Unmanned aerial vehicle obstacle avoidance system based on multi-device interconnection

Through the multi-device interconnected drone obstacle avoidance system, the majority voting principle is used to compare and verify the environmental map and obstacle avoidance paths, optimize the environment perception and path planning, and solve the problems of insufficient coordination and data consistency in traditional systems, improving the problem of improvement of the problem of insufficient coordination and data consistency in traditional systems. The obstacle avoidance efficiency and safety of drones.

CN118605569BActive Publication Date: 2024-11-19NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410821576.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-24
Publication Date
2024-11-19
Estimated Expiration
2044-06-24

AI Technical Summary

Technical Problem

Traditional UAV obstacle avoidance systems have insufficient coordination and data consistency problems in multiple UAV arrays, resulting in low environmental perception accuracy and inconsistent path planning, and prone to misjudgment of obstacles and path conflicts, and overall obstacle avoidance efficiency and safety. Insufficient sex.

Method used

Adopt a UAV obstacle avoidance system based on multi-device interconnection, including a UAV array module, environment perception module, path planning module, communication module and verification module, and compare and verify the environmental map and obstacle avoidance path through the majority voting principle. Optimize environment awareness and path planning algorithms to ensure environmental data sharing and path consistency.

Benefits of technology

It improves the accuracy of drone environment perception and the reliability of path planning, ensures the efficient and safe obstacle avoidance capabilities of multiple drones in complex environments, and enhances the stability and reliability of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118605569B_ABST
    Figure CN118605569B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of unmanned aerial vehicle obstacle avoidance, and in particular to an unmanned aerial vehicle obstacle avoidance system based on multi-device interconnection, wherein the unmanned aerial vehicle array module, the environment perception module, the path planning module, the communication module and the verification module are used for forming a plurality of unmanned aerial vehicles into an array, wherein there are N unmanned aerial vehicles in the array; the unmanned aerial vehicle in the environment perception module collects environment sensing data through an obstacle avoidance sensor, and generates its own environment map according to the environment sensing data; the path planning module calculates a self-obstacle avoidance path from a starting point to an end point based on its own environment map by using a path planning algorithm; the communication module transmits the self-environment map and the self-obstacle avoidance path to other unmanned aerial vehicles in the array as an environment map to be verified and an obstacle avoidance path to be verified; the verification module receives the environment map to be verified and the obstacle avoidance path to be verified sent by all other unmanned aerial vehicles, performs verification, and selects a path and optimizes it according to the verification result.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle obstacle avoidance, and in particular to a unmanned aerial vehicle obstacle avoidance system based on multi-device interconnection. Background Art

[0002] Traditional drone obstacle avoidance systems usually face problems of insufficient coordination and data consistency in multi-drone arrays. Traditional systems usually have each drone operate independently, relying on its own sensors and algorithms to perceive the environment and calculate obstacle avoidance paths. This approach leads to a lack of real-time environmental data sharing and path information exchange between drones, which makes data island problems prone to occur.

[0003] In terms of environmental perception, since there is no data support from other drones, the system's environmental perception accuracy only relies on the capabilities of a single sensor, which is easily affected by sensor noise and environmental changes, resulting in low accuracy of the generated environmental map.

[0004] In addition, the path planning process is only based on its own perception data, and it is unable to effectively coordinate and verify the paths of other drones. This method of independently calculating paths often leads to inconsistent obstacle avoidance paths between drones, which may cause path conflicts or low overall obstacle avoidance efficiency. In terms of verification, due to the lack of a systematic verification mechanism, the traditional system cannot cross-verify the paths and environmental perception results generated by each drone, lacks error correction and optimization means, and ultimately reduces the obstacle avoidance capability of the drone and the reliability of mission execution.

[0005] Therefore, when operating in complex environments, drones of traditional systems are prone to misjudging obstacles, unreasonable path planning, or inability to effectively avoid obstacles. The stability and safety of the overall system when multiple drones work together are obviously insufficient. Summary of the invention

[0006] The purpose of the present invention is to provide a drone obstacle avoidance system based on multi-device interconnection to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: a drone obstacle avoidance system based on multi-device interconnection, characterized in that it includes a drone array module, an environment perception module, a path planning module, a communication module and a verification module, wherein:

[0008] The drone array module is used to form a plurality of drones into an array, wherein there are N drones in the array;

[0009] The UAV in the environmental perception module collects environmental sensing data through obstacle avoidance sensors and generates its own environmental map based on the environmental sensing data;

[0010] The path planning module calculates a self-obstacle avoidance path from the starting point to the end point based on the self-environment map using a path planning algorithm;

[0011] The communication module transmits the self-environment map and the self-obstacle avoidance path to other drones in the array as the environment map to be verified and the obstacle avoidance path to be verified;

