An unmanned aerial vehicle autonomous exploration method based on voxel sphere and related equipment

Abstract

The application belongs to the technical field of unmanned aerial vehicle autonomous navigation and map construction, and particularly relates to an unmanned aerial vehicle autonomous exploration method based on voxel spheres and related equipment. Voxel point cloud data of environment point cloud is obtained through voxelization processing; a Euclidean distance field is calculated based on the voxel point cloud, a voxel point cloud set containing voxel-to-obstacle distance information is generated; a spatial constraint strategy is adopted to adaptively generate a topological map composed of voxel sphere nodes in combination with the current position of the unmanned aerial vehicle and preset safety parameters, each node representing a sphere space region and an exploration state of the voxel point cloud in the sphere space region; a path of the unmanned aerial vehicle to a target node is planned based on the node exploration state; the exploration states of the voxel point cloud and the nodes in the topological map are dynamically updated during movement; and finally, the nodes, states and topological structures are iteratively updated to realize autonomous exploration of the whole region. The method improves the unmanned aerial vehicle environment perception and path planning efficiency through voxelization and adaptive topological modeling.

Description

Technical Field

[0001] This invention belongs to the field of UAV autonomous navigation and map building technology, specifically involving a UAV autonomous exploration method and related equipment based on voxel spheres. Background Technology

[0002] With the rapid development of autonomous navigation and environmental perception technologies for unmanned aerial vehicles (UAVs), autonomous exploration of unknown environments has become a core capability for UAVs in performing tasks such as inspection, search and rescue, and mapping. Three-dimensional voxel maps, due to their ability to accurately represent environmental occupancy information, have become the mainstream environmental modeling method for autonomous UAV exploration. Furthermore, efficient methods for generating exploration nodes and quantifying information directly determine exploration efficiency and path planning performance.

[0003] Current mainstream technologies for exploring unknown environments mostly use discrete voxels or 3D grids as basic units, selecting exploration targets by calculating indicators such as the volume of local unknown regions and information gain. Some schemes use random sampling or uniform point distribution to generate exploration nodes, and combine them with topological graph structures to construct exploration paths; other methods rely on ESDF distance fields to ensure flight safety, using obstacle distance constraints to screen feasible exploration locations. These technologies can complete basic exploration tasks in simple scenarios, but they have significant limitations in complex branches and large-scale environments.

[0004] The existing technology has three main problems: First, discrete voxels lack high-level aggregation, resulting in fragmented exploration information representation and high complexity when calculating unknown information gain. Second, the generation of exploration nodes depends on a fixed number or random distribution, without adaptive adjustment based on the environmental structure, which easily leads to node redundancy or insufficient coverage. Third, the node position does not take into account both the safe zone and the exploration order, resulting in path fragmentation, inefficient exploration loops, and a significant increase in map construction time in large-scale scenarios. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and related equipment for autonomous exploration of unmanned aerial vehicles based on voxel spheres, which addresses the shortcomings of the prior art and solves the technical problem of incomplete exploration caused by static maps in traditional methods.

[0006] The objective of this invention is achieved through the following technical solutions: In a first aspect, the present invention provides an autonomous exploration method for unmanned aerial vehicles based on voxel spheres, comprising: Acquire environmental 3D LiDAR point cloud data, perform voxelization processing on the point cloud to obtain voxel point cloud data; Based on the voxel point cloud data, the Euclidean distance field in the environment is calculated, and a set of voxel point clouds containing Euclidean distance information from voxels to obstacles is generated. Based on the voxel point cloud set, the current location of the UAV, and preset safety parameters, a topology map composed of multiple voxel sphere nodes is generated through a spatial constraint strategy; each voxel sphere node is used to characterize the exploration status of the sphere spatial region and the voxel point cloud contained within the sphere spatial region; the spatial constraint strategy is used to adaptively determine the generation location and number of voxel sphere nodes according to the spatial attributes and distribution of the voxel point cloud. Based on the exploration status of each voxel sphere node in the topology map, a path is planned for the UAV to reach the target voxel sphere node; Control the drone to move along the path, and update the exploration status of the voxel point cloud and voxel sphere nodes in the topology map during the movement; Iteratively update the voxel sphere nodes, exploration status, and topology structure until all regions have been explored.

