Unmanned aerial vehicle exploration mapping method based on supplementary ray projection update

Abstract

This invention discloses a UAV exploration mapping method based on supplementary ray projection updates, belonging to the field of environmental perception and spatial modeling technology. It is used in the process of UAV LiDAR exploration mapping to perform reasonable and conservative state updates on occupied grid maps in the absence of effective echoes. This invention analyzes the visibility and occupancy relationships of candidate regions and introduces supplementary ray projection when preset conditions are met. It updates theoretically visible regions that have not received echoes through free space inference, thereby eliminating false boundaries introduced by the limitations of echo-driven update mechanisms and improving the accuracy and consistency of map states. This invention effectively compensates for the update defects of traditional occupancy grid mapping methods based on ray projection in echo-free directions without changing existing LiDAR hardware configurations or requiring additional external information or communication support, thus improving the reliability of map updates and the completeness of the final map.

Description

Technical Field

[0001] This invention belongs to the field of environmental perception and spatial modeling technology, specifically relating to the design of a UAV exploration mapping method based on supplementary ray projection updates. Background Technology

[0002] Autonomous exploration and mapping technology using UAVs equipped with LiDAR has been widely applied in the fields of environmental perception and spatial modeling. Existing methods typically model the environment based on occupancy grid maps and update the occupancy status of grid voxels using LiDAR point clouds. The updating of occupancy grids often employs an echo-driven ray update mechanism, which updates the voxel state to free or occupancy only along the ray path where valid point cloud echoes are obtained from the LiDAR. Due to its simplicity and high computational efficiency, this method is widely used in grid-based UAV exploration and mapping systems.

[0003] However, the aforementioned traditional ray casting mechanism has certain limitations in practical map exploration applications. Specifically, when the lidar does not obtain effective echoes in certain directions, such as in distant areas, open spaces, mirrored or low-reflectivity surfaces, slopes with unfavorable incident angles, or environments with sparse structures, the corresponding rays cannot trigger voxel state updates, causing voxels in these directions to remain in an unknown state for extended periods. Although geometrically, these areas may be theoretically within the visible range of the current UAV position, the lack of actual echo returns prevents the traditional ray-casting method from correcting their states. This forces the UAV to spend a long time exploring the vicinity, waiting for voxel and boundary state updates, further impacting the efficiency of exploration and mapping.

[0004] To address the difficulty of updating the direction without echoes, some existing technologies attempt to obtain effective echoes in subsequent observations by extending observation time, increasing the number of repeated scans, adjusting flight altitude, or changing flight paths, thereby indirectly completing voxel updates. However, these methods still fundamentally rely on acquiring real echoes and cannot solve the problem of updating the direction without echoes at the mapping mechanism level. Furthermore, they often require introducing additional flight distance and time overhead, reducing mapping efficiency.

[0005] Within the framework of grid-based mapping and boundary detection, the aforementioned unknown voxels that cannot be updated for extended periods are often continuously identified as boundary regions, resulting in a large number of persistent pseudo-boundaries. These pseudo-boundaries are not genuine entrances to unknown spaces, but rather introduced by the limitations of the echo-driven update mechanism. Existing technologies lack effective means to distinguish or eliminate these pseudo-boundaries, causing the system to still treat them as valid boundary regions during mapping or exploration.

[0006] The existence of these false boundaries has several adverse effects: First, the system may repeatedly perform path planning and update operations around the false boundaries, resulting in redundant processing of already observed areas. Second, map state determination is prone to errors, causing the system to mistakenly believe that a large number of unupdated areas still exist, thus affecting the continuity of the mapping process. Third, the final generated map is prone to areas with missing coverage or sparse structure, making it difficult to meet the requirements for map integrity and consistency. These problems are particularly pronounced in open or weakly structured environments where echo-free directions are more common, severely limiting the reliability and applicability of UAV-based occupancy grid-based exploration and mapping technology in complex scenarios.

