A UAV obstacle avoidance method and system based on bridge detection
Through the coordinated work of the three-level obstacle avoidance control center and multiple sensors, the autonomous obstacle avoidance of static and dynamic obstacles by drones in bridge detection is achieved, and the problem of low obstacle avoidance and coordination of drones in complex environments is solved. Improves safety and efficiency.
Patent Information
- Application Number
- CN202410757003.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-12
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-06-12
AI Technical Summary
In complex bridge environments, drone obstacle avoidance algorithms are difficult to deal with collisions and mission constraints between drones at the same time. Traditional methods have problems with low coordination and poor formation, especially when multiple drones work together. When it is difficult to effectively avoid static and dynamic obstacles, improve safety and efficiency.
A drone obstacle avoidance method and system based on bridge detection was designed. Through the three-level obstacle avoidance control center, the environment is sensed in real time, and the obstacle avoidance path is generated using multi-sensors and reinforcement learning algorithms to realize the autonomous obstacle avoidance of the drone, including Identification and obstacle avoidance strategies for unknown static obstacles, dynamic obstacles and other drones ensure that drones can safely bypass obstacles and complete detection tasks.
Improve the safety and autonomous obstacle avoidance capabilities of drones in unknown environments, ensure that drones can reach predetermined detection points, improve mission execution efficiency and safety, and solve the problems of coordination and formation maintenance in traditional methods. .
Smart Images

Figure CN118605608B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of unmanned aerial vehicle (UAV) path planning, and in particular relates to a UAV obstacle avoidance method and system based on bridge detection. Background Art
[0002] In recent years, drones are gradually replacing manned aircraft to perform high-risk inspection tasks due to their low cost, high mobility, and no casualties. As the working environment of drones gradually expands from medium and high altitudes to low altitudes, the number of environmental obstacles they encounter has also greatly increased. In the research of drone clusters, multi-drone obstacle avoidance control has become a hot topic. In the actual flight airspace, there may be obstacles such as buildings, mountains, and flocks of birds. The existence of these obstacles will threaten the flight safety of drones. In addition, during the obstacle avoidance process, the distance between drones will also change with the formation's obstacle avoidance maneuvers, and improper handling will easily cause collisions. Therefore, drone formations must be able to make decisions based on different environmental conditions while avoiding threatening obstacles and other drones.
[0003] At present, a lot of research has been done at home and abroad on the problems of UAV path planning algorithms and coordinated obstacle avoidance in adaptive environments, but there are generally defects such as low coordination and poor formation maintenance. Traditional UAV path planning algorithms often include A* algorithm, ant colony algorithm, particle swarm algorithm, etc. For UAV path planning tasks in specific environments, traditional algorithms are often improved and optimized. Many researchers consider the formation task and obstacle avoidance task of UAV swarms as a competitive relationship, and believe that the formation should be released first when avoiding obstacles, and then the formation should be restored after flying over the dangerous area to continue flying. However, under some environmental conditions, the UAV swarm can maintain the formation and fly through the obstacle space. In a complex bridge environment, the obstacle avoidance algorithm of the UAV must not only consider the collision problem of the UAV, but also meet some constraints of the UAV's own mission flight path.
[0004] Bridge structures belong to one of the large and medium-sized professional structures. Due to load effects, fatigue and corrosion effects, material embrittlement and lack of timely maintenance and repair, it is inevitable that damage accumulation and resistance decline will occur during the service life, shortening the service life of the bridge, and even causing safety accidents. Therefore, regular bridge inspections are very important. In the process of bridge inspections of different types of structures, traditional bridge inspections have many disadvantages. For example, when detecting dangerous places such as bridge piers, abutments, and belly bridges, there are safety hazards for inspectors. Therefore, in the face of complex bridge environments, using drones to conduct inspections can effectively avoid many disadvantages. When conducting some small bridge inspections, a single drone can complete the inspection quickly and effectively, but in the face of some large and complex bridges, multiple drones working together can improve the inspection efficiency.
[0005] When a drone is inspecting a bridge, obstacles such as large trees and high-voltage lines around the bridge will trigger the drone's emergency obstacle avoidance function. The current obstacle avoidance method for a single drone mainly relies on multi-sensor visual fusion and supporting deep learning or reinforcement learning algorithms. When multiple drones work together, the problem of collaborative obstacle avoidance between multiple drones will become more important. Therefore, it is worth exploring in depth to design a multi-drone collaborative obstacle avoidance method to achieve collaborative obstacle avoidance for multiple drones so that they can both maintain formation and avoid obstacles. Summary of the invention
[0006] The main purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art and to provide a UAV obstacle avoidance method and system based on bridge detection. The UAV navigates according to the obstacle avoidance path generated by path planning, and at the same time perceives changes in the surrounding environment in real time, and adjusts the path in real time according to sensor data to ensure that the UAV can safely bypass unknown static or dynamic obstacles, thereby improving the safety of the UAV in unknown environments and ensuring that the UAV can reach the predetermined detection point to complete the detection task. The autonomous obstacle avoidance capability of a single UAV and a UAV cluster can be improved, and the efficiency and safety of task execution can be improved.