[0007] As a further improvement of the present invention, the voxel sphere node is defined by the following data structure: sphere center coordinates, sphere radius, globally unique index, node exploration status, voxel point cloud exploration status mapping table, number of unexplored voxel point clouds, and total number of voxel point clouds. The node exploration status is updated based on whether the ratio of the number of unexplored voxel point clouds associated with the voxel sphere node to the total number of voxel point clouds reaches a preset threshold.

[0008] As a further improvement of the present invention, a topological map composed of multiple voxel sphere nodes is generated through a spatial constraint strategy, including: In the unexplored environment voxel point cloud, a priority weight is calculated for each unexplored voxel point cloud, and the priority weight is used to characterize the priority as the new sphere generation center; The unexplored voxel point clouds are sorted according to the priority weights to obtain a candidate point cloud sequence; Based on the order of the candidate point cloud sequence, and under the condition of satisfying the preset spatial constraints, a new voxel sphere node is generated at the candidate point cloud location.

[0009] As a further improvement of the present invention, the formula for calculating the priority weight is as follows:

[0010] In the formula, Voxel point cloud Priority weights; , This is the weighting adjustment coefficient; The ESDF distance value for the voxel point cloud; Voxel point cloud The three-dimensional Euclidean distance between the drone and the other drone.

[0011] As a further improvement of the present invention, under the condition of satisfying preset spatial constraints, a new voxel sphere node is generated at the candidate point cloud location, including: For each candidate point cloud in the candidate point cloud sequence, determine whether there is a number of unexplored voxel point clouds that do not belong to any existing voxel sphere node within a preset radius centered on the candidate point cloud, which is greater than a preset threshold. If the value is greater than the preset threshold, a new voxel sphere node is generated at the candidate point cloud location, and unexplored voxel point clouds within the preset radius that do not belong to any existing voxel sphere are included in the newly generated voxel sphere node.

[0012] As a further improvement of the present invention, generating new voxel sphere nodes at candidate point cloud locations while satisfying preset spatial constraints further includes: For each candidate point cloud in the candidate point cloud sequence, determine whether the distance between the candidate point cloud and the center of at least one existing voxel sphere node is within a preset distance constraint interval, and whether the distance between the voxel point cloud on the path connecting the two and the obstacle meets the safety conditions. If the conditions are met, a new voxel sphere node is generated at the candidate point cloud location, and a connection edge is established between the newly generated voxel sphere node and the existing voxel sphere node in the topology map. As a further improvement of the present invention, when generating the topology map, the method also includes using a spatial index structure to manage the position information of the generated voxel sphere nodes in order to accelerate the query of neighboring nodes and collision detection; the spatial index structure is a KD tree.

[0013] Secondly, the present invention provides an autonomous exploration system for unmanned aerial vehicles based on voxel spheres, comprising: The environmental perception module is used to acquire three-dimensional lidar point cloud data of the environment, perform voxelization processing on the point cloud to obtain voxel point cloud data; based on the voxel point cloud data, calculate the Euclidean distance field in the environment, and generate a set of voxel point clouds containing Euclidean distance information from voxels to obstacles. The topology map generation module generates a topology map composed of multiple voxel sphere nodes based on the voxel point cloud set, the current position of the UAV, and preset safety parameters, using a spatial constraint strategy. Each voxel sphere node is used to represent the exploration status of the sphere spatial region and the voxel point cloud contained within the sphere spatial region. The spatial constraint strategy is used to adaptively determine the generation location and number of voxel sphere nodes according to the spatial attributes and distribution of the voxel point cloud. The path planning module is used to plan a path for the UAV to the target voxel sphere node based on the exploration status of each voxel sphere node in the topology map. The state update module is used to control the UAV to move along the path and update the exploration state of the voxel point cloud and voxel sphere nodes in the topology map during the movement; iteratively update the voxel sphere nodes, exploration state and topology map structure until all areas are explored.

[0014] Thirdly, the present invention provides a computer device, comprising: a processor and a computer-readable storage medium; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the above-described voxel-based autonomous exploration method for unmanned aerial vehicles.

