping and path evaluation, comprising: At least one processor; and The memory stores instructions that, when executed by the at least one processor, cause the at least one processor to perform the autonomous return-to-home method for unmanned aerial vehicles in confined spaces based on real-time mapping and path evaluation as described above.

[0018] In another aspect of the invention, a machine-readable storage medium is also provided, which stores executable instructions that, when executed, cause the machine to perform the autonomous return-to-home method for a confined space UAV based on real-time mapping and path evaluation as described above.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention overcomes the limitations of a single method by comparing the dual strategies of "historical backtracking" and "real-time planning" in parallel, resulting in better decision-making results.

[0020] (2) The present invention introduces a map-based “path safety” score, which can proactively avoid guiding the drone back to narrow or dangerous areas and select a safer path.

[0021] (3) This invention introduces a "path smoothness" score, which prioritizes the selection of paths with fewer turns and more direct routes, thereby reducing unnecessary energy consumption and flight time and improving efficiency. This is especially critical in emergency return trips due to insufficient battery power. Attached Figure Description Figure 1 This is a flowchart of the autonomous return method for unmanned aerial vehicles in confined space based on real-time mapping and path evaluation, as described in this invention.

[0022] Figure 2 This is a schematic diagram of key waypoint recording and path deduplication in the return-to-home method of the UAV flight path described in this invention.

[0023] Figure 3 This is a schematic diagram showing the comprehensive evaluation and comparison of candidate return paths in the return method described in this invention. Detailed Implementation

[0024] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0025] Example 1 Reference Figure 1 , this embodiment provides an autonomous return method for an unmanned aerial vehicle (UAV) in a confined space based on real-time mapping and path evaluation. The method includes: S1. When the UAV flies in a GPS-free confined space, it senses the surrounding environment through on-board sensors, performs autonomous positioning, and constructs a three-dimensional grid map in real time. Specifically: The UAV is equipped with sensors such as lidar and inertial measurement unit (IMU). During flight, using SLAM technology, the pose (x, y, z, roll, pitch, yaw) of the UAV in its own established local coordinate system is estimated in real time, where x, y, and z are the position coordinates of the UAV in the coordinate system, roll is the roll angle of the UAV, pitch is the pitch angle of the UAV, and yaw is the heading angle of the UAV; at the same time, the three-dimensional occupancy grid map of the environment is constructed and updated in real time using the radar point cloud data, and this map identifies free space, obstacle space, and unknown space; S2. Starting from the take-off point, key flight waypoints are recorded in real time to form a historical flight path sequence. The specific recording strategy is as follows: S21. Straight flight determination: Continuously monitor the change rate of the heading angle of the UAV. When the change rate is lower than the preset threshold, it is determined that the UAV is in an approximately straight flight state; for example Figure 2 , the dashed line is the actual flight path of the UAV. The section from point D to the current position is a straight line, and the other parts are curves; S22. Straight segment recording: When in a straight flight state, only the starting coordinates and ending coordinates of the straight segment are recorded. This straight segment itself represents this section of the flight path; for a straight line, only the starting and ending points are recorded, which can reduce data storage and calculation amounts; S23. Curve segment recording: When the change rate of the heading angle exceeds the threshold, the UAV enters a non-straight flight state such as turning or avoiding obstacles. Stop recording the current straight segment and start the recording mode at a distance interval, that is, record a waypoint coordinate every preset fixed distance D (for example, 0.5 meters to 2 meters) until the flight tends to be straight again; S24. Path point duplicate removal and optimization: While recording new waypoints, the new waypoints are judged for consistency with the already stored path point sequence in real time. The judgment methods include but are not limited to: Distance judgment: Calculate the Euclidean distance between the new waypoint and the nearest stored waypoint. If the distance is less than the set threshold, it is determined that they coincide or are too close, and the new waypoint is discarded; Path coincidence judgment: When it is judged through positioning information that the current flight trajectory of the UAV coincides with a certain recorded path, all intermediate waypoints between the two coincident waypoints are automatically cleared from the recording sequence. When returning, the UAV will directly fly from the ending waypoint of the coincident path to the starting waypoint, thus shortening the return path and avoiding repeated flight; in Figure 2 In the diagram, the aircraft's flight path is from takeoff point > A > B > C > A > D. The segment A > B > C > A coincides with point A, so this segment is discarded during recording, and only the takeoff point > A > D segment is recorded. S25. Data Storage: All optimized key waypoints (including the start and end points of straight segments, waypoints of curved segments, and deduplicated pathpoints) are stored in chronological order into a circular queue or list within the flight control system, forming an optimized return pathpoint sequence. ,in The takeoff point is used as the reference point, and the minimum radius R of the free space between the location of each waypoint detected by the lidar and the plane perpendicular to the heading is recorded.