[0007] According to one aspect of the present invention, the present invention provides a method for avoiding obstacles of a UAV based on bridge detection, the method comprising the following steps:
[0008] The drone obtains perception information including obstacle location information through scanning, and feeds the perception information back to the third-level obstacle avoidance control center;
[0009] The third-level obstacle avoidance control center processes and analyzes the perception information, identifies the location information of the obstacle, marks it on the navigation map of the drone, and generates an obstacle avoidance path for the drone;
[0010] The UAV inspects according to the obstacle avoidance path, senses changes in the surrounding environment in real time, determines the type of the perceived obstacle, and feeds back the type of the obstacle to the three-level obstacle avoidance control center, and the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of the obstacle.
[0011] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0012] If the type of obstacle is an unknown static obstacle, the third-level obstacle avoidance control center activates the multi-sensor collaborative working mechanism, analyzes the location information of the static obstacle, replans the flight path of the drone, and enables the drone to complete autonomous obstacle avoidance when encountering unknown obstacles.
[0013] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0014] If the type of obstacle is an unknown dynamic obstacle, the third-level obstacle avoidance control center uses the reinforcement learning obstacle avoidance method and the interactive training between the UAV and the environment to obtain the flight strategy, and generates the UAV flight obstacle avoidance path based on the flight strategy to realize autonomous obstacle avoidance of the UAV.
[0015] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0016] If the type of obstacle is another drone group, the third-level obstacle avoidance control center activates the drone group obstacle avoidance method. When multiple drone groups jointly carry out detection tasks, each drone group is regarded as a whole for path planning, and collision detection is performed as a group. By dividing the drone into sub-groups, autonomous obstacle avoidance is achieved.
[0017] Preferably, the method of realizing autonomous obstacle avoidance of the drone by dividing into subgroups includes:
[0018] The drone team leader obtains the location information of the team members to identify the initial formation of the current drone group;
[0019] The drone leader and team members sense the obstacle information ahead respectively. The team members feed back the obstacle information to the drone leader. The drone leader re-determines the position of the new formation of the drone team based on the obstacle information, divides the team into sub-teams, and sends the target position at the next moment to the team members.
[0020] The drone team leader divides the drone into sub-groups according to the target position at the next moment, reselects the team leader, changes the drone team formation, and completes obstacle avoidance;
[0021] The UAV team leader feeds back the obstacle avoidance results to the third-level obstacle avoidance control center, and the third-level obstacle avoidance control center updates the data information and optimizes the obstacle avoidance strategy.
[0022] According to another aspect of the present invention, the present invention also provides a UAV obstacle avoidance system based on bridge detection, the system comprising a three-level obstacle avoidance control center and a UAV;
[0023] The drone obtains perception information including obstacle location information by scanning, and feeds the perception information back to the third-level obstacle avoidance control center;
[0024] The third-level obstacle avoidance control center processes and analyzes the perception information, identifies the location information of the obstacle, marks it on the navigation map of the UAV, and generates an obstacle avoidance path for the UAV;
[0025] The UAV inspects according to the obstacle avoidance path, senses changes in the surrounding environment in real time, determines the type of the perceived obstacle, and feeds back the type of the obstacle to the three-level obstacle avoidance control center. The three-level obstacle avoidance control center adjusts the flight path in real time according to the type of the obstacle.
[0026] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0027] If the type of obstacle is an unknown static obstacle, the third-level obstacle avoidance control center activates the multi-sensor collaborative working mechanism, analyzes the location information of the static obstacle, replans the flight path of the drone, and enables the drone to complete autonomous obstacle avoidance when encountering unknown obstacles.
[0028] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0029] If the type of obstacle is an unknown dynamic obstacle, the third-level obstacle avoidance control center uses the reinforcement learning obstacle avoidance method and the interactive training between the UAV and the environment to obtain the flight strategy, and generates the UAV flight obstacle avoidance path based on the flight strategy to realize autonomous obstacle avoidance of the UAV.
[0030] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0031] If the type of obstacle is another drone group, the third-level obstacle avoidance control center activates the drone group obstacle avoidance method. When multiple drone groups jointly carry out detection tasks, each drone group is regarded as a whole for path planning, and collision detection is performed as a group. By dividing the drone into sub-groups, autonomous obstacle avoidance is achieved.