[0015] Fourthly, the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and executed as described above for the voxel-based unmanned aerial vehicle autonomous exploration method.

[0016] The beneficial effects of this invention are as follows: This invention provides an autonomous exploration method for unmanned aerial vehicles (UAVs) based on voxel spheres. It generates a voxel point cloud set containing Euclidean distance information by voxelizing point cloud data. Based on this set and the UAV's current position and safety parameters, a voxel sphere node topology map representing the spatial region and exploration state of the sphere is adaptively generated using a spatial constraint strategy. This achieves direct conversion from point cloud data to a topology map, overcoming the technical problems of low point cloud data utilization and poor map construction efficiency in traditional path planning. The exploration state and spatial attributes associated with each voxel sphere node allow path planning to dynamically adjust target nodes based on the node's exploration state, compared to fixed-node mapping. The system improves the flexibility and environmental adaptability of path planning; the spatial constraint strategy adaptively determines the node generation location and number based on the voxel point cloud distribution, avoiding node redundancy or insufficient coverage. Compared with the preset fixed node strategy, it improves the efficiency of topology map construction and space utilization; iteratively updates the voxel point cloud and node exploration status in the topology map to achieve dynamic optimization of the exploration process, solving the problem of incomplete exploration caused by static maps in traditional methods; the synergistic effect of the above technical features enables UAVs to achieve efficient, safe, and adaptive exploration path planning and motion control in complex 3D environments, significantly improving the integrity of environmental exploration and the real-time performance of path planning compared with existing technologies. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of an embodiment of the present invention: an autonomous exploration method for unmanned aerial vehicles based on voxel spheres. Figure 2 This is a schematic diagram of the uniform expansion of voxel sphere nodes in the explorable region according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a three-dimensional ESDF voxel point cloud generated in the Gazebo environment according to an embodiment of the present invention; Figure 4 This is a voxel sphere map generated in a crossroads environment according to an embodiment of the present invention; Figure 5 This is a complete topological voxel sphere map of an embodiment of the present invention; Figure 6 These are the fitting results of curves with different complexity in embodiments of the present invention; Figure 7 This is a schematic diagram of the internal structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives and technical solutions of this invention clearer and easier to understand, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0020] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. The described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] Example 1 Traditional UAV autonomous exploration methods suffer from weak node exploration information representation capabilities, low exploration area coverage efficiency, and high algorithm computational complexity. Specifically, isolated voxel point clouds cannot be aggregated to form regional information representations, node generation positions and numbers lack adaptive constraints, and path planning and map construction are time-consuming, making it difficult to meet the needs of efficient autonomous exploration in large-scale complex environments. This application addresses these technical problems by proposing a voxel-based autonomous exploration method for UAVs. This method combines voxelization processing, Euclidean distance field calculation, spatial constraint strategies, and iterative update mechanisms to achieve rapid, safe, and full-coverage autonomous exploration of complex environments by UAVs.

[0022] Autonomous exploration methods for unmanned aerial vehicles based on voxel spheres, such as Figure 1 As shown, the specific implementation method is as follows.

[0023] S1: Acquire environmental 3D LiDAR point cloud data, perform voxelization processing on the point cloud to obtain voxel point cloud data.

[0024] Specifically, in some embodiments of this application, the UAV is equipped with a lidar sensor to acquire radar point cloud information in the environment. After acquiring the raw point cloud data, the point cloud data is voxelized. The voxelization process includes: discretizing the continuous space into uniform three-dimensional grid cells, each grid cell being called a voxel; determining the occupancy state of each voxel based on the distribution of point cloud data within the voxel; updating the obstacle occupancy state of each voxel in the environment using a ray projection method based on the ray path of the lidar, thereby obtaining a three-dimensional voxel map of the environment. This voxel map contains information on whether obstacles exist at the location of each voxel in the environment. By performing point cloudization on each voxel in the voxel map, that is, extracting the center point coordinates of each voxel as the coordinates of the voxel point cloud, the voxel point cloud data is finally obtained. Voxelization can convert sparse and irregularly distributed raw point cloud data into a regularized and structured voxel representation, facilitating subsequent distance field calculations and the application of spatial constraint strategies.