[0012] The verification module receives the environment maps to be verified and the obstacle avoidance paths to be verified sent by all other drones, and performs verification, specifically including:

[0013] The obstacle comparison and verification of the self-environment map and the environment maps to be verified sent by all other UAVs are performed in turn, and the majority comparison results are used as the verification results by using the majority voting principle, where:

[0014] If the environmental map comparison verification fails, the environmental perception module is optimized according to the records generated during the verification process;

[0015] If the environmental map comparison and verification is passed, the path planning algorithm in the path planning module is used to predict the obstacle avoidance paths of the environmental maps to be verified sent by all other drones in turn, and the obstacle avoidance paths in the predicted results are compared and verified with the obstacle avoidance paths to be verified sent by all other drones in turn. The majority voting principle is also used to take the majority of the comparison results as the verification results, where:

[0016] If the obstacle avoidance path verification passes, the obstacle avoidance path will be used as the final obstacle avoidance path for the drone;

[0017] If the obstacle avoidance path verification fails, the path planning module is optimized according to the records generated during the verification process.

[0018] As a further improvement of the technical solution, the environment perception module collects environment sensing data through an obstacle avoidance sensor, specifically including:

[0019] LiDAR: Use LiDAR to scan the surrounding environment and obtain a series of point cloud data, each point contains the location coordinates;

[0020] Visual sensor: Use a camera to capture images and obtain feature points in the image;

[0021] Ultrasonic sensor: Use ultrasonic sensor to measure the distance to obstacles and obtain one-dimensional distance data;

[0022] GPS and inertial measurement sensors: collect and obtain the drone's location information and attitude data to assist in environmental perception;

[0023] Environmental sensing data includes point cloud data, image feature points, one-dimensional distance, drone location information and attitude data.

[0024] As a further improvement of the technical solution, the environment perception module preprocesses and analyzes the collected environment sensor data to generate its own environment map, specifically including:

[0025] Noise filtering preprocessing: noise filtering is performed on the collected environmental sensor data, including mean filtering. By taking the values ​​of each data point and its neighboring points, the average value of these values ​​is calculated and the value of the original data point is replaced by this average value.

[0026] Environmental modeling: Divide the point cloud space into fixed-size voxels, calculate the centroid of the points in each voxel, and use the centroid to represent the voxel; initialize the environmental map into several fixed-size grids, and update the occupancy probability of each grid according to the environmental sensor data to build a complete environmental map, indicating whether each area is occupied by an obstacle;

[0027] Obstacle recognition: Based on the fixed-size grids in the complete environment map, multiple bounding boxes and corresponding confidence scores are predicted for each grid. A forward pass is performed using a convolutional neural network to obtain the bounding boxes and categories of all targets, convert the identified obstacles into spatial coordinates, calculate their position and size, generate their own obstacle data, and use the occupancy probability and obstacle data of each grid as their own environment map.

[0028] As a further improvement of the technical solution, the path planning module calculates a self-obstacle avoidance path from the starting point to the end point according to the self-environment map using the path planning algorithm. The steps of calculating the self-obstacle avoidance path are as follows:

[0029] Use a cost function to evaluate each node, including the actual cost and the estimated cost, where the actual cost is the actual cost of moving from the starting point to the current node; the estimated cost is the estimated cost from the current node to the target node; the cost function is the sum of these two parts;

[0030] Manhattan distance is chosen as the heuristic function, which is the sum of horizontal and vertical distances from the current node to the target node;

[0031] Add the starting node to the open list and set the cost of the starting node; initialize the closed list to be empty; the open list is used to store the nodes to be checked, and the closed list is used to store the nodes that have been checked;

[0032] Take the node with the lowest cost from the open list as the current node; if the current node is the end point, terminate the search and the path is found; move the current node from the open list to the closed list; perform the following operations on each neighbor node of the current node:

[0033] If the neighbor node is in the closed list, it is skipped;

[0034] If the neighbor node is an obstacle in its own environment map, that is, the occupancy probability is higher than the standard threshold, it will be skipped;

[0035] Calculate the temporary cost of the neighbor node, that is, the cost from the starting point through the current node to the neighbor node;

[0036] If the neighbor node is not in the open list, or a path with a lower cost is found to reach the neighbor node, update the parent node of the neighbor node to the current node; update the actual cost of the neighbor node; calculate and update the total cost of the neighbor node; if the neighbor node is not in the open list, add it to the open list;

[0037] When the end node is found, trace back from the end node along the parent node mapping to the starting node, collect the coordinates of all nodes, and generate a self-obstacle avoidance path from the starting point to the end point.