[0032] Preferably, the method of realizing autonomous obstacle avoidance of the drone by dividing into subgroups includes:
[0033] The drone team leader obtains the location information of the team members to identify the initial formation of the current drone group;
[0034] The drone leader and team members sense the obstacle information ahead respectively. The team members feed back the obstacle information to the drone leader. The drone leader re-determines the position of the new formation of the drone team based on the obstacle information, divides the team into sub-teams, and sends the target position at the next moment to the team members.
[0035] The drone team leader divides the drone into sub-groups according to the target position at the next moment, reselects the team leader, changes the drone team formation, and completes obstacle avoidance;
[0036] The UAV team leader feeds back the obstacle avoidance results to the third-level obstacle avoidance control center, and the third-level obstacle avoidance control center updates the data information and optimizes the obstacle avoidance strategy.
[0037] Beneficial effects: The UAV in the present invention navigates according to the obstacle avoidance path generated by path planning, and at the same time perceives changes in the surrounding environment in real time, and adjusts the path in real time according to sensor data, ensuring that the UAV can safely bypass unknown static or dynamic obstacles, thereby improving the safety of the UAV in unknown environments, and ensuring that the UAV can reach the predetermined detection point to complete the detection task, which can improve the autonomous obstacle avoidance capability of a single UAV or a UAV cluster, and improve the efficiency and safety of task execution.
[0038] The features and advantages of the present invention will become clear through reference to the following drawings and detailed description of specific embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 This is a flow chart of the obstacle avoidance method for UAV based on bridge detection;
[0040] Figure 2 It is a flow chart of the method implemented by the three-level obstacle avoidance control center;
[0041] Figure 3 This is a schematic diagram of the obstacle avoidance system for UAVs based on bridge detection;
[0042] Figure 4 Schematic diagram of multi-UAV bridge inspection and obstacle avoidance. DETAILED DESCRIPTION
[0043] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only partial embodiments of the present invention, not all 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.
[0044] Example 1
[0045] Figure 1 This is a flow chart of the obstacle avoidance method for UAVs based on bridge detection. Figure 1 As shown, this embodiment provides a UAV obstacle avoidance method based on bridge detection, and the method includes the following steps:
[0046] S1: The drone obtains perception information including obstacle location information through scanning, and feeds the perception information back to the third-level obstacle avoidance control center.
[0047] Specifically, the concepts involved in this embodiment are as follows:
[0048] Bridge modeling: The mission control center issues a bridge inspection mission, and the drone conducts a full-scale scan of the bridge and its surroundings, perceives the surroundings in real time, obtains the three-dimensional coordinates of tall obstacles such as bridges, and forms an electronic map. The collected data is processed and analyzed using image processing, deep learning and other methods, and tall obstacles such as bridges are identified and marked on the drone's navigation map, and fed back to the third-level obstacle avoidance control center.
[0049] Drone: Mainly responsible for implementing the inspection task of the target bridge, executing the planned flight path, electing a drone team leader in the drone cluster, and the team leader and team members work together. The team leader assigns the next inspection task.
[0050] Three-level obstacle avoidance control center: The three-level collision structure of drones and static obstacles, drones and dynamic obstacles (single drone and ordinary dynamic obstacles), and drone groups and drone groups can cope with different types of collisions in complex bridge environments. It is mainly responsible for collision detection, data processing, task release, etc. During the bridge detection process, the drone performs collision detection on obstacles, determines the collision safety distance, and allocates and executes the next collision avoidance plan. The drone replans the mission flight path according to the collision avoidance type, and at the same time perceives the changes in the surrounding environment in real time and adjusts the path in real time.
[0051] Monitoring strategy: Mainly responsible for monitoring the inspection effect of the entire bridge inspection process, receiving monitoring instructions from the mission control center, issuing monitoring tasks, and feeding back monitoring results. After the UAV achieves obstacle avoidance, it will feed back the obstacle avoidance effect to the obstacle avoidance control center to continuously optimize the obstacle avoidance performance of the UAV.
[0052] In this step, the mission control center determines the inspection task and sends out a drone to scan and model the target bridge and the surrounding environment. The drone generates an electronic map, obtains the three-dimensional coordinates of the main obstacles, and feeds them back to the third-level obstacle avoidance control center to analyze the data.
[0053] The specific steps are as follows:
[0054] Step 1.1: The drone is equipped with multiple sensors such as high-definition cameras, lidars, and IMU inertial meters to collect bridge data. The obstacle point cloud data obtained by the camera and lidar and the drone speed and attitude data obtained by the IMU inertial meter can perceive the surrounding environment in real time and obtain the three-dimensional coordinates of known obstacles.