[0025] Optionally, as an example, the voxel resolution of the voxelization process can be preset according to the environmental complexity and the UAV's computing resources, for example, set to 0.2 meters, 0.5 meters or 1.0 meters.

[0026] S2: Based on voxel point cloud data, calculate the Euclidean distance field in the environment and generate a set of voxel point clouds containing Euclidean distance information from voxels to obstacles.

[0027] like Figure 3As shown, the Euclidean distance field uses the Euclidean Signed Distance Field (ESDF) to represent the Euclidean distance value from each voxel point cloud to the nearest obstacle. The voxel point cloud set contains voxel point clouds from all unexplored areas, and each voxel point cloud carries a corresponding ESDF distance value. Specifically, a breadth-first search (BFS) method is used to propagate distance information outward from the obstacle voxels, updating the Euclidean distance from each voxel to the nearest obstacle, thus obtaining the ESDF. The value of each voxel in the ESDF represents the Euclidean distance from that voxel's location to the nearest obstacle. Each voxel in the ESDF is converted into a point cloud, and the center point coordinates and their corresponding ESDF distance values ​​are extracted to generate a voxel point cloud set containing Euclidean distance information from the voxel to the obstacle. Each voxel point cloud in this set has globally unique coordinate information and carries an ESDF distance value: the larger the distance value, the closer the voxel point cloud is to the center of the passable area, and the safer it is; the smaller the distance value, the closer it is to the obstacle, and the lower the safety. This distance information is used to evaluate the safety and centrality of node positions during the subsequent voxel sphere node generation process.

[0028] Optionally, as an embodiment, a truncated distance can be used when calculating the Euclidean distance field, that is, only the voxel distance values ​​within a certain range around the obstacle are calculated, and the voxel distance values ​​outside this range are set to a preset maximum value to reduce the amount of calculation.

[0029] S3: Based on the voxel point cloud set, the current position of the UAV, and preset safety parameters, a topology map composed of multiple voxel sphere nodes is generated through a spatial constraint strategy.

[0030] Voxel sphere nodes are used to characterize the exploration status of the sphere spatial region and the voxel point cloud contained within the sphere spatial region. The spatial constraint strategy adaptively determines the generation location and number of voxel sphere nodes based on the spatial attributes and distribution characteristics of the voxel point cloud. The preset safety parameters include the UAV safety radius, voxel sphere radius, voxel sphere center distance constraint interval, and minimum generation threshold for sphere point cloud.

[0031] Specifically, such as Figure 2 As shown, the definition process of a voxel sphere is as follows: Define a voxel sphere B in the definition space as a quintuple, as shown in the following equation:

[0032] In the formula, Represents the three-dimensional coordinates of the center of the sphere in space; This represents the radius of the sphere; the sphere itself will enclose some voxel point clouds. All three conditions must be met: .

[0033] Point cloud of each voxel Two exploration states are defined: unexplored (0) and explored (1). Since each voxel point cloud in space has globally unique coordinate information, the exploration state is mapped through a function. Perform the calculation. The domain is voxel point cloud The coordinates, with a range representing the two exploration states of the voxel point cloud, such as... .

[0034] The voxel sphere itself also contains an exploration state. When the number of explored voxel point clouds in the sphere reaches a certain proportion At that time, the sphere as a whole will change from an unexplored (0) state to an explored (1) state, that is:

[0035] In the formula, Represents a set of voxel point clouds The number of point clouds it possesses.

[0036] The data structure of the voxel sphere is shown in Table 1.

[0037] Table 1 Data Structure of Voxel Spheres

[0038] The pos and radius attributes in Table 1 correspond to the voxel sphere quintuples, respectively. In , In the mapping function In its implementation, the globally unique nature of voxel point cloud coordinates is fully utilized, employing a hash function (hash table). To integrate with the previously discussed topology map concepts, each voxel sphere is assigned a globally unique index to uniquely identify it within the space. The number of unexplored voxel points (unexplored_num) and the total number of voxel points (total_num) are introduced to describe the exploration completion rate of each voxel sphere. The exploration completion rate is defined as follows. The mathematical expression is as follows:

[0039] In the formula, This indicates the number of unexplored voxel point clouds within the voxel sphere. This represents the total number of voxel point clouds contained within the voxel sphere. When The value has reached the exploration completion threshold. When the voxel sphere is in the exploration state, it will change from unexplored (0) to explored (1).