[0026] S3. Return-to-Home Trigger Steps: Monitor the drone's battery level and communication status. When the preset return-to-home conditions are met, automatically trigger the return-to-home procedure and initiate hybrid path planning, generating two candidate paths: Path 1: Directly call the path point sequence stored in step S2 Reverse the order of the data to generate a historical backtracking path from the current position to the takeoff point. ; Path 2: Starting from the current location of the drone, the takeoff point... With the destination as the starting point, on the currently constructed 3D raster map, use the A* search algorithm or its variants for global path planning to generate a new optimal forward path. .

[0027] S4. Establish a comprehensive scoring function for the two candidate paths. and An evaluation is conducted, with scoring criteria including path smoothness, path safety, and path length; ultimately, the path with the higher score is selected as the execution path. The path smoothness refers to the number and angle of turning points in the path. It calculates the turning angles between all segments of the path, and a function penalizes paths with a high number of turns and large turning angles, assigning a score component to these paths. It is inversely proportional to the total turning angle and the number of turning points. The specific calculation method is as follows: For a given path, it is defined as a series of ordered waypoints. Composition, in which As the starting point of the path, This is the endpoint of the path. Adjacent waypoints form the path segment vector. For each internal waypoint in the path (i = 1, 2, ..., n-1), calculate the vectors of the two path segments before and after it. and The angle between As a turning point, The value range is [0°, 180°]. Turning angle. The larger, the greater the cost The higher, For the entire path, Where N is the number of internal waypoints where the path turns, i.e. The number of points, or approximately. > Quantity, A threshold close to 0 is used to ignore minute jitter, set according to the actual scenario. α is a fixed cost weighting coefficient that penalizes each inflection point, which can be set according to the specific drone platform. The term is a direct penalty for the number of turns, even if all turns are smooth. While the number of turns is small, continuous turns (where N is large) can also lead to an uneven path and low flight efficiency. This ensures that the system will tend to choose a path with fewer turns.

[0028] The path safety refers to the spaciousness of the space traversed by the path. It analyzes the grids and their neighborhoods along the path, and calculates the spatial scale cost of historically traced paths using the minimum radius R of the recorded free space. Paths traversing narrow spaces are penalized, and their score components are used to determine the safety factor. It is proportional to the minimum radius of the average free space of the area traversed by the path or the width of the passage. The specific calculation method is as follows: calculate the distance from each grid passed by the path to the nearest obstacle in its vicinity, and for those distances less than a threshold... (Based on scene settings) All grids, their distance from the nearest obstacle is ,but , where i is 1 to n.

[0029] The scoring component of the path length It is inversely proportional to the total path length L. =1 / L.

[0030] The total score is the weighted sum of all components: Score = w1 + w2 + w3

[0031] Where w1, w2, and w3 are weighting coefficients, which can be adjusted according to task requirements; Compare the scores of the two paths and select the path with the higher score as the final return path. .

[0032] For example in Figure 3There are 3 planned paths. The dashed path is the shortest, but it needs to pass through the narrow space between obstacles 3 and 5, so the path safety is low. The solid path has more turns, so the path smoothness score is low. The dotted-line path does not pass through the narrow path and has relatively few turns, but the path length is longer. The final path selection needs to be determined based on the comprehensive score.

[0033] S5. Combine local obstacle avoidance algorithm to return safely. During the return process, the UAV continues to execute the process in step S2 and records the flight path. When an obstacle that the local planner cannot bypass appears on the final execution path, steps S3 and S4 are re-executed to re-plan and evaluate a new final execution path and continue execution. The safe return to base using the combined local obstacle avoidance algorithm refers to the drone returning along the finally selected path. During flight, the radar continuously detects obstacles and integrates a local planner (such as using Timed Elastic Band or model predictive control algorithms) to handle unforeseen dynamic obstacles on the global path, ensuring flight safety.