[0055] Step 1.2: The third-level obstacle avoidance control center uses image processing, deep learning and other methods to process and analyze the collected data. By mapping the point cloud data, it can identify tall obstacles such as bridges and mark them on the drone's navigation map.
[0056] Step 1.3: Based on the obstacle information on the electronic map, the mission control center and the third-level obstacle avoidance control center use the path planning algorithm to generate an obstacle avoidance path for the drone based on the 3D point cloud map to ensure that the drone can bypass different types of tall obstacles such as bridges. If the obstacle information is updated, execute step 1.2.
[0057] S2: The third-level obstacle avoidance control center processes and analyzes the perception information, identifies the location information of the obstacle, marks it on the navigation map of the UAV, and generates an obstacle avoidance path for the UAV.
[0058] Specifically, for unknown obstacles that appear in the obstacle avoidance path, the various sensors on board are used to identify them and determine the type and category of the obstacle (dynamic obstacle, static obstacle or drone group). For example, the displacement of the same obstacle is compared in multiple sampling data of the sensor, and the actual displacement is calculated by transforming the relative coordinates with the drone into absolute coordinates relative to the ground. If an offset occurs, it means that it is a dynamic obstacle. There are mature algorithms for tracking point cloud data of the same object.
[0059] S3: The UAV inspects according to the obstacle avoidance path, senses changes in the surrounding environment in real time, determines the type of the perceived obstacle, and feeds back the type of the obstacle to the three-level obstacle avoidance control center. The three-level obstacle avoidance control center adjusts the flight path in real time according to the type of the obstacle.
[0060] Specifically, for unknown obstacles, data analysis is used to determine whether the safety distance threshold h has been reached. If so, feedback is sent to the third-level obstacle avoidance control center, the map information is updated, and it is marked on the drone's navigation map. Otherwise, the drone will fly along the established path.
[0061] If the obstacle appearing in the expected obstacle avoidance path is an unknown static obstacle, the UAV will urgently implement autonomous obstacle avoidance and adjust the flight direction.
[0062] If it is an unknown dynamic obstacle, the drone uses the various sensors it carries to analyze and predict the obstacle's location information at the next moment to achieve emergency obstacle avoidance.
[0063] If two groups of drones meet, the detection capabilities of the two groups of drones at the next moment are analyzed according to the complexity of the drone groups, and one of the groups of drones is divided into subgroups, and the flight direction is adjusted to achieve obstacle avoidance. For example, A[k] is drone group A, B[k] is drone group B, k is the number within the drone group, and by comparing the number of drones in drone groups A and B, the group with more drones is divided into subgroups.
[0064] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0065] If the type of obstacle is an unknown static obstacle, the third-level obstacle avoidance control center activates the multi-sensor collaborative working mechanism, analyzes the location information of the static obstacle, replans the flight path of the drone, and enables the drone to complete autonomous obstacle avoidance when encountering unknown obstacles.
[0066] Specifically, refer to Figure 2 ,The third level obstacle avoidance control center activates a multi-sensor collaborative working mechanism to analyze the location information of static obstacles to support the re-planning of the flight trajectory of the UAV, thereby enabling the UAV to complete autonomous obstacle avoidance when encountering unknown obstacles.
[0067] The specific steps are as follows:
[0068] The drone is equipped with multiple sensors such as high-definition cameras and two-dimensional lidars. It integrates the point cloud position information and IMU inertial meter information identified by multiple sensors, feeds them back to the three-level obstacle avoidance control center and converts them to the same coordinate system.
[0069] The drone is equipped with a variety of sensors to continuously collect point cloud information of obstacles in the environment around the bridge, identify the location information of the static obstacle P, obtain the real-time coordinates of P, and pre-process and analyze the data.
[0070] Update the coordinate information P (x, y, z) of the obstacle P on the 3D point cloud electronic map of the target bridge. Specifically, extract the ORB feature points in each frame of the image, and its depth value can be calculated by the depth map or binocular matching, so the 3D coordinates P of the obstacle under the bridge electronic map can be obtained. The specific calculation formula is as follows:
[0071]
[0072] k refers to the drone number, t refers to the tth moment, (x k ,y k , z k ) refers to the coordinates of drone k, (x kt ,y kt , z kt ) refers to the coordinates of UAV k at time t, and exp is the exponential operation function.
[0073] Taking the current position of the drone as the starting point, the path planning algorithm is used to replan the flight path and generate a new task detection path to achieve the obstacle avoidance effect. Among them, a certain point on the flight path needs to calculate its safety function to ensure that it will not encounter obstacles. H(x, y, z) represents the safety function of a certain point on the three-dimensional electronic map, N is the number of bridge detection points, M is the number of obstacle points, and α is the safety factor. The specific calculation formula is as follows:
[0074]
[0075] Create two new arrays D1 and D2. D1 is used to store the information of the detection points passed by the drone, and D2 is used to store the obstacle information. By traversing all the data in D1, find the location information of the obstacle P, delete it from D1 and add it to D2, and iterate in a loop to find all the detection point information sets to help the drone formation complete obstacle avoidance.
