Autonomous positioning and obstacle crossing method and device for unmanned aerial vehicle group, electronic equipment and medium

Abstract

The embodiment of the invention provides an unmanned aerial vehicle group autonomous positioning and obstacle crossing method and device, electronic equipment and a medium, and the method comprises the steps: obtaining the local pose of a single unmanned aerial vehicle based on VINS-Fusion, generating a raster map through combining point cloud data, and converting the raster map into an ESDF map according to a distance transformation algorithm; based on the ESDF map, converting the pose information of each unmanned aerial vehicle into a global pose by using a TF mechanism in an ROS, and obtaining the global pose information of the unmanned aerial vehicle cluster; determining the number of road sign points through a cost function of an improved A * algorithm and a topological equivalence algorithm; according to the global pose information of the unmanned aerial vehicle cluster and the number of road sign points, adaptively allocating obstacle crossing time and speed to each unmanned aerial vehicle based on the environment information, and generating an optimal course and path; and initializing the takeoff height, the maximum flight speed and the expected target attitude of each unmanned aerial vehicle cluster, and completing autonomous obstacle crossing and exploration of the unmanned aerial vehicle clusters in combination with the optimal course and path.

Description

Technical Field

[0001] This document relates to the technical field of obstacle avoidance for drone swarms, and in particular to a method, device, electronic equipment and medium for autonomous positioning and obstacle surmounting for drone swarms. Background Art

[0002] Drones have been widely used in civil, industrial and military fields. Multi-drone clusters will be able to achieve more complex group writing tasks through information exchange and sharing combined with flight control, distributed positioning and other technologies. Compared with a single drone system, the cluster system will be more robust in performing tasks and more efficient.

[0003] For complex scenarios, how multiple drones can safely and efficiently cross obstacles will be the focus of research. To realize the operation of the multi-drone system, each drone needs to know its global position in the system, which is usually based on external positioning systems such as GNSS. However, in an environment where global satellite navigation signals are missing, such systems will fail.

[0004] In addition, the complex and changing environment will also affect the drone's efficient crossing of obstacles, which brings challenges to mission execution. How to build a more reasonable path planning solution will be an urgent problem to be solved. Summary of the invention

[0005] The purpose of the present invention is to provide a method, device, electronic equipment and medium for autonomous positioning and obstacle crossing of a drone swarm, aiming to solve the above-mentioned problems in the prior art.

[0006] The present invention provides a method for autonomous positioning and obstacle crossing of a drone group, comprising:

[0007] Based on VINS-Fusion, the local pose of a single UAV is obtained, and a raster map is generated by combining the point cloud data. The raster map is then converted into an ESDF map based on the distance transformation algorithm.

[0008] Based on the ESDF map, the TF mechanism in ROS is used to convert the pose information of each drone into a global pose, and the global pose information of the drone cluster is obtained;

[0009] Determine the number of landmarks by improving the cost function of the A* algorithm and the topological equivalence algorithm;

[0010] According to the global posture information of the drone cluster and the number of landmarks, the obstacle crossing time and speed are adaptively allocated to each drone based on environmental information to generate the optimal heading and path;

[0011] Initialize the take-off height, maximum flight speed, and expected target posture of each drone cluster, and combine the optimal heading and path to complete the autonomous obstacle crossing and exploration of the drone cluster.

[0012] The present invention provides an autonomous positioning and obstacle-crossing device for a drone group, comprising:

[0013] Generate a conversion module, which is used to obtain the local pose of a single UAV based on VINS-Fusion, generate a raster map based on point cloud data, and convert the raster map into an ESDF map based on the distance transformation algorithm;

[0014] An acquisition module is used to convert the pose information of each drone into a global pose based on the ESDF map by using the TF mechanism in ROS to acquire the global pose information of the drone cluster;

[0015] A determination module, used for determining the number of landmark points by improving the cost function of the A* algorithm and the topological equivalence algorithm;

[0016] The optimal heading and path generation module is used to adaptively allocate the obstacle crossing time and speed of each drone based on the global posture information of the drone cluster and the number of landmarks, and based on the environmental information to generate the optimal heading and path;

[0017] The autonomous obstacle crossing and exploration module is used to initialize the take-off height, maximum flight speed, and expected target posture of each drone cluster, and complete the autonomous obstacle crossing and exploration of the drone cluster in combination with the optimal heading and path.