[0034] Example 2 This embodiment also provides an apparatus for implementing an autonomous return-to-home method for unmanned aerial vehicles in confined spaces based on real-time mapping and path evaluation, comprising: At least one processor; and The memory stores instructions that, when executed by the at least one processor, cause the at least one processor to perform the autonomous return-to-home method for unmanned aerial vehicles in confined spaces based on real-time mapping and path evaluation as described above.

[0035] In this embodiment, the electronic device may include, but is not limited to: personal computer, server computer, workstation, desktop computer, laptop computer, notebook computer, mobile computing device, smartphone, tablet computer, cellular phone, personal digital assistant (PDA), handheld device, messaging device, wearable computing device, consumer electronic device, etc.

[0036] Example 3 This embodiment also provides a machine-readable storage medium storing executable instructions that, when executed, cause the machine to perform the autonomous return-to-home method for a UAV in a confined space based on real-time mapping and path evaluation as described above.

[0037] Specifically, a system or apparatus equipped with a readable storage medium may be provided, on which software program code implementing the functions of any of the embodiments described above is stored, and the computer or processor of the system or apparatus can read and execute the instructions stored in the readable storage medium.

[0038] In this case, the program code read from the readable medium itself can perform the functions of any of the above embodiments, and therefore the machine-readable code and the readable storage medium storing the machine-readable code constitute a part of this specification.

[0039] Examples of readable storage media include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD-RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer or the cloud via a communication network.

[0040] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the technical solutions of the present invention, and are not intended to limit the specific implementation of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the claims of the present invention should be included within the protection scope of the claims of the present invention.

Claims

1. A method for autonomous return-to-home of a UAV in a confined space based on real-time mapping and path evaluation, the method comprising: S1. When the drone is flying in a limited space without GPS, it can perceive the surrounding environment through airborne sensors, perform autonomous positioning, and build a three-dimensional grid map in real time. S2. Starting from the takeoff point, the UAV records key waypoints in real time to form a historical flight path sequence; S3. Monitor the drone's battery level and communication status. When the preset return-to-home conditions are met, automatically trigger the return-to-home procedure and start hybrid path planning to generate two candidate paths: a historical backtracking path and a new optimal forward path. S4. Establish a comprehensive scoring function to evaluate the two candidate paths. The scoring criteria include path smoothness, path safety, and path length. Finally, select the path with the higher score as the final execution path. S5. Combine local obstacle avoidance algorithm for safe return. During the return process, the UAV continues to execute the process in step S2 and records the flight path. When an obstacle that the local planner cannot bypass appears on the final execution path, steps S3 and S4 are re-executed to re-plan and evaluate a new final execution path and continue execution.

2. The autonomous return-to-home method for unmanned aerial vehicles in confined spaces based on real-time mapping and path evaluation according to claim 1, characterized in that, Step S1 specifically refers to: The drone is equipped with lidar and inertial measurement unit sensors. During flight, it uses SLAM technology to estimate the drone's pose in its own local coordinate system in real time. At the same time, it uses radar point cloud data to build and update a three-dimensional occupancy grid map of the environment in real time. This map identifies free space, obstacle space and unknown space.

3. The autonomous return-to-home method for unmanned aerial vehicles in confined spaces based on real-time mapping and path evaluation according to claim 1, characterized in that, The specific recording strategy in step S2 is as follows: S21. Straight Flight Judgment: Continuously monitor the rate of change of the drone's heading angle. When the rate of change is lower than a preset threshold, the drone is judged to be in a state of near-straight flight. S22. Straight line segment recording: When in a straight flight state, only the starting coordinates and ending coordinates of the straight line segment are recorded. The straight line segment itself represents this flight path. S23. Curve segment recording: When the rate of change of heading angle exceeds the threshold, the UAV enters a non-straight flight state, stops the current straight segment recording, and starts the distance interval recording mode, that is, every preset fixed distance D, a waypoint coordinate is recorded until the flight tends to be straight again. S24. Pathpoint Deduplication Optimization: While recording new waypoints, the consistency between the new waypoints and the already stored pathpoint sequences is checked in real time. The checking methods include: Distance judgment: Calculate the Euclidean distance between the new waypoint and the nearest stored waypoint. If the distance is less than a set threshold, it is judged as overlapping or too close, and the new waypoint is discarded. Path overlap detection: When the location information determines that the current flight path of the drone overlaps with a recorded path segment, all intermediate waypoints between the two overlapping waypoints are automatically removed from the recording sequence. When returning, the drone will fly directly from the end waypoint of the overlapping path to the starting waypoint. S25. Data Storage: All optimized key waypoints, including the start and end points of straight segments, waypoints of curved segments, and deduplicated pathpoints, are stored in chronological order into a circular queue or list within the flight control system, forming an optimized return pathpoint sequence. ,in The takeoff point is used as the reference point, and the minimum radius R of the free space between the location of each waypoint detected by the lidar and the plane perpendicular to the heading is recorded.