[0007] Therefore, the existing occupancy grid mapping method based on echo-driven ray updating has insurmountable technical defects when dealing with the absence of effective echo directions. There is an urgent need for a mapping and updating method that can reasonably correct the map state, effectively eliminate false boundaries, and improve the reliability of map updates even in the absence of echoes. Summary of the Invention

[0008] The purpose of this invention is to address the shortcomings of existing UAV occupancy grid mapping methods based on LiDAR in terms of map update mechanisms. It proposes a UAV exploration mapping method based on supplementary ray projection updates. Unlike traditional echo-driven updates based on actual measured rays, this method does not use the actual echo endpoint as the sole update criterion. Instead, when candidate boundary clusters meet the theoretically visible and unobstructed conditions, supplementary rays are actively constructed. Free-space inference updates are performed on unknown areas that have not received effective echoes but should geometrically be within the visible range. This effectively compensates for the update defects of traditional occupancy grid mapping methods in echo-free directions, improving the reliability of map updates and the completeness of the final map.

[0009] The technical solution of this invention is: a UAV exploration mapping method based on supplementary ray projection update, comprising the following steps: S1. Based on the occupancy grid map of the current UAV equipped with LiDAR, extract candidate boundary voxels and cluster them to obtain candidate boundary clusters.

[0010] S2. For each candidate boundary cluster, select at least one representative point to describe the spatial location and orientation characteristics of the boundary cluster.

[0011] S3. Perform visibility condition judgment on the representative points of each candidate boundary cluster.

[0012] S4. Perform supplementary ray casting and line-of-sight detection on the candidate boundary clusters where representative points that meet the visibility execution conditions are located.

[0013] S5. Update the free space inference of the supplementary rays detected by line of sight.

[0014] S6. In response to the voxels in the neighborhood of the representative point in the candidate boundary cluster being updated to the free space state, the boundary state corresponding to the candidate boundary cluster is adjusted.

[0015] S7. During the UAV mapping process, as the UAV's position changes, steps S1 to S6 are repeated periodically or as needed to continuously update the occupied grid map and boundary status as the mapping process progresses.

[0016] Further, step S1 includes the following sub-steps: S11. Based on the updated occupied grid map of the UAV equipped with LiDAR in the current frame, perform incremental detection on the boundary state and extract voxels at the junction of free space and unknown space as candidate boundary voxels.

[0017] S12. Based on the spatial connectivity between voxels, cluster the candidate boundary voxels to form several candidate boundary clusters.

[0018] Furthermore, each candidate boundary cluster represents a spatial region that has not yet been updated or determined to be unknown in the current map state.

[0019] Furthermore, in step S2, the representative point is selected as the geometric centroid of the candidate boundary cluster, or one or more boundary voxel points selected from the candidate boundary cluster.

[0020] Furthermore, the visibility execution condition in step S3 is: the representative point is located within the theoretical field of view of the lidar sensor, and according to the current occupied grid map, there is no occlusion structure composed of known occupied voxels in the direction of the line connecting the current position of the UAV to the representative point.

[0021] Furthermore, step S4 includes the following sub-steps: S41. Using the current position of the UAV as the starting point of the ray, emit at least one supplementary ray toward the representative point that meets the visibility execution conditions.

[0022] S42. Along the propagation path of the supplementary ray, perform line-of-sight detection based on the occupied grid map. When no known occupied voxel is detected on the path of the supplementary ray, determine that the direction corresponding to the supplementary ray is a theoretically visible and unobstructed direction. When an obstructing voxel is detected on the path of the supplementary ray, terminate the update operation of the supplementary ray.

[0023] Further, step S5 specifically involves selecting a set of voxels within a preset length range as update objects along the propagation path of the supplementary ray detected by the line of sight, and updating the voxels to a free space state.

[0024] Furthermore, the adjustment in step S6 includes clearing boundary markers and eliminating false boundaries introduced by the non-echo direction.

[0025] The beneficial effects of this invention are: (1) This invention can reasonably correct the map state under the condition of no effective echo, overcoming the inherent limitations of the traditional ray-based update mechanism. Existing grid-based mapping methods only update the voxel state along the ray path where the laser echo is actually obtained, and cannot perform any update operation in the direction without echo, resulting in the theoretically visible area remaining in an unknown state for a long time. This invention analyzes the visibility and occlusion relationship of candidate areas, and introduces supplementary ray projection when preset conditions are met, and performs free space inference update for the spatial area that is theoretically visible but without echo, thus making up for the shortcomings of the traditional echo-driven update method in the direction without return at the mechanism level.

[0026] (2) This invention effectively eliminates false boundaries introduced by the echo-free direction, improving the accuracy of map state determination. Because voxels in the echo-free direction cannot be updated for a long time, existing technologies easily create a large number of false boundaries in the grid map that are not true unknown spatial entrances. This invention adjusts or clears the corresponding boundary states after supplementing the ray update coverage of key voxels, enabling the timely elimination of false boundaries caused by the limitations of the update mechanism, thereby preventing their long-term existence and interference with map state determination.