[0038] As a further improvement of the technical solution, the verification module includes an environment map verification unit, which receives the environment maps to be verified and the obstacle avoidance paths to be verified sent by all other drones in the array, and performs verification, specifically including:

[0039] Preliminary comparison: gradually compare the obstacles of each received environment map to be verified with the environment map itself, and compare the occupancy status of each grid in each two maps one by one, as follows:

[0040] The environment map is represented as a two-dimensional grid, and each element in the grid represents the occupancy status of the location;

[0041] Grid-by-grid comparison: create a consistent counter to record the number of consistent grids, and create a comparison counter to record the number of all compared grids;

[0042] Comparison logic: For each grid position in the matrix, compare the values ​​of the environment map at that position with the value of the environment map to be verified. If the two values ​​are equal, increase the consistent grid counter; regardless of whether the two values ​​are equal, increase the comparison counter;

[0043] Result statistics: Calculate the consistent grid ratio, that is, the number of consistent grids divided by the number of all compared grids, and judge the consistency, set the consistency threshold, if the consistent grid ratio is higher than the consistency threshold, then the two maps are considered Figure 1 If they are consistent, they are considered inconsistent;

[0044] Environmental map comparison and verification: record the number of results that are consistent with the environmental map comparison results of all other drones, and use the majority voting principle to determine the final environmental map comparison and verification results.

[0045] As a further improvement of the technical solution, the verification module includes an obstacle avoidance path verification unit, which uses a path planning algorithm to predict obstacle avoidance paths for the environment maps to be verified sent by all other drones in turn, and compares and verifies the obstacle avoidance paths in the prediction results with the obstacle avoidance paths to be verified in turn. The specific steps are as follows:

[0046] Determine the key point sequence of the predicted obstacle avoidance path and the obstacle avoidance path to be verified, where the key point sequence includes turning points and waypoints in the path, representing the points and components where the path changes in direction. The turning point is the position where the direction changes in the path, and the waypoint is the middle point connecting two turning points in the path, which is used to describe the path in more detail.

[0047] Create a structure to record the comparison results, which is used to record the comparison results of each key point;

[0048] Node-by-node comparison: For each key point in the predicted obstacle avoidance path, the distance between the two key points is calculated. If the distance is within the standard error range, the two key points are considered similar, otherwise they are considered dissimilar.

[0049] Record the obstacle avoidance path comparison results: record each comparison result, and calculate the similarity of the overall path based on the comparison results. The similarity of the overall path is the percentage of similar key points to the total number of key points. Set a similarity threshold. If the ratio of similar key points exceeds the similarity threshold, the paths are considered similar, and the predicted obstacle avoidance path is consistent with the obstacle avoidance path to be verified; otherwise, they are considered dissimilar, and the predicted obstacle avoidance path is inconsistent with the obstacle avoidance path to be verified.

[0050] Finally, the majority voting principle is also used to take the majority comparison results as the verification results of the obstacle avoidance path.

[0051] As a further improvement of the present technical solution, when the environmental map comparison verification calculated by the environmental map verification unit fails, the record result of the environmental map comparison verification is sent to the drone corresponding to the inconsistent obstacle comparison result, and the corresponding environmental perception module is optimized according to the record result, and the record includes the inconsistent grid position in the environmental map and the deviation information of the obstacle and its specific position; the drone receiving the comparison result will optimize its environmental perception module according to the record, and the specific steps of the optimization include:

[0052] Recalibrate sensor: adjust the calibration of the sensor;

[0053] Adjust algorithm parameters: adjust the parameters of the algorithm in the environment perception module according to the deviation information in the comparison record;

[0054] Add more filtering steps: Add more filtering steps during data processing;

[0055] Rescan the environment: Rescan the surrounding environment to update and correct the environment map.

[0056] As a further improvement of the present technical solution, when the verification result of the obstacle avoidance path calculated by the obstacle avoidance path verification unit fails, the record result of the obstacle avoidance path comparison is sent to the drone corresponding to the inconsistent obstacle avoidance path comparison result, and the corresponding path planning module is optimized according to the record result, and the record result includes the key point deviation and the overall similarity of the predicted obstacle avoidance path and the obstacle avoidance path to be verified; the drone receiving the comparison result will optimize its path planning module according to these records, and the specific steps include:

[0057] Optimize the path planning algorithm: adjust the path planning algorithm based on the key point deviation information in the comparison record;

[0058] Adjust cost function parameters: Optimize the evaluation criteria of path planning by modifying the parameters in the cost function;

[0059] Re-plan the path: Based on the updated environmental perception data and the optimized algorithm, a new obstacle avoidance path is re-planned and the obstacle avoidance path verification is re-performed until all drones confirm that the obstacle avoidance path verification has passed.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] 1. The self-environmental map generated by the environmental perception module in the obstacle avoidance system of the drone based on multi-device interconnection can not only derive its own obstacle avoidance path, but also can be compared and verified with the environmental maps to be verified sent by all other drones in sequence; this method ensures that each drone can share and compare environmental data. By comparing grid by grid, the system can accurately judge the consistency of the environmental map and ensure that all drones have consistent perception of the environment; when the environmental map comparison fails, the system will send the comparison result to the relevant drones, providing inconsistent grid positions and obstacle deviation information; based on this information, the drone recalibrates the sensor, adjusts the algorithm parameters and increases the filtering steps, thereby improving the accuracy of environmental perception and ensuring that the final environmental map data is accurate and reliable.