[0018] An embodiment of the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the above-mentioned method for autonomous positioning and obstacle surmounting of a swarm of unmanned aerial vehicles.

[0019] An embodiment of the present invention further provides a computer-readable storage medium, on which a program for implementing information transmission is stored. When the program is executed by a processor, the steps of the above-mentioned method for autonomous positioning and obstacle surmounting of a swarm of unmanned aerial vehicles are implemented.

[0020] By adopting the embodiment of the present invention, an ESDF map is constructed, the A* algorithm is improved and redundant target points are eliminated, and the flight speed of the drone is adaptively allocated based on the environment, the efficiency of the drone's obstacle avoidance and autonomous exploration is improved, and the safety risk is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0022] Figure 1 is a flow chart of a method for autonomous positioning and obstacle surmounting of a drone swarm according to an embodiment of the present invention;

[0023] Figure 2 is a flow chart of ESDF map update according to an embodiment of the present invention;

[0024] Figure 3 is a topological equivalent schematic diagram of an embodiment of the present invention;

[0025] Figure 4 is a schematic diagram of a drone swarm autonomous positioning and obstacle surmounting device according to an embodiment of the present invention;

[0026] Figure 5 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will be combined with the drawings in one or more embodiments of this specification to clearly and completely describe the technical solutions in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this document.

[0028] Method Embodiment

[0029] According to an embodiment of the present invention, a method for autonomous positioning and obstacle crossing of a drone swarm is provided. Figure 1 is a flow chart of the autonomous positioning and obstacle surmounting method of a drone swarm according to an embodiment of the present invention. Figure 1 As shown, the autonomous positioning and obstacle crossing method of a drone swarm according to an embodiment of the present invention specifically includes:

[0030] Step S101, based on VINS-Fusion, obtain the local pose of a single UAV, generate a grid map in combination with point cloud data, and convert the grid map into an ESDF map according to a distance transformation algorithm; specifically includes:

[0031] Convert the point cloud information from the drone coordinate system to the world coordinate system, convert the point cloud into voxels, and add the voxel information to the insertQueue and deleteQueue queues;

[0032] Initialize the ESDF map, merge the two queues into the updateQueue queue, and use the breadth-first algorithm to update all possible changed voxel data;

[0033] The updated voxel data is combined with the depth information, and the depth information inside the obstacle is defined as negative, and finally the ESDF map is generated.

[0034] Step S102, based on the ESDF map, the TF mechanism in ROS is used to convert the pose information of each drone into a global pose, and obtain the global pose information of the drone cluster; specifically includes:

[0035] The TF mechanism in ROS is adopted, and the TF library is used to convert the relationship between different coordinate systems in different drone systems. Among them, the TF library uses the publish and subscribe mechanism to allow coordinate transformation information to be transmitted between multiple nodes.

[0036] Step S103, determining the number of landmark points by improving the cost function of the A* algorithm and the topological equivalence algorithm; specifically includes:

[0037] The cost item of the node of the improved A* algorithm is:

[0038] f(n)=g(n)+h(n)+α·c(n) Formula 1;

[0039] Among them, g(n) and h(n) represent the actual cost from the initial node to node n and the estimated cost from node n to the target node, respectively. α·c(n) is a newly added security cost item, which is used to reflect the security level of node n. Here:

[0040]

[0041] Among them, esdf(n) represents the distance from node n to the nearest obstacle surface. This value is directly obtained based on the ESDF map. For more dangerous nodes, the larger c(n) is, the greater the overall cost of selecting the node. In addition, grids with esdf(n) less than 0 are considered inaccessible areas during the search process, so c(n) is always positive. α is a non-negative weight factor used to adjust the emphasis on path safety during planning. The smaller α is, the closer the planned path is to the shortest path, but the safety level will also decrease. Increasing α can improve the safety level of the path, and the length of the planned path will also increase. When α is 0, the improved A* algorithm will degenerate into the traditional A* algorithm.