[0076] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0077] If the type of obstacle is an unknown dynamic obstacle, the third-level obstacle avoidance control center uses the reinforcement learning obstacle avoidance method and the interactive training between the UAV and the environment to obtain the flight strategy, and generates the UAV flight obstacle avoidance path based on the flight strategy to realize autonomous obstacle avoidance of the UAV.
[0078] Specifically, the third-level obstacle avoidance control center starts the obstacle avoidance method based on reinforcement learning. The reinforcement learning obstacle avoidance method is to continuously interact with the environment, perceive changes in surrounding environmental information, evaluate the resulting state changes, and optimize through training. Ultimately, the drone can take the best action in each state. After completing obstacle avoidance, it can return to the original planned path to continue the mission, thereby improving detection efficiency.
[0079] The specific steps are as follows:
[0080] The third-level obstacle avoidance control center uses reinforcement learning to avoid obstacles and uses interactive training between the drone and the environment to obtain flight strategies.
[0081] A UAV obstacle avoidance model is established in the three-dimensional space under the bridge environment. The number and type of dynamic obstacles, the position P, and the starting point O of the UAV are randomly generated. The autonomous obstacle avoidance capability of the UAV is continuously trained and optimized to form a mature obstacle avoidance model.
[0082] Analyze and determine the flight path of the dynamic obstacle at the next moment t, obtain the position information of the path collision point Q, and perform data preprocessing.
[0083] Update the position coordinate information of the path collision point Q on the electronic map of the three-dimensional model of the target bridge. The specific formula is as follows.
[0084] Q(x,y,z)=O(x,y,z)+P(x,y,z)
[0085] After analyzing the motion trajectory of the dynamic obstacle, the drone adjusts its flight direction in time through the optimized training model. For example, it uses the Minimum Snap method to generate a dynamically feasible drone flight obstacle avoidance path in the direction away from the dynamic obstacle trajectory. The starting and ending points of the path are both on the originally planned route, achieving autonomous obstacle avoidance. After the obstacle avoidance is completed, the drone returns to the original path in time to continue to complete the detection task.
[0086] In reinforcement learning, the state S is used as the optimization indicator to achieve the optimal path for the drone to avoid obstacles. The drone's three-level obstacle avoidance control center performs a lot of calculations and proofreading. The specific calculation formula is as follows.
[0087]
[0088] G(S) refers to the value function of state S, which indicates the characteristic manifestation of the optimal result of the obstacle avoidance model based on reinforcement learning during the training process. Y is the optimal index coefficient. By adjusting the size of the Y coefficient, the maturity of the obstacle avoidance model training is continuously tested and adjusted. k is the drone number, r is the k The detection capability of UAV k.
[0089] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0090] If the type of obstacle is another drone group, the third-level obstacle avoidance control center activates the drone group obstacle avoidance method. When multiple drone groups jointly carry out detection tasks, each drone group is regarded as a whole for path planning, and collision detection is performed as a group. By dividing the drone into sub-groups, autonomous obstacle avoidance is achieved.
[0091] Specifically, the third-level obstacle avoidance control center activates the obstacle avoidance method for the drone group. When multiple groups of drones carry out inspection tasks together, each drone group is regarded as a whole for path planning, and collision detection is performed on a group basis. By dividing the group into sub-groups, the effect of autonomous obstacle avoidance is achieved.
[0092] The specific steps are as follows:
[0093] A minimum UAV path trajectory circle collision detection mechanism is designed to effectively detect potential collisions between different UAV groups in linear time. The UAV groups are allowed to autonomously divide into subgroups to reduce the probability of deadlock between UAV groups.
[0094] A control model between drone groups is designed to control the overall avoidance tendency between drone groups or subgroups. The speeds of all drones in the drone group are vector-added and divided by the number of drones to generate the preferred group speed V.
[0095] The UAV group control model is designed, and the preferred group speed V is the UAV receiving group speed, so as to handle the collision detection of each UAV group and sub-group. The specific formula is as follows
[0096]
[0097] θ k It refers to the steering angle of UAV k, η is the steering angle coefficient, and V0 is the initial speed of the UAV.
[0098] A multi-task model optimization method is designed, and the inter-group and intra-group control models are calibrated simultaneously to achieve stable trajectory planning within and between UAV groups.
[0099] The drone group senses the surrounding environment and performs collision detection. It determines the group boundary by calculating the minimum outer circle size of the drone group and detects the collision probability between drone groups based on the minimum outer circle center point distance.