[0062] 2. The path planning algorithm used by the path planning module in the drone obstacle avoidance system based on multi-device interconnection can not only calculate its own obstacle avoidance path, but also predict the obstacle avoidance path of the environment map to be verified sent by all other drones in turn; by comparing the key point sequence of the predicted obstacle avoidance path with the obstacle avoidance path to be verified, the system can accurately calculate the similarity of the path to ensure that the predicted path is highly consistent with the actual path; when the path verification fails, the comparison result will be fed back to the relevant drones to optimize their path planning algorithms and adjust the cost function parameters; through this method, each drone can continuously improve itself to make the generated obstacle avoidance path closer to actual needs, thereby improving the efficiency and safety of the overall obstacle avoidance task.

[0063] 3. The obstacle avoidance system based on multi-device interconnection ensures that multiple drones can perform obstacle avoidance tasks efficiently and safely by coordinating the linkage of multiple modules. The drone array module is responsible for the organization and initialization of the array, the environmental perception module collects and processes environmental data, the path planning module calculates the obstacle avoidance path, the communication module realizes data sharing and transmission, and the verification module is responsible for the verification of the path and environmental map. Through noise filtering, environmental modeling and obstacle identification, the system can generate an accurate environmental map and calculate the optimal path through the path planning algorithm. Each verification step uses the majority voting principle to ensure the reliability of the verification results. The optimized sensors and algorithms improve the overall performance of the system, thereby ensuring that the drone array can operate stably and effectively avoid obstacles in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a schematic diagram of the overall module of the present invention;

[0065] Figure 2 It is a schematic diagram of the verification module unit of the present invention.

[0066] In the figure: 100, drone array module; 200, environment perception module; 300, path planning module; 400, communication module; 500, verification module; 501, environment map verification unit; 502, obstacle avoidance path verification unit. DETAILED DESCRIPTION

[0067] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] See also Figure 1-2The present invention provides a technical solution: a drone obstacle avoidance system based on multi-device interconnection, including a drone array module 100, an environment perception module 200, a path planning module 300, a communication module 400 and a verification module 500.

[0069] The drone array module 100 is used to form a coordinated array of multiple drones to ensure that the drones can cooperate with each other and efficiently perform obstacle avoidance tasks. Each drone is included in the array system, registered and initialized, and a unique identification ID is set for each drone for subsequent communication and data exchange.

[0070] The environment perception module 200 is a key component of the system. Its main function is to perceive and collect obstacle data in the surrounding environment in real time to ensure that the drone can safely plan its path and avoid obstacles. First, the environment sensing data is collected through the obstacle avoidance sensor, which includes:

[0071] LiDAR: Use LiDAR to scan the surrounding environment and obtain a series of point cloud data, each point contains the location coordinates;

[0072] Visual sensor: Use a camera to capture images or videos and obtain feature points in the image;

[0073] Ultrasonic sensor: Use ultrasonic sensor to measure the distance to obstacles and obtain one-dimensional distance data;

[0074] GPS and inertial measurement sensors: collect and obtain the drone's location information and attitude data to assist in environmental perception.

[0075] Environmental sensing data includes point cloud data, image feature points, one-dimensional distance, drone location information and attitude data.

[0076] The environment perception module 200 pre-processes and analyzes the collected environment sensor data to generate its own environment map, which specifically includes:

[0077] Noise filtering preprocessing: noise filtering is performed on the collected environmental sensor data, including mean filtering. By taking the values ​​of each data point and its neighboring points, the average of these values ​​is calculated, and the value of the original data point is replaced with this average value to smooth the data and reduce the impact of random noise;

[0078] Environmental modeling: Divide the point cloud space into cubes (voxels) of fixed size, calculate the center of gravity of the points in each voxel, and use the center of gravity to represent the voxel to reduce the amount of point cloud data and improve computing efficiency; initialize the environmental map into several fixed-size grids, and update the occupancy probability of each grid according to the environmental sensor data to build a complete environmental map, indicating whether each area is occupied by an obstacle;

[0079] Obstacle recognition: Based on the fixed-size grids in the complete environment map, multiple bounding boxes and corresponding confidence scores are predicted for each grid. A forward pass is performed using a convolutional neural network to obtain the bounding boxes and categories of all targets, convert the identified obstacles into spatial coordinates, calculate their position and size, generate their own obstacle data, and use the occupancy probability and obstacle data of each grid as their own environment map.