[0042] For any two paths τ1(s),τ2(s),,s∈[0,1], if τ1(0)=τ2(0),τ1(1)=τ2(1), then for any s, the line segment If both paths are collision-free, then the paths τ1(s) and τ2(s) are topologically equivalent, that is, when a one-to-one mapping is performed on the points on the two paths, if the line between any pair of corresponding points does not collide with an obstacle, then the two paths are topologically equivalent;

[0043] Set the starting point S as the key landmark point K, and then traverse the subsequent landmark points in sequence. Assume that the next point of landmark point K is M. If the line segment If no obstacle is hit, then calculate Distance to the nearest obstacle Calculate the weight of point M This weight is a trade-off between the sparsity of the points and the safety of the path after extracting the key landmark points. The point with the largest weight is further selected as the next key point, and the above process is repeated;

[0044] For each point O in K and G, calculate the weight In each iteration, the point set O with the smallest weight w will be used as the key landmark point set, where d KO Represents a line segment Distance to the nearest obstacle d OG Represents a line segment Distance to the nearest obstacle |KO| represents the Euclidean distance from landmark point K to landmark point O, and |OG| represents the Euclidean distance from landmark point O to landmark point G;

[0045] Step S104, according to the global posture information of the drone cluster and the number of landmarks, adaptively allocate the obstacle crossing time and speed to each drone based on the environmental information, and generate the optimal heading and path; specifically includes:

[0046] Assume that the total length of a certain section of the path that the drone crosses is L, and the maximum flight speed of the given drone is v max , the total time to cross the obstacle is T. This strategy sets T = 1.2T, and further divides T evenly to each segment according to the proportion of the length of each segment to the total length, and obtains the initial time allocation. Further, the safety level of each segment is calculated. Adaptively adjust the time allocation for each segment, where for the i-th path τ i , its safety is measured by its closest distance to obstacles:

[0047]

[0048] Where n represents τ i For any grid passed, after obtaining the shortest distance between each path and the obstacle, the time allocation of the i-th path is adjusted to:

[0049]

[0050] Among them, d(τ j ) represents the i-th path τ i degree of safety.

[0051] After the above adjustments, areas closer to obstacles will be allocated more time, so that the drone can obtain a lower speed in dangerous areas, improve the accuracy of trajectory tracking, and prevent the drone from colliding with obstacles due to control errors or environmental errors when the drone is close to obstacles.

[0052] Step S105, initializing the take-off height, maximum flight speed, and expected target posture of each drone cluster, and combining the optimal heading and path to complete autonomous obstacle crossing and exploration of the drone cluster.

[0053] The technical solution of the embodiment of the present invention is described in detail below with reference to the accompanying drawings.

[0054] Based on VINS-Fusion, the local position of a single UAV is obtained, and the grid map is generated by combining the point cloud data. The grid map is further converted into an ESDF map by combining the distance transformation algorithm. Specifically, Figure 2 As shown, the point cloud information is converted from the drone coordinate system to the world coordinate system, and the point cloud is converted into voxels, and the voxel information is added to the insertQueue and deleteQueue queues; the ESDF map is initialized, the two queues are merged into the updateQueue queue, and all possible voxel data that may be changed are updated using a breadth-first algorithm; the updated voxel data is combined with the depth information, and the depth information inside the obstacle is defined as negative, and finally the ESDF map is generated.

[0055] Based on the TF mechanism in ROS, the pose information of each drone is converted into a global pose, and the global pose information of the drone cluster is obtained; the TF mechanism refers to a coordinate transformation mechanism provided by the Transform library, which can be used to transform the relationship between different coordinate systems in different drone systems. The TF library uses a publish and subscribe mechanism to allow coordinate transformation information to be transmitted between multiple nodes.