[0100] The inter-group control model is responsible for providing the group-level collision avoidance strategy of the drones and determining the group-level speed as the input of the intra-group control model to guide the overall movement of the drones within a group. The intra-group control model provides the collision avoidance speed of a single drone and controls the movement of individual drones within the drone group to achieve obstacle avoidance between groups. The cost function E is generated, and the specific formula is as follows.
[0101] E=∑(V-V0)*(KP)
[0102] E refers to the cost function generated by achieving the separation of a single drone from the drone group, K is the position information of drone k, and P is the position information of the obstacle.
[0103] Preferably, the method of realizing autonomous obstacle avoidance of the drone by dividing into subgroups includes:
[0104] The drone team leader obtains the location information of the team members to identify the initial formation of the current drone group;
[0105] The drone leader and team members sense the obstacle information ahead respectively. The team members feed back the obstacle information to the drone leader. The drone leader re-determines the position of the new formation of the drone team based on the obstacle information, divides the team into sub-teams, and sends the target position at the next moment to the team members.
[0106] The drone team leader divides the drone into sub-groups according to the target position at the next moment, reselects the team leader, changes the drone team formation, and completes obstacle avoidance;
[0107] The UAV team leader feeds back the obstacle avoidance results to the third-level obstacle avoidance control center, and the third-level obstacle avoidance control center updates the data information and optimizes the obstacle avoidance strategy.
[0108] Specifically, refer to Figure 4,For the situation of multi-UAV collaborative obstacle avoidance, the UAV team leader strategy is used to realize information sharing and coordination among UAVs, improve the efficiency of UAV division into sub-groups, quickly complete autonomous obstacle avoidance, and ensure the stability and safety of the entire system. At the same time, the UAV team leader strategy is used to generate a charging plan, optimizing and improving the efficiency of the UAV in the next detection task.
[0109] The specific steps are as follows:
[0110] A leader A0 is selected from the drone group, and other drones are group members. The drone leader identifies the initial formation of the current drone group by obtaining the position information of the group members.
[0111] The team leader and team members sense the obstacle information ahead respectively, and the team members feed back the obstacle information to the team leader. The team leader re-determines the position of the new formation of the drone team based on the obstacle information, divides the sub-groups, and sends the target position at the next moment to the team members. The specific formula is as follows.
[0112]
[0113] U(x, y, z) is the position information of the new drone group at the next moment, and U(x0, y0, z0) is the position information of the current drone group.
[0114] The drone group is divided into sub-groups according to the target position at the next moment, the team leader is re-selected, the drone group formation is changed, and obstacle avoidance is completed.
[0115] The drone team leader will feed back the obstacle avoidance results to the third-level obstacle avoidance control center, update the data information, and continuously optimize the obstacle avoidance method model.
[0116] The charging plan is generated by using the drone team leader strategy, and the charging order of the drones is reasonably allocated. The drones that cannot complete the next inspection task are given priority to return to the base station for charging. At the same time, the division of labor within the drone group is adjusted to optimize the efficiency of bridge inspection.
[0117] In this embodiment, the application of drone technology in bridge detection will break the limitations of traditional bridge detection methods, further develop the field of bridge detection, and be able to handle the detection tasks of various types of bridge structures, effectively avoiding the drawbacks of traditional detection methods. The multi-drone obstacle avoidance method based on bridge detection proposed in the present invention can be applied to fields such as bridge inspection and bridge maintenance, and can improve the autonomous obstacle avoidance capabilities of single drones and drone clusters, and improve the efficiency and safety of task execution. Traditional drone obstacle avoidance algorithms have certain limitations and are difficult to meet the detection tasks of drones in complex environments. The three-level obstacle avoidance control center structure proposed in the present invention systematically coordinates the obstacle avoidance algorithm, can effectively solve a series of complex collision types that occur in bridge detection, can cope with unknown obstacles that appear, and to a certain extent effectively ensure the safety of the drone execution task environment, and can improve the efficiency of drones in performing detection tasks.
[0118] The obstacle avoidance control concept of the drone group proposed in this embodiment is that within the drone group, each single drone can achieve local information interaction through distributed communication topology to avoid collision and maintain the task formation; outside the drone group, the virtual potential field is constructed using the improved artificial potential field method, so that the target position and obstacle information are reflected in the potential field value of each point. The drone determines the direction and speed of flight according to the change of the potential field value, so that after the drone group is divided into subgroups, it can avoid airspace obstacles and fly to the task detection point even without pre-planning the flight path. In the actual bridge inspection task, the drone navigates according to the obstacle avoidance path generated by the path planning, and at the same time perceives the changes in the surrounding environment in real time, adjusts the path in real time according to the sensor data, and takes the drone leader as the main strategy to ensure that the drone can safely bypass unknown static or dynamic obstacles, improve the safety of the drone in an unknown environment, and ensure that the drone can reach the predetermined detection point to complete the inspection task.