[0027] (3) This invention improves the reliability and consistency of map updates while ensuring the conservatism of the update process. Before performing supplementary ray projection, this invention judges the relationship between the sensor's theoretical field of view and occlusion, and only performs inference updates when the target area is theoretically visible and unobstructed, thus avoiding incorrect corrections to invisible or occluded areas. Through this conservative update strategy, the reliability and consistency of map state updates are ensured while introducing inference update capabilities. Attached Figure Description

[0028] Figure 1 The diagram shows a flowchart of a UAV exploration and mapping method based on supplementary ray projection update provided by an embodiment of the present invention. Detailed Implementation

[0029] Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described in the drawings are merely exemplary and are intended to illustrate the principles and spirit of the invention, and are not intended to limit the scope of the invention.

[0030] This invention provides a UAV exploration mapping method based on supplementary ray projection updates, such as... Figure 1 As shown, the process includes the following steps S1 to S7: S1. Based on the occupancy grid map of the current UAV equipped with LiDAR, extract candidate boundary voxels and cluster them to obtain candidate boundary clusters.

[0031] S1 includes the following sub-steps S11~S12: S11. Based on the updated occupied grid map of the UAV equipped with LiDAR in the current frame, perform incremental detection on the boundary state and extract voxels at the junction of free space and unknown space as candidate boundary voxels.

[0032] In this embodiment of the invention, the boundary is constructed incrementally. Instead of globally reconstructing the boundaries of the entire occupied grid map at every moment, it dynamically adjusts the boundary based on the boundary state at the previous moment and the update results of the occupied grid map from the current frame's LiDAR. When an unknown voxel in the neighboring region of the boundary at the previous moment is updated to free space by the current frame's observation, the corresponding original boundary voxel becomes invalid and is cleared. Simultaneously, new candidate boundary voxels are extracted at the boundary between the newly added free space and the unknown space, thereby achieving frame-by-frame recursive updating of the boundary.

[0033] Let the occupation of the grid map be denoted as , Indicates the number of occupies in the raster map Each voxel has a corresponding occupancy status marker. The occupied state includes at least free space, unknown space, and occupied space. When a voxel... A voxel is considered a candidate boundary voxel if it meets the following conditions: in Represents free space. Represents unknown space. Indicates the number of occupies in the raster map Individual factors, Voxel representation Neighborhood. Then we have , This represents the set of candidate boundary voxels, composed of newly added candidate boundary voxels in the current frame and valid boundary voxels retained from the previous time step. For original boundary voxels that no longer satisfy the boundary condition between free space and unknown space due to map updates in the current frame, their boundary markers are cleared. This allows the boundary to continuously advance outwards during UAV observation without requiring a full re-initialization of the boundaries across the entire map.

[0034] S12. Based on the spatial connectivity between voxels, cluster the candidate boundary voxels to form several candidate boundary clusters. Then, use the PAC method to process the larger candidate boundary clusters and divide them into multiple smaller boundary clusters to reduce computational complexity.

[0035] In this embodiment of the invention, the 26-adjacency method is used to process the candidate boundary voxel set. Clustering is performed to form several candidate boundary clusters: in Represents the set of candidate boundary clusters. Indicates the first Candidate boundary clusters, , This represents the total number of candidate boundary clusters.

[0036] Each candidate boundary cluster represents a spatial region that has not yet been updated or has been determined to be unknown in the current map state, and is used for subsequent visibility analysis and update determination.

[0037] S2. For each candidate boundary cluster, select at least one representative point to describe the spatial location and orientation characteristics of the boundary cluster.

[0038] In this embodiment of the invention, the representative point is selected as the geometric centroid of the candidate boundary cluster, or one or more boundary voxel points selected from the candidate boundary cluster. By selecting the representative point, the spatial region at the boundary cluster level is mapped to a finite number of spatial reference points, thereby reducing the computational complexity of subsequent visibility determination and ray projection.

[0039] In this embodiment of the invention, the formula for calculating the geometric centroid is: in Voxel representation spatial coordinate vector, Represents candidate boundary clusters Number of voxels included Indicates the first The position vectors of the geometric centroids of candidate boundary clusters.

[0040] S3. Perform visibility condition judgment on the representative points of each candidate boundary cluster.