[0056] The cost function of the A* algorithm is improved, and the number of landmarks is reduced based on the topological equivalence algorithm. The difference from the traditional A* algorithm lies in the design of the node cost. In the present invention, the node cost term is designed as:

[0057] f(n)=g(n)+h(n)+α·c(n)

[0058] Wherein, g(n) and h(n) are the same as the original A* algorithm, representing the actual cost from the initial node to node n and the estimated cost from node n to the target node, respectively. α·c(n) is a new security cost item added in the present invention, which is used to reflect the security level of node n. Here:

[0059]

[0060] esdf(n) represents the distance from node n to the nearest obstacle surface. This value can be directly obtained based on the ESDF map. For more dangerous nodes, the larger c(n) is, the greater the overall cost of selecting the node is. Therefore, in the planning process, the node will be skipped to improve the safety of the path. In addition, grids with esdf(n) less than 0 will be considered as inaccessible areas during the search process, so c(n) is always positive. α is a non-negative weight factor used to adjust the importance of path safety during planning. The smaller α is, the closer the planned path is to the shortest path, but the safety level will also decrease. Increasing α can improve the safety level of the path, and the length of the planned path will also increase. When α is 0, the improved A* algorithm will degenerate into the traditional A* algorithm.

[0061] Reducing the number of landmarks based on the topological equivalence algorithm specifically includes the following: Figure 3 As shown, for any two paths τ1(s), τ2(s), s∈[0,1], if τ1(0)=τ2(0), τ1(1)=τ2(1), if for any s, the line segment In short, when the points on the two paths are mapped one-to-one, if the line between any pair of corresponding points does not collide with an obstacle, then the two paths are topologically equivalent.

[0062] In the actual flight process of the drone, in order to improve safety and flight efficiency, a suitable path needs to be found. The present invention first sets the starting point S as the key landmark point K, and then traverses the subsequent landmark points in sequence. Assuming that the next point of the landmark point K is M, if the line segment If no obstacle is hit, then calculate Distance to the nearest obstacle Calculate the weight of point M The weight is a trade-off between the sparsity of the points and the safety of the path after extracting the key landmarks. The point with the largest weight is further selected as the next key point and the above process is repeated. For each point O in K and G, the weight is calculated. In each iteration, the point set O with the smallest weight w Will serve as a set of key waypoints.

[0063] Based on the environmental information, the time and speed of obstacle crossing are adaptively allocated to generate the optimal heading and path strategy, which specifically includes the following sub-steps: Assuming that the total length of a certain section of the path for the drone to cross obstacles is L, and the maximum flight speed of the drone is v max, the total time to cross the obstacle is T. This strategy sets T = 1.2T, and further divides T evenly to each segment according to the proportion of the length of each segment to the total length, to obtain the initial time allocation, and further calculates the safety level of each segment. Adaptively adjust the time allocation for each segment. For the i-th path τ i , its safety is measured by its closest distance to obstacles:

[0064]

[0065] Where n represents τ i After obtaining the shortest distance between each path and the obstacle, the time allocation of the i-th path is adjusted to:

[0066]

[0067] After the above adjustments, areas closer to obstacles will be allocated more time, so that the drone can obtain a lower speed in dangerous areas, improve the accuracy of trajectory tracking, and prevent the drone from colliding with obstacles due to control errors or environmental errors when the drone is close to obstacles.

[0068] Finally, the take-off height, maximum flight speed, and expected target posture of each drone cluster are initialized, and the drone cluster’s autonomous obstacle crossing and exploration is completed in combination with the optimal heading and path.

[0069] In summary, the technical solution of the embodiment of the present invention utilizes image enhancement and VINS-Fusion algorithm to realize the pose estimation of a single UAV, and obtains the global pose of each UAV based on the TF (Transformations Frames) mechanism of ROS. At the same time, a raster map and an ESDF map are generated based on the depth point cloud obtained by each UAV. Furthermore, a two-stage global path planning strategy is proposed. In the front-end path exploration stage, a safety cost factor is constructed in the cost function of the traditional A* exploration algorithm, and the number of redundant landmarks is reduced based on the topological equivalence algorithm to improve the efficiency of the algorithm; in the back-end trajectory generation stage, an environment-adaptive time and speed allocation strategy is proposed, so that each UAV can autonomously adjust the flight speed and time consumption based on the complexity of the environment, thereby improving the safety of autonomous exploration.