[0040] Specifically, voxel sphere nodes are defined using a fixed data structure, which includes the sphere's center coordinates, radius, globally unique index, node exploration status, a voxel point cloud exploration status mapping table, the number of unexplored voxel point clouds, and the total number of voxel point clouds. The node exploration status is divided into two types: explored and unexplored. It is updated based on whether the ratio of the number of unexplored voxel point clouds associated with the voxel sphere node to the total number of voxel point clouds reaches an exploration completion threshold. When the ratio reaches the threshold, the node's exploration status changes from unexplored to explored. The voxel point cloud exploration status is divided into two types: unexplored (0) and explored (1). A hash table is used to construct the mapping relationship between voxel point cloud coordinates and exploration status, and the global uniqueness of voxel point cloud coordinates enables fast querying of exploration status.

[0041] In some embodiments of this application, the spatial constraint strategy includes a priority-weighted sphere position selection strategy and a voxel density-based adaptive sphere number strategy. These two strategies work together to determine the generation location and number of voxel sphere nodes.

[0042] Specifically, the process of generating a topology map using spatial constraint strategies is as follows: In the unexplored environment voxel point cloud, a priority weight is calculated for each unexplored voxel point cloud. This weight characterizes the priority of that voxel point cloud as a new sphere generation center. The formula for calculating the priority weight is:

[0043] In the formula, Voxel point cloud Priority weights; , This is the weighting adjustment coefficient; The ESDF distance value for the voxel point cloud; Voxel point cloud The three-dimensional Euclidean distance between the drone and the other drone.

[0044] An adaptive sphere numbering strategy based on voxel density iterates through the candidate point cloud sequence and determines whether the number of unexplored voxel point clouds not belonging to any existing voxel sphere node within a preset radius centered on a candidate point cloud is greater than the minimum generation threshold for sphere point clouds. If the number is greater than the threshold, the generation quantity condition is met. Simultaneously, it checks whether the distance between the center of a candidate point cloud and at least one existing voxel sphere node is within a preset distance constraint interval, and whether the ESDF distance values ​​of the voxel point clouds on the path connecting the two are greater than the UAV's safe radius. If these conditions are met, the position and safety constraints are satisfied. When the quantity, position, and safety constraints are all satisfied, a new voxel sphere node is generated at the candidate point cloud location, the corresponding unexplored voxel point cloud is included in the new node, and a connection edge is established between the new node and existing nodes in the topology map. Intersections are allowed between voxel spheres, and the voxel point clouds in the intersection portion uniquely belong to a single voxel sphere node to avoid duplicate counting. The final generation effect is as follows. Figure 4 As shown.

[0045] In some embodiments of this application, when generating the topology map, a KD-tree spatial index structure is used to manage the position information of the generated voxel sphere nodes. The KD-tree is used to accelerate neighbor node queries and collision detection, thereby improving the efficiency of topology map construction. In the initial state, an initial voxel sphere node is generated with the voxel point cloud where the UAV is currently located as the center, serving as the starting node of the topology map.

[0046] Specifically, an empty KD-tree is initialized. For each voxel sphere node generated, its center coordinates are inserted into the KD-tree. When querying other voxel spheres within a certain distance range of a given voxel point cloud, the nearest neighbor search function of the KD-tree is used to quickly obtain the result, thereby accelerating the algorithm execution.