[0041] In this embodiment of the invention, the visibility execution condition is: the representative point is located within the theoretical field of view of the lidar sensor, and according to the currently occupied grid map, there is no occlusion structure composed of known occupied voxels in the direction of the line connecting the current position of the UAV to the representative point.

[0042] Supplementary ray projection updates are allowed for candidate boundary clusters corresponding to representative points only when all of the above conditions are met. For candidate boundary clusters where representative points do not meet the execution conditions, their update operations are skipped to avoid incorrect inferences about invisible or occluded areas.

[0043] In this embodiment of the invention, let the current position vector of the UAV be... Then, represents the position vector of the point relative to the drone. for: When the distance between the representative point and the drone meets the requirement When this occurs, it indicates that the representative point is within the sensor's theoretical field of view, where, and These represent the preset minimum detection distance and the maximum detection distance, respectively.

[0044] Based on the currently occupied grid map, the current position of the drone... Pointing to the representative point Lines are used for line detection. If a known occupying voxel exists in the voxel sequence that the line passes through, the representative point is determined to be occluded; if no known occupying voxel exists, the representative point is determined to be unoccluded.

[0045] Along the line Voxel sequences obtained by sampling ,in Represents the line segment parameter, when At that time, the corresponding drone's current position ;when At that time, the corresponding representative point position . Indicates the segment along the line The sampled first m Line segment parameters, Represents line segment parameters If there exists a sampling point that satisfies the following conditions: If the point does not meet the unoccupied constraint, then it is considered that the representative point does not satisfy the unoccupied constraint. It indicates that space is occupied.

[0046] S4. Perform supplementary ray casting and line-of-sight detection on the candidate boundary clusters where representative points that meet the visibility execution conditions are located.

[0047] Step S4 includes the following sub-steps S41~S42: S41. Based on the current location of the drone As the starting point of the ray, it points towards a representative point that satisfies the visibility execution condition. At least one supplementary ray is emitted in the direction of the emission.

[0048] In this embodiment of the invention, the unit direction vector of the supplementary ray It can be represented as: S42. Along the propagation path of the supplementary ray, perform line-of-sight detection based on the occupied grid map. When no known occupied voxel is detected on the path of the supplementary ray, determine that the direction corresponding to the supplementary ray is a theoretically visible and unobstructed direction. When an obstructing voxel is detected on the path of the supplementary ray, terminate the update operation of the supplementary ray.

[0049] In this embodiment of the invention, for candidate boundary clusters that satisfy the visibility execution condition, a parameterized ray is constructed along the supplementary ray direction. : in, This represents the distance parameter along the direction of the supplementary ray. This indicates the update length of the supplementary ray, i.e., the maximum update range.

[0050] Follow the virtual ray at a preset step size Perform discrete sampling: in, Indicates the first The position vector of each discrete sampling point Indicates the sequence number of the discrete sampling point. Indicates the first The supplementary ray is at the preset update length The corresponding total number of sampling steps within the range satisfies , This indicates rounding down to the nearest integer.

[0051] For each discrete sampling point, a line-of-sight detection is performed on the map voxel. When any voxel corresponding to a discrete sampling point is determined to be a known occupied voxel, the update operation of the supplementary ray is terminated. When no known occupied voxel is detected along the supplementary ray path, the direction is determined to be a theoretically visible and unobstructed direction, allowing the subsequent free space inference update stage to proceed.

[0052] It should be noted that the supplementary raycasting update in this embodiment of the invention differs from the traditional raycasting update based on real laser echoes. Traditional raycasting typically uses the actual measured ray and its echo endpoint as the basis for map updates; however, this embodiment of the invention does not use the actual echo endpoint as the sole prerequisite for updates. Instead, for candidate boundary clusters, in the absence of valid echoes, it first determines whether they meet the supplementary update conditions based on the sensor's theoretical field of view and the occlusion relationships in the currently occupied grid map. Supplementary raycasting is triggered only for theoretically visible and unoccupied candidate areas, and free space inference updates are performed for the corresponding unknown areas.

[0053] S5. Update the free space inference of the supplementary rays detected by line of sight.

[0054] In this embodiment of the invention, a set of voxels within a preset length range are selected as update objects along the propagation path of the supplementary ray detected by line of sight, and the voxels are updated to a free space state.

[0055] Through the above updates, spatial regions that were previously unknown due to a lack of effective echoes but were geometrically within the theoretical visible range can be reasonably modified into free space.