[0070] Device Example 1

[0071] According to an embodiment of the present invention, a drone group autonomous positioning and obstacle crossing device is provided. Figure 4 Schematic diagram of a drone group autonomous positioning and obstacle surmounting device according to an embodiment of the present invention. Figure 4 As shown, the autonomous positioning and obstacle-crossing device for a drone swarm according to an embodiment of the present invention specifically includes:

[0072] The generation conversion module 40 is used to obtain the local pose of a single UAV based on VINS-Fusion, generate a grid map in combination with the point cloud data, and convert the grid map into an ESDF map according to the distance transformation algorithm; specifically used for:

[0073] Convert the point cloud information from the drone coordinate system to the world coordinate system, convert the point cloud into voxels, and add the voxel information to the insertQueue and deleteQueue queues;

[0074] Initialize the ESDF map, merge the two queues into the updateQueue queue, and use the breadth-first algorithm to update all possible changed voxel data;

[0075] Combine the updated voxel data with the depth information, define the depth information inside the obstacle as negative, and finally generate the ESDF map;

[0076] The acquisition module 42 is used to convert the posture information of each drone into a global posture based on the ESDF map by using the TF mechanism in ROS to obtain the global posture information of the drone cluster; specifically used for:

[0077] The TF mechanism in ROS is adopted, and the TF library is used to convert the relationship between different coordinate systems in different drone systems. Among them, the TF library uses the publish and subscribe mechanism to allow coordinate transformation information to be transmitted between multiple nodes.

[0078] The determination module 44 is used to determine the number of landmark points by improving the cost function of the A* algorithm and the topological equivalence algorithm; specifically, it is used to:

[0079] The cost item of the node of the improved A* algorithm is:

[0080] f(n)=g(n)+h(n)+α·c(n) Formula 1;

[0081] Among them, g(n) and h(n) represent the actual cost from the initial node to node n and the estimated cost from node n to the target node, respectively. α·c(n) is a newly added security cost item, which is used to reflect the security level of node n. Here:

[0082]

[0083] Among them, esdf(n) represents the distance from node n to the nearest obstacle surface. This value is directly obtained based on the ESDF map. For more dangerous nodes, the larger c(n) is, the greater the overall cost of selecting the node. In addition, grids with esdf(n) less than 0 are considered inaccessible areas during the search process, so c(n) is always positive. α is a non-negative weight factor used to adjust the emphasis on path safety during planning. The smaller α is, the closer the planned path is to the shortest path, but the safety level will also decrease. Increasing α can improve the safety level of the path, and the length of the planned path will also increase. When α is 0, the improved A* algorithm will degenerate into the traditional A* algorithm.

[0084] For any two paths τ1(s),τ2(s),,s∈[0,1], if τ1(0)=τ2(0),τ1(1)=τ2(1), then for any s, the line segment If both paths are collision-free, then the paths τ1(s) and τ2(s) are topologically equivalent, that is, when a one-to-one mapping is performed on the points on the two paths, if the line between any pair of corresponding points does not collide with an obstacle, then the two paths are topologically equivalent;

[0085] Set the starting point S as the key landmark point K, and then traverse the subsequent landmark points in sequence. Assume that the next point of landmark point K is M. If the line segment If no obstacle is hit, then calculate Distance to the nearest obstacle Calculate the weight of point M This weight is a trade-off between the sparsity of the points and the safety of the path after extracting the key landmark points. The point with the largest weight is further selected as the next key point, and the above process is repeated;

[0086] For each point O in K and G, calculate the weight In each iteration, the point set O with the smallest weight w will be used as the key landmark point set, where d KO Represents a line segment Distance to the nearest obstacle d OG Represents a line segment Distance to the nearest obstacle |KO| represents the Euclidean distance from landmark point K to landmark point O, and |OG| represents the Euclidean distance from landmark point O to landmark point G;