[0047] Optionally, as an embodiment, the specific algorithm flow for generating voxel spheres is as follows: Initialize an empty graph and an empty KD-tree; generate the first voxel sphere with the voxel point cloud where the UAV is currently located as the center and the UAV's safe radius as the radius, update the sphere attributes, add the sphere as a node to the graph, and add its center coordinates to the KD-tree; traverse the voxel point cloud set, calculate the priority weight of each voxel point cloud, and obtain a descending-ordered candidate point cloud sequence; sequentially traverse the candidate point cloud sequence, and for each candidate point cloud, use the KD-tree to query whether it already has a voxel sphere within the distance constraint interval. If not, determine whether the number of unexplored voxel point clouds that do not belong to other spheres contained in the sphere with the point cloud as the center and radius as the radius is greater than the minimum threshold for generating sphere point clouds. If the conditions are met, initialize an empty candidate edge set, traverse the existing spheres in the graph, and determine whether the line connecting the current point cloud to the center of an existing sphere satisfies the condition that the edge length is within [D_min, D_max] and the ESDF distance value of all voxel point clouds on the edge is greater than the safe radius of the UAV. If the conditions are met, add the edge to the candidate edge set. After the traversal, if the candidate edge set is not empty, generate a new voxel sphere with the point cloud as the center, update the attributes, add the new sphere as a node and the candidate edge set as an edge to the graph, and insert the coordinates of the new sphere's center into the KD tree. Repeat until the candidate point cloud sequence is traversed, and return the voxel sphere map.

[0048] S4: Based on the exploration status of each voxel sphere node in the topology map, plan the path for the UAV to the target voxel sphere node.

[0049] The target voxel sphere node is the node in the topology map that is unexplored and has the highest exploration value. Exploration value is quantified by the ratio of the number of unexplored voxel point clouds within the node to the total number of voxel point clouds. Path planning must ensure that all voxel point clouds along the path meet safe passage conditions. Specifically, the exploration completion rate of a voxel sphere node is calculated by dividing the number of unexplored voxel point clouds by the total number of voxel point clouds. The exploration completion rate directly quantifies the amount of unknown information that the UAV can obtain by exploring this node, serving as the core basis for path planning. Unexplored nodes with high exploration completion rates are prioritized as target nodes. Figure 5 As shown.

[0050] S5: Control the drone to move along the path and update the exploration status of voxel point cloud and voxel sphere nodes in the topology map during the movement.

[0051] In some embodiments of this application, the UAV moves along the path planned in S4. During the movement, the UAV continuously acquires new environmental point cloud information through lidar. For newly observed areas, the exploration state of the corresponding voxel point cloud is updated: through a hash table mapping function, the exploration state corresponding to the observed voxel point cloud coordinates is updated from unexplored (0) to explored (1). For each voxel sphere node, the exploration state of all voxel point clouds inside it is re-counted, and the number of unexplored voxel point clouds of the sphere node is updated. When the exploration completion degree P_unexplored of the sphere node reaches a preset threshold, the node exploration state of the sphere node is updated from unexplored (0) to explored (1).

[0052] S6: Iteratively update the voxel sphere nodes, exploration status, and topology structure until all regions have been explored.

[0053] The iterative process continuously executes spatial constraint strategies, path planning, and state update operations, cyclically expanding the exploration area until no new unexplored voxel nodes are generated. Specifically, the iterative process forms a cyclical flow of path search, motion exploration, and map update, with voxel nodes always distributed in safe positions at the center of the explorable area, avoiding obstacle expansion zones. When all voxel nodes have switched to the explored state and no new unexplored voxel nodes are generated, the entire environment exploration is considered complete.

[0054] To verify the performance of the above method, this embodiment uses simulation experiments to illustrate the fitting results. In the simulation experiments, the total number of spheres generated in each increment and the total time consumed in the generation process are recorded (no data is recorded when no spheres are generated). Based on this, different complexity curves are used, including a first-order logarithmic curve. conic section and cubic curves The data is fitted, and the fitting result is used to... Figure 6 As shown, the first-order logarithmic curve can be seen. and quadratic curve The determination coefficients are relatively high, reaching 0.974 and 0.987 respectively. Based on this, it can be inferred that the time complexity of the voxel sphere map path generation algorithm decreases as the number of generated voxel spheres increases. ~ This trend indicates that the voxel sphere map path generation algorithm has better generation efficiency than random search algorithms, especially in large-scale environmental exploration tasks, which can significantly reduce the time spent by UAVs in building exploration maps.