[0080] The path planning module 300 calculates a self-obstacle avoidance path from the starting point to the end point based on the self-environment map using the path planning algorithm. The self-environment map contains the occupancy probability and obstacle data of each grid. The steps of calculating the self-obstacle avoidance path are as follows:

[0081] Use a cost function to evaluate each node, including the actual cost and the estimated cost, where the actual cost is the actual cost of moving from the starting point to the current node; the estimated cost is the estimated cost from the current node to the target node; the cost function is the sum of these two parts;

[0082] Manhattan distance is chosen as the heuristic function, which is the sum of horizontal and vertical distances from the current node to the target node;

[0083] Add the starting node to the open list and set the cost of the starting node; initialize the closed list to be empty; the open list is used to store the nodes to be checked, and the closed list is used to store the nodes that have been checked;

[0084] Take the node with the lowest cost from the open list as the current node; if the current node is the end point, terminate the search and the path is found; move the current node from the open list to the closed list; perform the following operations on each neighbor node of the current node (in the four directions of up, down, left, and right):

[0085] If the neighbor node is in the closed list, it is skipped;

[0086] If the neighbor node is an obstacle in its own environment map, that is, the occupancy probability is higher than the standard threshold, it will be skipped;

[0087] Calculate the temporary cost of the neighbor node, that is, the cost from the starting point through the current node to the neighbor node;

[0088] If the neighbor node is not in the open list, or a path with a lower cost is found to reach the neighbor node, update the parent node of the neighbor node to the current node; update the actual cost of the neighbor node; calculate and update the total cost of the neighbor node; if the neighbor node is not in the open list, add it to the open list;

[0089] When the end node is found, trace back to the starting node from the end node along the parent node mapping, collect the coordinates of all nodes, and generate a self-obstacle avoidance path from the starting point to the end point, which consists of a series of grids, such as: each node passed from the starting point to the end point.

[0090] The communication module 400 transmits its own environment map and its own obstacle avoidance path to other drones in the array through a wireless network as an environment map to be verified and an obstacle avoidance path to be verified.

[0091] The environment map verification unit 501 in the verification module 500 receives the environment maps to be verified and the obstacle avoidance paths to be verified sent by all other drones in the array, and performs verification, specifically including:

[0092] Preliminary comparison: gradually compare the obstacles of each received environment map to be verified with the environment map itself, and compare the occupancy status of each grid in each two maps one by one, as follows:

[0093] The environment map is represented as a two-dimensional grid, and the elements in each grid represent the occupancy status of the location, for example: 0 means that the grid is empty, and 1 means that the grid is occupied by an obstacle;

[0094] Grid-by-grid comparison: create a consistent counter to record the number of consistent grids, and create a comparison counter to record the number of all compared grids;

[0095] Comparison logic: For each grid position in the matrix, compare the values ​​of the environment map at that position with the value of the environment map to be verified. If the two values ​​are equal, increase the consistent grid counter; regardless of whether the two values ​​are equal, increase the comparison counter;

[0096] Result statistics: Calculate the consistent grid ratio, that is, the number of consistent grids divided by the number of all compared grids, and judge the consistency, set the consistency threshold, if the consistent grid ratio is higher than the consistency threshold, then the two maps are considered Figure 1 If they are consistent, they are considered inconsistent;

[0097] Environmental map comparison verification: record the number of results that are consistent with the environmental map comparison of all other drones, and use the majority voting principle to determine the final environmental map comparison verification result. For example, the environmental map comparison result with x drones is consistent, while the environmental map comparison result with y drones is inconsistent. If x is greater than y, and x+y=N-1, where N is the total number of all drones, and N minus 1 is because it does not include itself, then the final verification result is consistent, that is, the verification is passed; if x is less than or equal to y, the verification fails, where:

[0098] When the environmental map comparison verification fails, the record result of the environmental map comparison verification is sent to the drone corresponding to the inconsistent obstacle comparison result, and the corresponding environmental perception module 200 is optimized according to the record result, and the record includes the inconsistent grid position in the environmental map and the deviation information of the obstacle and its specific position; the drone receiving the comparison result will optimize its environmental perception module 200 according to these records, and the specific steps of the optimization include:

[0099] Recalibrate the sensor: Adjust the calibration of the sensor to ensure it provides more accurate data;

[0100] Adjust algorithm parameters: According to the deviation information in the comparison records, adjust the parameters of the environment perception algorithm to make it more suitable for the actual environment;

[0101] Add filtering steps: Add more filtering steps during data processing to remove noise and improve data accuracy;

[0102] Rescan the environment: Rescan the surrounding environment, update and correct the environment map to ensure the accuracy of the map information.