[0056] S6. In response to the voxels in the neighborhood of the representative point in the candidate boundary cluster being updated to the free space state, the boundary state corresponding to the candidate boundary cluster is adjusted.

[0057] In this embodiment of the invention, the adjustment includes clearing boundary markers and eliminating false boundaries introduced by the non-echo direction.

[0058] S7. During the UAV mapping process, as the UAV's position changes, steps S1 to S6 are repeated periodically or as needed to continuously update the occupied grid map and boundary status as the mapping process progresses.

[0059] Through the aforementioned incremental update mechanism, reasonable corrections can be made to the direction without echoes, without relying on additional sensor inputs or real laser echoes, thus maintaining the continuity and consistency of the map state.

[0060] Through the above technical solutions, the embodiments of the present invention, while maintaining the traditional echo-driven ray update mechanism, introduce a supplementary ray projection update method based on visibility analysis, enabling the UAV lidar mapping system to conservatively update the theoretically visible area in the absence of effective echoes, effectively eliminating false boundaries and improving the accuracy and completeness of the occupied grid map.

[0061] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A UAV exploration mapping method based on supplementary ray projection updates, characterized in that, Includes the following steps: S1. Based on the occupancy grid map of the current UAV equipped with LiDAR, extract candidate boundary voxels and cluster them to obtain candidate boundary clusters; S2. For each candidate boundary cluster, select at least one representative point to describe the spatial location and orientation characteristics of the boundary cluster; S3. Perform visibility condition judgment on the representative points of each candidate boundary cluster; S4. Perform supplementary ray casting and line-of-sight detection on the candidate boundary clusters where representative points that meet the visibility execution conditions are located; S5. Update the free space inference of the supplementary rays detected by line of sight; S6. In response to the voxels in the neighborhood of the representative point in the candidate boundary cluster being updated to the free space state, the boundary state corresponding to the candidate boundary cluster is adjusted. S7. During the UAV mapping process, as the UAV's position changes, steps S1 to S6 are repeated periodically or as needed to continuously update the occupied grid map and boundary status as the mapping process progresses.

2. The UAV exploration mapping method based on supplementary ray projection update according to claim 1, characterized in that, Step S1 includes the following sub-steps: S11. Based on the updated grid map of the drone equipped with the lidar in the current frame, incremental detection is performed on the boundary state, and voxels located at the boundary between free space and unknown space are extracted as candidate boundary voxels. S12. Based on the spatial connectivity between voxels, the candidate boundary voxels are clustered to form several candidate boundary clusters.

3. The UAV exploration mapping method based on supplementary ray projection update according to claim 1 or 2, characterized in that, Each candidate boundary cluster represents a spatial region that has not been updated or has been determined to be unknown in the current map state.

4. The UAV exploration mapping method based on supplementary ray projection update according to claim 1, characterized in that, In step S2, the representative point is selected as the geometric centroid of the candidate boundary cluster, or one or more boundary voxel points selected from the candidate boundary cluster.

5. The UAV exploration mapping method based on supplementary ray projection update according to claim 1, characterized in that, The visibility execution condition in step S3 is: the representative point is located within the theoretical field of view of the lidar sensor, and according to the current occupied grid map, there is no occlusion structure composed of known occupied voxels in the direction of the line connecting the current position of the UAV to the representative point.

6. The UAV exploration mapping method based on supplementary ray projection update according to claim 1, characterized in that, Step S4 includes the following sub-steps: S41. Using the current position of the UAV as the starting point of the ray, emit at least one supplementary ray toward the representative point that meets the visibility execution conditions; S42. Along the propagation path of the supplementary ray, perform line-of-sight detection based on the occupied grid map. When no known occupied voxel is detected on the supplementary ray path, determine that the direction corresponding to the supplementary ray is a theoretically visible and unobstructed direction. When an obstructing voxel is detected on the supplementary ray path, terminate the update operation of the supplementary ray.

7. The UAV exploration mapping method based on supplementary ray projection update according to claim 1, characterized in that, Step S5 specifically involves selecting a set of voxels within a preset length range as update objects along the propagation path of the supplementary ray detected by the line of sight, and updating the voxels to a free space state.

8. The UAV exploration mapping method based on supplementary ray projection update according to claim 1, characterized in that, The adjustment in step S6 includes clearing boundary markers and eliminating false boundaries introduced by the non-echo direction.

Images