[0119] The drone team leader strategy proposed in this embodiment, on the one hand, can quickly divide subgroups when multiple drones are avoiding obstacles, generate new team leaders to complete the next detection task, optimize the obstacle avoidance performance, and improve the detection efficiency; on the other hand, it can generate a drone charging plan to ensure the continuity of the bridge detection process and improve the work efficiency of the drone.
[0120] Example 2
[0121] Figure 3 This is a schematic diagram of a UAV obstacle avoidance system based on bridge detection. Figure 3 As shown, this embodiment provides a UAV obstacle avoidance system based on bridge detection, the system includes a three-level obstacle avoidance control center and a UAV;
[0122] The drone obtains perception information including obstacle location information by scanning, and feeds the perception information back to the third-level obstacle avoidance control center;
[0123] The third-level obstacle avoidance control center processes and analyzes the perception information, identifies the location information of the obstacle, marks it on the navigation map of the UAV, and generates an obstacle avoidance path for the UAV;
[0124] The UAV inspects according to the obstacle avoidance path, senses changes in the surrounding environment in real time, determines the type of the perceived obstacle, and feeds back the type of the obstacle to the three-level obstacle avoidance control center. The three-level obstacle avoidance control center adjusts the flight path in real time according to the type of the obstacle.
[0125] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0126] If the type of obstacle is an unknown static obstacle, the third-level obstacle avoidance control center activates the multi-sensor collaborative working mechanism, analyzes the location information of the static obstacle, replans the flight path of the drone, and enables the drone to complete autonomous obstacle avoidance when encountering unknown obstacles.
[0127] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0128] If the type of obstacle is an unknown dynamic obstacle, the third-level obstacle avoidance control center uses the reinforcement learning obstacle avoidance method and the interactive training between the UAV and the environment to obtain the flight strategy, and generates the UAV flight obstacle avoidance path based on the flight strategy to realize autonomous obstacle avoidance of the UAV.
[0129] Preferably, the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including:
[0130] If the type of obstacle is another drone group, the third-level obstacle avoidance control center activates the drone group obstacle avoidance method. When multiple drone groups jointly carry out detection tasks, each drone group is regarded as a whole for path planning, and collision detection is performed as a group. By dividing the drone into sub-groups, autonomous obstacle avoidance is achieved.
[0131] Preferably, the method of realizing autonomous obstacle avoidance of the drone by dividing into subgroups includes:
[0132] The drone team leader obtains the location information of the team members to identify the initial formation of the current drone group;
[0133] The drone leader and team members sense the obstacle information ahead respectively. The team members feed back the obstacle information to the drone leader. The drone leader re-determines the position of the new formation of the drone team based on the obstacle information, divides the team into sub-teams, and sends the target position at the next moment to the team members.
[0134] The drone team leader divides the drone into sub-groups according to the target position at the next moment, reselects the team leader, changes the drone team formation, and completes obstacle avoidance;
[0135] The UAV team leader feeds back the obstacle avoidance results to the third-level obstacle avoidance control center, and the third-level obstacle avoidance control center updates the data information and optimizes the obstacle avoidance strategy.
[0136] The specific implementation process of the operations performed by the three-level obstacle avoidance control center, the drone, and the drone team leader in this embodiment 2 is the same as the implementation process in embodiment 1, and will not be repeated here.
[0137] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.
Claims
1. A UAV obstacle avoidance method based on bridge detection, characterized in that: The method comprises the following steps: The drone obtains perception information including obstacle location information through scanning, and feeds the perception information back to the third-level obstacle avoidance control center; The third-level obstacle avoidance control center processes and analyzes the perception information, identifies the location information of the obstacle, marks it on the navigation map of the drone, and generates an obstacle avoidance path for the drone; The UAV inspects according to the obstacle avoidance path, senses the changes in the surrounding environment in real time, determines the type of the perceived obstacle, and feeds back the type of the obstacle to the three-level obstacle avoidance control center, which adjusts the flight path in real time according to the type of the obstacle; the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of the obstacle, including: If the type of the obstacle is another drone group, the third-level obstacle avoidance control center starts the drone group obstacle avoidance method. When multiple drone groups jointly carry out the detection task, each drone group is regarded as a whole for path planning, and collision detection is performed on a group basis. By dividing the drones into subgroups, autonomous obstacle avoidance of the drones is achieved. The autonomous obstacle avoidance of the drones by dividing the drones into subgroups includes: The drone team leader obtains the location information of the team members to identify the initial formation of the current drone group; The drone leader and team members sense the obstacle information ahead respectively. The team members feed back the obstacle information to the drone leader. The drone leader re-determines the position of the new formation of the drone team based on the obstacle information, divides the team into sub-teams, and sends the target position at the next moment to the team members. The drone team leader divides the drone into sub-groups according to the target position at the next moment, reselects the team leader, changes the drone team formation, and completes obstacle avoidance; The UAV team leader feeds back the obstacle avoidance results to the third-level obstacle avoidance control center, and the third-level obstacle avoidance control center updates the data information and optimizes the obstacle avoidance strategy.