[0087] The optimal heading and path generation module 46 is used to adaptively allocate the obstacle crossing time and speed of each drone based on the global posture information of the drone cluster and the number of landmarks, and generate the optimal heading and path based on the environmental information; specifically used for:

[0088] Assume that the total length of a certain section of the path that the drone crosses is L, and the maximum flight speed of the given drone is v max , the total time to cross the obstacle is T. This strategy sets T = 1.2T, and further divides T evenly to each segment according to the proportion of the length of each segment to the total length, and obtains the initial time allocation. Further, the safety level of each segment is calculated. Adaptively adjust the time allocation for each segment, where for the i-th path τ i , its safety is measured by its closest distance to obstacles:

[0089]

[0090] Where n represents τ i For any grid passed, after obtaining the shortest distance between each path and the obstacle, the time allocation of the i-th path is adjusted to:

[0091]

[0092] Among them, d(τ j ) represents the i-th path τ i degree of safety.

[0093] The autonomous obstacle crossing and exploration module 48 is used to initialize the take-off height, maximum flight speed, and expected target posture of each drone cluster, and complete the autonomous obstacle crossing and exploration of the drone cluster in combination with the optimal heading and path.

[0094] The embodiment of the present invention is a device embodiment corresponding to the above-mentioned method embodiment. The specific operations of each module can be understood by referring to the description of the method embodiment, which will not be repeated here.

[0095] Device Example 2

[0096] An embodiment of the present invention provides an electronic device, such as Figure 5 As shown, it includes: a memory 50, a processor 52, and a computer program stored in the memory 50 and executable on the processor 52. When the computer program is executed by the processor 52, the steps described in the method embodiment are implemented.

[0097] Device Example 3

[0098] An embodiment of the present invention provides a computer-readable storage medium, on which a program for implementing information transmission is stored. When the program is executed by the processor 52, the steps described in the method embodiment are implemented.

[0099] The computer-readable storage medium in this embodiment includes, but is not limited to, ROM, RAM, magnetic disk or optical disk, etc.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for autonomous positioning and obstacle crossing of a drone swarm, characterized in that: include: Based on VINS-Fusion, the local pose of a single UAV is obtained, and a raster map is generated by combining the point cloud data. The raster map is then converted into an ESDF map based on the distance transformation algorithm. Based on the ESDF map, the TF mechanism in ROS is used to convert the pose information of each drone into a global pose, and the global pose information of the drone cluster is obtained; Determine the number of landmarks by improving the cost function of the A* algorithm and the topological equivalence algorithm; According to the global posture information of the drone cluster and the number of landmarks, the obstacle crossing time and speed are adaptively allocated to each drone based on environmental information to generate the optimal heading and path; Initialize the take-off height, maximum flight speed, and expected target posture of each drone cluster, and combine the optimal heading and path to complete the autonomous obstacle crossing and exploration of the drone cluster.

2. The method according to claim 1, characterized in that Based on VINS-Fusion, the local pose of a single UAV is obtained, and a raster map is generated by combining the point cloud data. The raster map is then converted into an ESDF map according to the distance transformation algorithm. Specifically, the following steps are involved: Convert the point cloud information from the drone coordinate system to the world coordinate system, convert the point cloud into voxels, and add the voxel information to the insertQueue and deleteQueue queues; Initialize the ESDF map, merge the two queues into the updateQueue queue, and use the breadth-first algorithm to update all possible changed voxel data; The updated voxel data is combined with the depth information, and the depth information inside the obstacle is defined as negative, and finally the ESDF map is generated.

3. The method according to claim 1, characterized in that The TF mechanism in ROS is used to convert the posture information of each drone into the global posture, including: The TF mechanism in ROS is adopted, and the TF library is used to convert the relationship between different coordinate systems in different drone systems. Among them, the TF library uses the publish and subscribe mechanism to allow coordinate transformation information to be transmitted between multiple nodes.