[0103] When the environmental map comparison verification in the environmental map verification unit 501 passes, the obstacle avoidance path verification unit 502 in the verification module 500 is started, and the obstacle avoidance path verification unit 502 uses the path planning algorithm in the path planning module 300 to predict the obstacle avoidance paths of all other unmanned aerial vehicles to be verified. The obstacle avoidance paths in the prediction results are compared and verified with the obstacle avoidance paths to be verified. The specific steps are as follows:

[0104] Determine the key point sequence of the predicted obstacle avoidance path and the obstacle avoidance path to be verified, where the key point sequence includes turning points and waypoints in the path, representing the points and components where the path changes in direction. The turning point is the position where the direction changes in the path, and the waypoint is the middle point connecting two turning points in the path, which is used to describe the path in more detail.

[0105] Create a structure to record the comparison results, which is used to record the comparison results of each key point;

[0106] Node-by-node comparison: For each key point in the predicted obstacle avoidance path, the distance between the two key points is calculated. If the distance is within the standard error range, the two key points are considered similar, otherwise they are considered dissimilar.

[0107] Record the obstacle avoidance path comparison results: record each comparison result, and calculate the similarity of the overall path based on the comparison results. The similarity of the overall path is the percentage of similar key points to the total number of key points. Set a similarity threshold. If the ratio of similar key points exceeds the similarity threshold, the paths are considered similar, and the predicted obstacle avoidance path is consistent with the obstacle avoidance path to be verified; otherwise, they are considered dissimilar, and the predicted obstacle avoidance path is inconsistent with the obstacle avoidance path to be verified.

[0108] Finally, the majority voting principle is also used to take the majority comparison results as the verification results of the obstacle avoidance path, where:

[0109] When the obstacle avoidance path verification result passes, the obstacle avoidance path is used as the final obstacle avoidance path for the drone;

[0110] When the verification result of the obstacle avoidance path fails, the record result of the obstacle avoidance path comparison is sent to the drone corresponding to the inconsistent obstacle avoidance path comparison result, and the corresponding path planning module 300 is optimized according to the record result, and the record result includes the key point deviation and the overall similarity of the predicted obstacle avoidance path and the obstacle avoidance path to be verified; the drone receiving the comparison result will optimize its path planning module 300 according to these records, and the specific steps include:

[0111] Optimize the path planning algorithm: adjust the path planning algorithm based on the key point deviation information in the comparison record to make the generated path more in line with actual needs;

[0112] Adjust the cost function parameters: By modifying the parameters in the cost function, the evaluation criteria of path planning are optimized to make it closer to the current environmental requirements and task objectives;

[0113] Re-plan the path: Based on the updated environmental perception data and the optimized algorithm, a new obstacle avoidance path is re-planned to ensure that the new path is more accurate, safe, and in line with the actual environmental conditions, until all drones confirm that the obstacle avoidance path verification has passed.

[0114] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. The obstacle avoidance system for unmanned aerial vehicles based on multi-device interconnection is characterized by: The system comprises a drone array module (100), an environment perception module (200), a path planning module (300), a communication module (400) and a verification module (500), wherein: The drone array module (100) is used to form N drones into an array; The drone in the environment perception module (200) collects environment sensing data through an obstacle avoidance sensor, and generates its own environment map based on the environment sensing data, wherein the environment map uses a grid of a fixed size to represent space, and an element of each grid represents whether the position is occupied by an obstacle; The path planning module (300) calculates a self-obstacle avoidance path from a starting point to an end point based on the self-environment map using a path planning algorithm, wherein the obstacle avoidance path is composed of a series of key points, and the key points represent points where the path changes in direction and their components; The communication module (400) transmits the self-environment map and the self-obstacle avoidance path to other drones in the array; and the self-environment map and the self-obstacle avoidance path sent by other drones as the environment map to be verified and the obstacle avoidance path to be verified; The verification module (500) is used to perform environmental map comparison verification and obstacle avoidance path verification on the environmental map to be verified and the obstacle avoidance path to be verified, specifically comprising: The obstacle comparison and verification is performed on the own environment map and the environment maps to be verified sent by all other drones in sequence, wherein the comparison process specifically compares each grid in the own environment map with the corresponding grid in the environment maps to be verified sent by other drones, and the comparison content includes the occupation status of obstacles in the grid; using the majority voting principle, the majority comparison results are used as the verification results, wherein: If the environmental map comparison verification fails, the environmental perception module (200) is optimized according to the inconsistent grid positions generated during the verification process, at which time the obstacle avoidance path is not verified, and after the environmental perception module (200) is optimized, the environmental map comparison verification is performed again; If the environmental map comparison and verification is passed, the obstacle avoidance path is verified, and the path planning algorithm in the path planning module (300) is used to predict the obstacle avoidance paths of the environmental maps to be verified sent by all other drones in turn, and the obstacle avoidance paths in the prediction results are compared and verified with the obstacle avoidance paths to be verified sent by all other drones in turn, wherein the comparison process specifically compares the key points of the predicted obstacle avoidance path with the corresponding key points in the obstacle avoidance path to be verified, and the comparison content includes the position of the key points and the shape of the path, and the majority of the comparison results are used as the verification results by using the majority voting principle, wherein: If the obstacle avoidance path verification passes, the obstacle avoidance path will be used as the final obstacle avoidance path for the drone; If the obstacle avoidance path verification fails, the path planning module (300) is optimized according to the key point deviations generated during the verification process.