2. The method according to claim 1, characterized in that: The three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including: If the type of obstacle is an unknown static obstacle, the third-level obstacle avoidance control center activates the multi-sensor collaborative working mechanism, analyzes the location information of the static obstacle, replans the flight path of the drone, and enables the drone to complete autonomous obstacle avoidance when encountering unknown obstacles.
3. The method according to claim 1, characterized in that The three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including: If the type of obstacle is an unknown dynamic obstacle, the third-level obstacle avoidance control center uses the reinforcement learning obstacle avoidance method and the interactive training between the UAV and the environment to obtain the flight strategy, and generates the UAV flight obstacle avoidance path based on the flight strategy to realize autonomous obstacle avoidance of the UAV.
4. A UAV obstacle avoidance system based on bridge detection, characterized in that: The system includes a three-level obstacle avoidance control center and a drone; The drone obtains perception information including obstacle location information by scanning, and feeds the perception information back to the third-level obstacle avoidance control center; The third-level obstacle avoidance control center processes and analyzes the perception information, identifies the location information of the obstacle, marks it on the navigation map of the UAV, and generates an obstacle avoidance path for the UAV; The UAV inspects according to the obstacle avoidance path, senses changes in the surrounding environment in real time, determines the type of the perceived obstacle, and feeds back the type of the obstacle to the three-level obstacle avoidance control center, and the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of the obstacle; the three-level obstacle avoidance control center adjusts the flight path in real time according to the type of the obstacle, including: If the type of the obstacle is another drone group, the third-level obstacle avoidance control center starts the drone group obstacle avoidance method. When multiple drone groups jointly carry out the detection task, each drone group is regarded as a whole for path planning, and collision detection is performed on a group basis. By dividing the drones into subgroups, autonomous obstacle avoidance of the drones is achieved. The autonomous obstacle avoidance of the drones by dividing the drones into subgroups includes: The drone team leader obtains the location information of the team members to identify the initial formation of the current drone group; The drone leader and team members sense the obstacle information ahead respectively. The team members feed back the obstacle information to the drone leader. The drone leader re-determines the position of the new formation of the drone team based on the obstacle information, divides the team into sub-teams, and sends the target position at the next moment to the team members. The drone team leader divides the drone into sub-groups according to the target position at the next moment, reselects the team leader, changes the drone team formation, and completes obstacle avoidance; The UAV team leader feeds back the obstacle avoidance results to the third-level obstacle avoidance control center, and the third-level obstacle avoidance control center updates the data information and optimizes the obstacle avoidance strategy.
5. The system according to claim 4, characterized in that The three-level obstacle avoidance control center adjusts the flight path in real time according to the type of obstacle, including: If the type of obstacle is an unknown static obstacle, the third-level obstacle avoidance control center activates the multi-sensor collaborative working mechanism, analyzes the location information of the static obstacle, replans the flight path of the drone, and enables the drone to complete autonomous obstacle avoidance when encountering unknown obstacles.
6. The system according to claim 4, characterized in that The three-level obstacle avoidance control center makes real-time adjustments to the flight path according to the type of the obstacle, including: if the type of the obstacle is an unknown dynamic obstacle, the three-level obstacle avoidance control center uses a reinforcement learning obstacle avoidance method, utilizes interactive training between the UAV and the environment to obtain a flight strategy, generates a UAV flight obstacle avoidance path based on the flight strategy, and realizes autonomous obstacle avoidance of the UAV.
Citation Information
Patent Citations
Unmanned aerial vehicle cluster obstacle avoidance method
CN114428516A
Unmanned aerial vehicle control method and device and obstacle prompting method and device
CN114815863A
Multi-unmanned aerial vehicle global and local path intelligent planning method and system
CN115494866A
Unmanned aerial vehicle system with autonomous path planning and obstacle avoidance system
CN116755458A
Roadblock detection system, online autonomous decision-making unmanned aerial vehicle comprising detection system and application of online autonomous decision-making unmanned aerial vehicle
CN116880554A
Cited By
Pier fly-around path planning method based on Archimedes spiral
CN120722950A