4. The method according to claim 1, characterized in that: The number of landmark points is determined by improving the cost function of the A* algorithm and the topological equivalence algorithm, including: The cost item of the node of the improved A* algorithm is: f(n)=g(n)+h(n)+α·c(n) Formula 1; Among them, g(n) and h(n) represent the actual cost from the initial node to node n and the estimated cost from node n to the target node, respectively. α·c(n) is a newly added security cost item, which is used to reflect the security level of node n. Here: Among them, esdf(n) represents the distance from node n to the nearest obstacle surface. This value is directly obtained based on the ESDF map. For more dangerous nodes, the larger c(n) is, the greater the overall cost of selecting the node. In addition, grids with esdf(n) less than 0 are considered inaccessible areas during the search process, so c(n) is always positive. α is a non-negative weight factor used to adjust the emphasis on path safety during planning. The smaller α is, the closer the planned path is to the shortest path, but the safety level will also decrease. Increasing α can improve the safety level of the path, and the length of the planned path will also increase. When α is 0, the improved A* algorithm will degenerate into the traditional A* algorithm. For any two paths τ1(s), τ2(s), s∈[0,1], if τ1(0)=τ2(0), τ1(1)=τ2(1), if for any s, the line segment If both paths are collision-free, then the paths τ1(s) and τ2(s) are topologically equivalent, that is, when a one-to-one mapping is performed on the points on the two paths, if the line between any pair of corresponding points does not collide with an obstacle, then the two paths are topologically equivalent; Set the starting point S as the key landmark point K, and then traverse the subsequent landmark points in sequence. Assume that the next point of landmark point K is M. If the line segment If no obstacle is hit, then calculate Distance to the nearest obstacle Calculate the weight of point M This weight is a trade-off between the sparsity of the points and the safety of the path after extracting the key landmark points. The point with the largest weight is further selected as the next key point, and the above process is repeated; For each point O in K and G, calculate the weight In each iteration, the point set O with the smallest weight w will be used as the key landmark point set, where d KO Represents a line segment Distance to the nearest obstacle d OG Represents a line segment Distance to the nearest obstacle |KO| represents the Euclidean distance from landmark point K to landmark point O, and |OG| represents the Euclidean distance from landmark point O to landmark point G.

5. The method according to claim 1, characterized in that According to the global posture information of the drone cluster and the number of landmarks, the obstacle crossing time and speed of each drone are adaptively allocated based on the environmental information to generate the optimal heading and path. Specifically, it includes: Assume that the total length of a certain section of the path that the drone crosses is L, and the maximum flight speed of the given drone is v max , the total time to cross the obstacle is T. This strategy sets T = 1.2T, and further divides T evenly to each segment according to the proportion of the length of each segment to the total length, and obtains the initial time allocation. Further, the safety level of each segment is calculated. Adaptively adjust the time allocation for each segment, where for the i-th path τ i , its safety is measured by its closest distance to obstacles: Where n represents τ i For any grid passed, after obtaining the shortest distance between each path and the obstacle, the time allocation of the i-th path is adjusted to: Among them, d(τ j ) represents the i-th path τ i degree of safety.

6. An autonomous positioning and obstacle-crossing device for drone swarms, characterized in that: include: Generate a conversion module, which is used to obtain the local pose of a single UAV based on VINS-Fusion, generate a raster map based on point cloud data, and convert the raster map into an ESDF map based on the distance transformation algorithm; An acquisition module is used to convert the pose information of each drone into a global pose based on the ESDF map by using the TF mechanism in ROS to acquire the global pose information of the drone cluster; A determination module, used for determining the number of landmark points by improving the cost function of the A* algorithm and the topological equivalence algorithm; The optimal heading and path generation module is used to adaptively allocate the obstacle crossing time and speed of each drone based on the global posture information of the drone cluster and the number of landmarks, and generate the optimal heading and path based on the environmental information; The autonomous obstacle crossing and exploration module is used to initialize the take-off height, maximum flight speed, and expected target posture of each drone cluster, and complete the autonomous obstacle crossing and exploration of the drone cluster in combination with the optimal heading and path.