4. The autonomous return-to-home method for unmanned aerial vehicles in confined spaces based on real-time mapping and path evaluation according to claim 3, characterized in that, In step S3, the historical backtracking path refers to directly calling the path point sequence stored in step S2. Reverse the order of the data to generate a historical backtracking path from the current position to the takeoff point. ; The new optimal forward path refers to: taking the current location of the UAV as the starting point, and the takeoff point... With the destination as the starting point, on the currently constructed 3D raster map, use the A* search algorithm or its variants for global path planning to generate a new optimal forward path. .

5. The autonomous return-to-home method for unmanned aerial vehicles in confined spaces based on real-time mapping and path evaluation according to claim 4, characterized in that, In step S4, the path smoothness refers to the number and angle of turning points in the path. The turning angles between all segments of the path are calculated, and a function penalizes paths with a high number of turns and large turning angles, which are then included in the score component. It is inversely proportional to the total angle of change and the number of changes; The path safety refers to the spaciousness of the space traversed by the path. It analyzes the grids and their neighborhoods along the path, and calculates the spatial scale cost of historically traced paths using the minimum radius R of the recorded free space. Paths traversing narrow spaces are penalized, and their score components are used to determine the safety factor. It is proportional to the minimum radius of the average free space of the area traversed by the path or the width of the passage; The scoring component of the path length It is inversely proportional to the total path length L, that is =1 / L; The total score is the weighted sum of all components: Score = w1 + w2 + w3 Where w1, w2, and w3 are weighting coefficients, which can be adjusted according to task requirements; Finally, the scores of the two paths are compared, and the path with the higher score is selected as the final return path. .

6. The autonomous return method for unmanned aerial vehicles in confined space based on real-time mapping and path evaluation according to claim 5, characterized in that, The specific method for calculating the path smoothness is as follows: For a given path, it is defined as a series of ordered waypoints. Composition, in which As the starting point of the path, The path endpoint is the path segment vector formed by adjacent waypoints. For each internal waypoint in the path Calculate the vectors of the two path segments before and after it. and The angle between As a turning point, The value range is [0°, 180°]; Turning angle The larger, the greater the cost The higher, For the entire path: Where N is the number of internal waypoints where the path turns, i.e. The number of points, or approximately. > Quantity, A threshold close to 0 is used to ignore minute jitter, set according to the actual scenario. α is a fixed cost weighting coefficient that penalizes each inflection point, which can be set according to the specific drone platform. The item is a direct penalty for the number of transitions; The specific method for calculating path safety is as follows: calculate the distance from each grid along the path to its nearest surrounding obstacle, and for those distances less than a preset threshold... All grids, whose distance from the nearest obstacle is... ,but , where i is 1 to n.

7. The autonomous return-to-home method for unmanned aerial vehicles in confined spaces based on real-time mapping and path evaluation according to claim 6, characterized in that, The safe return to base using the combined local obstacle avoidance algorithm refers to the drone returning along the finally selected path. During flight, the radar continuously detects obstacles and integrates a local planner, employing Timed Elastic Band or model predictive control algorithms to handle unforeseen dynamic obstacles on the global path, ensuring flight safety.

8. An apparatus for realizing an autonomous return-to-home method for unmanned aerial vehicles in confined spaces based on real-time mapping and path evaluation, characterized in that, The device includes: At least one processor; and The memory stores instructions that, when executed by the at least one processor, cause the at least one processor to perform the autonomous return-to-home method for a confined space UAV based on real-time mapping and path evaluation as described in any one of claims 1 to 7.

9. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores executable instructions that, when executed, cause the machine to perform the autonomous return-to-home method for a confined space UAV based on real-time mapping and path evaluation as described in any one of claims 1 to 7.

Images