2. The obstacle avoidance system for unmanned aerial vehicles based on multi-device interconnection according to claim 1 is characterized in that: The environment perception module (200) collects environment sensing data through an obstacle avoidance sensor, specifically comprising: LiDAR: Use LiDAR to scan the surrounding environment and obtain a series of point cloud data, each point contains the location coordinates; Visual sensor: Use a camera to capture images and obtain feature points in the image; Ultrasonic sensor: Use ultrasonic sensor to measure the distance to obstacles and obtain one-dimensional distance data; GPS and inertial measurement sensors: collect and obtain the drone's location information and attitude data to assist in environmental perception; Among them, point cloud data, image feature points, one-dimensional distance, drone location information and posture data together constitute environmental sensing data.

3. The unmanned aerial vehicle obstacle avoidance system based on multi-device interconnection according to claim 2 is characterized in that: The environment perception module (200) pre-processes and analyzes the collected environment sensing data to generate its own environment map, specifically including: Noise filtering preprocessing: noise filtering is performed on the collected environmental sensor data, including mean filtering. By taking the values ​​of each data point and its neighboring points, the average value of these values ​​is calculated and the value of the original data point is replaced by this average value. Environmental modeling: Divide the point cloud space into fixed-size voxels, calculate the centroid of the points in each voxel, and use the centroid to represent the voxel; initialize the environmental map into several fixed-size grids, and update the occupancy probability of each grid according to the environmental sensor data to build a complete environmental map, indicating whether each area is occupied by an obstacle; Obstacle recognition: Based on the fixed-size grids in the complete environment map, multiple bounding boxes and corresponding confidence scores are predicted for each grid. A forward pass is performed using a convolutional neural network to obtain the bounding boxes and categories of all targets, convert the identified obstacles into spatial coordinates, calculate their position and size, generate their own obstacle data, and use the occupancy probability and obstacle data of each grid as their own environment map.

4. The obstacle avoidance system for unmanned aerial vehicles based on multi-device interconnection according to claim 1 is characterized in that: The path planning module (300) calculates a self-obstacle avoidance path from a starting point to an end point according to the self-environment map using a path planning algorithm. The steps of calculating the self-obstacle avoidance path are as follows: Use a cost function to evaluate each node, including the actual cost and the estimated cost, where the actual cost is the actual cost of moving from the starting point to the current node; The estimated cost is the estimated cost from the current node to the target node; The cost function is the sum of these two parts; Manhattan distance is chosen as the heuristic function, which is the sum of horizontal and vertical distances from the current node to the target node; Add the starting node to the open list and set the cost of the starting node; Initialize the shutdown list to be empty; The open list is used to store the nodes to be checked, and the closed list is used to store the nodes that have been checked; Take the node with the smallest cost from the open list and use it as the current node; If the current node is the end point, terminate the search and the path is found; move the current node from the open list to the closed list; perform the following operations on each neighbor node of the current node: If the neighbor node is in the closed list, it is skipped; If the neighbor node is an obstacle in its own environment map, that is, the occupancy probability is higher than the standard threshold, it will be skipped; Calculate the temporary cost of the neighbor node, that is, the cost from the starting point through the current node to the neighbor node; If the neighbor node is not in the open list, or a path with a lower cost is found to reach the neighbor node, the parent node of the neighbor node is updated to the current node; The actual cost of updating neighbor nodes; Calculate and update the total cost of neighbor nodes; If the neighbor node is not in the open list, add it to the open list; When the end node is found, trace back from the end node along the parent node mapping to the starting node, collect the coordinates of all nodes, and generate a self-obstacle avoidance path from the starting point to the end point.