[0055] Example 2 This embodiment provides an autonomous exploration system for unmanned aerial vehicles based on voxel spheres, including: The environmental perception module is used to acquire three-dimensional lidar point cloud data of the environment, perform voxelization processing on the point cloud to obtain voxel point cloud data; based on the voxel point cloud data, calculate the Euclidean distance field in the environment, and generate a set of voxel point clouds containing Euclidean distance information from voxels to obstacles. The topology map generation module generates a topology map composed of multiple voxel sphere nodes based on the voxel point cloud set, the current position of the UAV, and preset safety parameters, using a spatial constraint strategy. Each voxel sphere node is used to represent the exploration status of the sphere spatial region and the voxel point cloud contained within the sphere spatial region. The spatial constraint strategy is used to adaptively determine the generation location and number of voxel sphere nodes according to the spatial attributes and distribution of the voxel point cloud. The path planning module is used to plan a path for the UAV to the target voxel sphere node based on the exploration status of each voxel sphere node in the topology map. The state update module is used to control the UAV to move along the path and update the exploration state of the voxel point cloud and voxel sphere nodes in the topology map during the movement; iteratively update the voxel sphere nodes, exploration state and topology map structure until all areas are explored.

[0056] Specific limitations regarding the voxel-based autonomous UAV exploration system can be found in the above description of the voxel-based autonomous UAV exploration method; the corresponding technical effects are equivalent and will not be repeated here. Each module in the aforementioned voxel-based autonomous UAV exploration system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, allowing the processor to call and execute the corresponding operations of each module.

[0057] Example 3 Figure 7 An internal structural diagram of a computer device is shown in one embodiment. This computer device may specifically be a terminal or a server. Figure 7 As shown, the computer device includes a processor, memory, network interface, display, camera, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a voxel-based autonomous exploration method for unmanned aerial vehicles. The display screen can be an LCD screen or an e-ink display. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0058] As will be understood by those skilled in the art, computer equipment Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computing device may include more or fewer components than those shown in the figure, or combine certain components, or have the same component arrangement.

[0059] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described above.

[0060] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0061] In summary, the present application provides a method, system, computer device, and storage medium for autonomous UAV exploration based on voxel spheres. This method aggregates discrete voxel point clouds into geometric spheres using voxel sphere nodes, setting an exploration state for each voxel point cloud. The exploration value is obtained by counting the number of unexplored voxel point clouds within the sphere, and the exploration completion rate reveals the exploration status of the region. Compared to traditional methods that calculate the volume of unexplored areas, this application significantly reduces algorithm complexity. The spatial constraint strategy adaptively determines the generation position and number of sphere nodes based on the priority weights of the voxel point clouds (combining ESDF distance and UAV distance) and voxel density, achieving coverage of more exploration areas with as few nodes as possible and solving the problems of node position and quantity uncertainty. The sphere nodes are evenly distributed in the center of the safe and passable area, with uniformity ensured by distance constraint intervals and path safety guaranteed by safety conditions.

[0062] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0063] The above-described embodiments are merely preferred embodiments of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. A method for autonomous exploration of unmanned aerial vehicles based on voxel sphere, characterized in that, include: Acquire environmental 3D LiDAR point cloud data, perform voxelization processing on the point cloud to obtain voxel point cloud data; Based on the voxel point cloud data, the Euclidean distance field in the environment is calculated, and a set of voxel point clouds containing Euclidean distance information from voxels to obstacles is generated. Based on the voxel point cloud set, the current location of the UAV, and preset safety parameters, a topology map composed of multiple voxel sphere nodes is generated through a spatial constraint strategy; each voxel sphere node is used to characterize the exploration status of the sphere spatial region and the voxel point cloud contained within the sphere spatial region; the spatial constraint strategy is used to adaptively determine the generation location and number of voxel sphere nodes according to the spatial attributes and distribution of the voxel point cloud. Based on the exploration status of each voxel sphere node in the topology map, a path is planned for the UAV to reach the target voxel sphere node; Control the drone to move along the path, and update the exploration status of the voxel point cloud and voxel sphere nodes in the topology map during the movement; Iteratively update the voxel sphere nodes, exploration status, and topology structure until all regions have been explored.