7. The device according to claim 6, characterized in that The generation conversion module is specifically used for: Convert the point cloud information from the drone coordinate system to the world coordinate system, convert the point cloud into voxels, and add the voxel information to the insertQueue and deleteQueue queues; Initialize the ESDF map, merge the two queues into the updateQueue queue, and use the breadth-first algorithm to update all possible changed voxel data; Combine the updated voxel data with the depth information, define the depth information inside the obstacle as negative, and finally generate the ESDF map; The acquisition module is specifically used for: The TF mechanism in ROS is adopted, and the TF library is used to convert the relationship between different coordinate systems in different drone systems. Among them, the TF library uses the publish and subscribe mechanism to allow coordinate transformation information to be transmitted between multiple nodes.

8. The device according to claim 6, characterized in that The determination module is specifically used for: The cost item of the node of the improved A* algorithm is: f(n)=g(n)+h(n)+α·c(n) Formula 1; Among them, g(n) and h(n) represent the actual cost from the initial node to node n and the estimated cost from node n to the target node, respectively. α·c(n) is a newly added security cost item, which is used to reflect the security level of node n. Here: Among them, esdf(n) represents the distance from node n to the nearest obstacle surface. This value is directly obtained based on the ESDF map. For more dangerous nodes, the larger c(n) is, the greater the overall cost of selecting the node. In addition, grids with esdf(n) less than 0 are considered inaccessible areas during the search process, so c(n) is always positive. α is a non-negative weight factor used to adjust the emphasis on path safety during planning. The smaller α is, the closer the planned path is to the shortest path, but the safety level will also decrease. Increasing α can improve the safety level of the path, and the length of the planned path will also increase. When α is 0, the improved A* algorithm will degenerate into the traditional A* algorithm. For any two paths τ1(s),τ2(s),,s∈[0,1], if τ1(0)=τ2(0),τ1(1)=τ2(1), then for any s, the line segment If both paths are collision-free, then the paths τ1(s) and τ2(s) are topologically equivalent, that is, when a one-to-one mapping is performed on the points on the two paths, if the line between any pair of corresponding points does not collide with an obstacle, then the two paths are topologically equivalent; Set the starting point S as the key landmark point K, and then traverse the subsequent landmark points in sequence. Assume that the next point of landmark point K is M. If the line segment If no obstacle is hit, then calculate Distance to the nearest obstacle Calculate the weight of point M This weight is a trade-off between the sparsity of the points and the safety of the path after extracting the key landmark points. The point with the largest weight is further selected as the next key point, and the above process is repeated; For each point O in K and G, calculate the weight In each iteration, the point set O with the smallest weight w will be used as the key landmark point set, where d KO Represents a line segment Distance to the nearest obstacle d OG Represents a line segment Distance to the nearest obstacle |KO| represents the Euclidean distance from landmark point K to landmark point O, and |OG| represents the Euclidean distance from landmark point O to landmark point G; The optimal heading and path generation module is specifically used for: Assume that the total length of a certain section of the path that the drone crosses is L, and the maximum flight speed of the given drone is v max , the total time to cross the obstacle is T. This strategy sets T = 1.2T, and further divides T evenly to each segment according to the proportion of the length of each segment to the total length, and obtains the initial time allocation. Further, the safety level of each segment is calculated. Adaptively adjust the time allocation for each segment, where for the i-th path τ i , its safety is measured by its closest distance to obstacles: Where n represents τ i For any grid passed, after obtaining the shortest distance between each path and the obstacle, the time allocation of the i-th path is adjusted to: Among them, d(τ j ) represents the i-th path τ i degree of safety.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the autonomous positioning and obstacle surmounting method for a swarm of unmanned aerial vehicles as described in any one of claims 1 to 5 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an implementation program for information transmission, and when the program is executed by the processor, the steps of the autonomous positioning and obstacle surmounting method for a swarm of unmanned aerial vehicles as described in any one of claims 1 to 5 are implemented.

Images