5. The unmanned aerial vehicle obstacle avoidance system based on multi-device interconnection according to claim 1 is characterized in that: The verification module (500) comprises an environment map verification unit (501), wherein the environment map verification unit (501) receives the environment maps to be verified and the obstacle avoidance paths to be verified sent by all other drones in the array, and performs verification, specifically comprising: Preliminary comparison: gradually compare the obstacles of each received environment map to be verified with the environment map itself, and compare the occupancy status of each grid in each two maps one by one, as follows: The environment map is represented as a two-dimensional grid, and each element in the grid represents the occupancy status of the location; Grid-by-grid comparison: create a consistent counter to record the number of consistent grids, and create a comparison counter to record the number of all compared grids; Comparison logic: For each grid position in the matrix, compare the values ​​of the environment map at that position with the value of the environment map to be verified. If the two values ​​are equal, increase the consistent grid counter; regardless of whether the two values ​​are equal, increase the comparison counter; Result statistics: Calculate the consistent grid ratio, that is, the number of consistent grids divided by the number of all compared grids, and judge the consistency. Set a consistency threshold. If the consistent grid ratio is higher than the consistency threshold, the two maps are considered consistent, otherwise they are considered inconsistent. Environmental map comparison and verification: record the number of results that are consistent with the environmental map comparison results of all other drones, and use the majority voting principle to determine the final environmental map comparison and verification results.

6. The unmanned aerial vehicle obstacle avoidance system based on multi-device interconnection according to claim 5 is characterized in that: The verification module (500) comprises an obstacle avoidance path verification unit (502), wherein the obstacle avoidance path verification unit (502) uses a path planning algorithm to predict obstacle avoidance paths for the environment maps to be verified sent by all other drones in sequence, and compares and verifies the obstacle avoidance paths in the prediction results with the obstacle avoidance paths to be verified in sequence, and the specific steps are as follows: Determine the key point sequence of the predicted obstacle avoidance path and the obstacle avoidance path to be verified, where the key point sequence includes turning points and waypoints in the path, representing the points and components where the path changes in direction. The turning point is the position where the direction changes in the path, and the waypoint is the middle point connecting two turning points in the path, which is used to describe the path in more detail. Create a structure to record the comparison results, which is used to record the comparison results of each key point; Node-by-node comparison: For each key point in the predicted obstacle avoidance path, the distance between the two key points is calculated. If the distance is within the standard error range, the two key points are considered similar, otherwise they are considered dissimilar. Record the obstacle avoidance path comparison results: record each comparison result, and calculate the similarity of the overall path based on the comparison results. The similarity of the overall path is the percentage of similar key points to the total number of key points. Set a similarity threshold. If the ratio of similar key points exceeds the similarity threshold, the paths are considered similar, and the predicted obstacle avoidance path is consistent with the obstacle avoidance path to be verified; otherwise, they are considered dissimilar, and the predicted obstacle avoidance path is inconsistent with the obstacle avoidance path to be verified. Finally, the majority voting principle is also used to take the majority comparison results as the verification results of the obstacle avoidance path.

7. The unmanned aerial vehicle obstacle avoidance system based on multi-device interconnection according to claim 6 is characterized in that: When the environmental map comparison verification calculated by the environmental map verification unit (501) fails, the record result of the environmental map comparison verification is sent to the drone corresponding to the inconsistent obstacle comparison result, and the corresponding environmental perception module (200) is optimized according to the record result, and the record includes the inconsistent grid position in the environmental map and the deviation information of the obstacle and its specific position; the drone receiving the comparison result will optimize its environmental perception module (200) according to the record, and the specific steps of the optimization include: Recalibrate sensor: adjust the calibration of the sensor; Adjusting algorithm parameters: adjusting the parameters of the algorithm in the environment perception module (200) according to the deviation information in the comparison record; Add more filtering steps: Add more filtering steps during data processing; Rescan the environment: Rescan the surrounding environment to update and correct the environment map.

8. The unmanned aerial vehicle obstacle avoidance system based on multi-device interconnection according to claim 7 is characterized in that: When the obstacle avoidance path verification result calculated by the obstacle avoidance path verification unit (502) fails to pass, the obstacle avoidance path comparison record result is sent to the drone corresponding to the inconsistent obstacle avoidance path comparison result, and the corresponding path planning module (300) is optimized according to the record result, and the record result includes the key point deviation and the overall similarity of the predicted obstacle avoidance path and the obstacle avoidance path to be verified; the drone receiving the comparison result will optimize its path planning module (300) according to these records, and the specific steps include: Optimize the path planning algorithm: adjust the path planning algorithm based on the key point deviation information in the comparison record; Adjust cost function parameters: Optimize the evaluation criteria of path planning by modifying the parameters in the cost function; Re-plan the path: Based on the updated environmental perception data and the optimized algorithm, a new obstacle avoidance path is re-planned and the obstacle avoidance path verification is re-performed until all drones confirm that the obstacle avoidance path verification has passed.