2. The voxel sphere based autonomous exploration method for UAVs according to claim 1, wherein, The voxel sphere node is defined by the following data structure: sphere center coordinates, sphere radius, globally unique index, node exploration status, voxel point cloud exploration status mapping table, number of unexplored voxel point clouds, and total number of voxel point clouds. The node exploration status is updated based on whether the ratio of the number of unexplored voxel point clouds associated with the voxel sphere node to the total number of voxel point clouds reaches a preset threshold. 3.The method of claim 1, wherein, A topology map consisting of multiple voxel sphere nodes is generated using a spatial constraint strategy, including: In the unexplored environment voxel point cloud, a priority weight is calculated for each unexplored voxel point cloud, and the priority weight is used to characterize the priority as the new sphere generation center; The unexplored voxel point clouds are sorted according to the priority weights to obtain a candidate point cloud sequence; Based on the order of the candidate point cloud sequence, and under the condition of satisfying the preset spatial constraints, a new voxel sphere node is generated at the candidate point cloud location.

4. The voxel sphere based autonomous exploration method for UAVs according to claim 3, wherein, The formula for calculating the priority weight is: In the formula, is a priority weight of the voxel point cloud; is a priority weight of the voxel point cloud; , is a weight adjustment coefficient; is an ESDF distance value of the voxel point cloud; is a priority weight of the voxel point cloud; is a three-dimensional Euclidean distance between the UAV and the voxel point cloud.

5. The autonomous exploration method for unmanned aerial vehicles based on voxel spheres according to claim 3, characterized in that, Under the condition of satisfying the preset spatial constraints, new voxel sphere nodes are generated at the candidate point cloud locations, including: For each candidate point cloud in the candidate point cloud sequence, determine whether there is a number of unexplored voxel point clouds that do not belong to any existing voxel sphere node within a preset radius centered on the candidate point cloud, which is greater than a preset threshold. If the value is greater than the preset threshold, a new voxel sphere node is generated at the candidate point cloud location, and unexplored voxel point clouds within the preset radius that do not belong to any existing voxel sphere are included in the newly generated voxel sphere node.

6. The autonomous exploration method for unmanned aerial vehicles based on voxel spheres according to claim 5, characterized in that, In addition to generating new voxel sphere nodes at candidate point cloud locations while satisfying preset spatial constraints, the process also includes: For each candidate point cloud in the candidate point cloud sequence, determine whether the distance between the candidate point cloud and the center of at least one existing voxel sphere node is within a preset distance constraint interval, and whether the distance between the voxel point cloud on the path connecting the two and the obstacle meets the safety conditions. If the conditions are met, a new voxel sphere node is generated at the candidate point cloud location, and a connection edge is established between the newly generated voxel sphere node and the existing voxel sphere node in the topology map.

7. The autonomous exploration method for unmanned aerial vehicles based on voxel spheres according to claim 1, characterized in that, When generating the topology map, the method also includes using a spatial index structure to manage the position information of the generated voxel sphere nodes in order to accelerate neighbor node queries and collision detection; the spatial index structure is a KD tree.

8. An autonomous exploration system for unmanned aerial vehicles based on voxel spheres, characterized in that, include: The environmental perception module is used to acquire 3D LiDAR point cloud data of the environment, perform voxelization processing on the point cloud to obtain voxel point cloud data. Based on the voxel point cloud data, the Euclidean distance field in the environment is calculated, and a set of voxel point clouds containing Euclidean distance information from voxels to obstacles is generated. The topology map generation module generates a topology map composed of multiple voxel sphere nodes based on the voxel point cloud set, the current position of the UAV, and preset safety parameters, using a spatial constraint strategy. Each voxel sphere node is used to represent the exploration status of the sphere spatial region and the voxel point cloud contained within the sphere spatial region. The spatial constraint strategy is used to adaptively determine the generation location and number of voxel sphere nodes according to the spatial attributes and distribution of the voxel point cloud. The path planning module is used to plan a path for the UAV to the target voxel sphere node based on the exploration status of each voxel sphere node in the topology map. The state update module is used to control the UAV to move along the path and update the exploration state of the voxel point cloud and voxel sphere nodes in the topology map during the movement; iteratively update the voxel sphere nodes, exploration state and topology map structure until all areas are explored.

9. A computer device, characterized in that, include: Processor and computer-readable storage media; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the voxel-based unmanned aerial vehicle autonomous exploration method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1 to 7, for use in autonomous exploration of unmanned aerial vehicles based on voxel spheres.

Images