Positioning and navigation system and method for autonomous inspection of unmanned aerial vehicle, and computer program product

Abstract

The invention discloses a positioning and navigation system and method for autonomous inspection of an unmanned aerial vehicle and a computer program product, and the system comprises a data collection module, a data processing module, an SLAM core processing module, an unmanned aerial vehicle control execution module, a fault-tolerant module and a standby module. According to the method, the confidence coefficient is monitored in real time in the SLAM positioning link, the standby module is started immediately when the confidence coefficient is lower than the threshold value, hovering or hovering is triggered automatically, air crash and equipment loss can be avoided without manual intervention, meanwhile, the inspection task continuity and data integrity are guaranteed, and the safety, reliability and operation efficiency of tunnel unmanned aerial vehicle operation and maintenance are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of autonomous inspection technology for unmanned aerial vehicles (UAVs), specifically to a positioning and navigation system, method, and computer program product for autonomous inspection of UAVs. Background Technology

[0002] In tunnel operation, maintenance, and construction supervision, traditional manual inspections suffer from low efficiency, high operational risks, and delayed data processing. To address these issues, drone-based autonomous tunnel inspection technology has emerged: drones equipped with sensors such as lidar and high-definition cameras collect real-time data on tunnel structure and the environment, which is then wirelessly transmitted to the control center for processing and analysis. This provides an efficient solution for inspections, significantly improving work efficiency and data reliability.

[0003] Currently, this technology has been applied in fields such as industrial inspection (e.g., mines and chemical tunnels), building construction (tunnel surrounding rock deformation monitoring), and large venue safety (underground passage inspection).

[0004] However, the core simultaneous localization and mapping (SLAM) navigation system of the current mainstream solutions does not have an effective fault tolerance mechanism: tunnel closures and complex signals can easily lead to SLAM failures, causing drones to crash due to inability to navigate in emergencies and making them difficult to recall, resulting in equipment loss and interruption of inspections. Summary of the Invention

[0005] The technical problem to be solved by the embodiments of the present invention is to provide a positioning and navigation system, method and computer program product for autonomous inspection of unmanned aerial vehicles (UAVs) to improve the safety and reliability of UAV operation and maintenance in closed tunnel environments.

[0006] To address the aforementioned technical problems, this invention provides a positioning and navigation system for autonomous UAV inspection, comprising a data acquisition module, a data processing module, a SLAM core processing module, a UAV control and execution module, a fault-tolerant module, and a backup module, wherein: The data acquisition module is used to collect inspection environment data within the inspection area; The data processing module is electrically connected to the data acquisition module and is used to perform preliminary processing on the inspection environment data acquired by the data acquisition module. The SLAM core processing module is electrically connected to the data processing module and is used to calculate the UAV pose and construct a 3D map based on the pre-processed inspection environment data. The fault-tolerant module is connected to the SLAM core processing module and is used to monitor the location confidence in real time and output a switching command when the location confidence is lower than the threshold. The backup module is connected to the fault-tolerant module and the UAV control execution module, and is used to immediately activate the redundant sensor or backup positioning strategy after receiving the switching command, and trigger the UAV to hover or land to a preset safe point. The UAV control execution module is used to convert hovering or landing commands into flight control commands to drive the UAV to complete safe actions.

[0007] Preferably, the data acquisition module is a multi-sensor integrated unit, including at least a lidar, a vision sensor, and an IMU inertial measurement unit; The lidar is used to collect depth information and 3D point cloud data of the inspection environment, the vision sensor is used to collect visual feature data of the inspection environment, and the IMU inertial measurement unit is used to collect the motion acceleration, angular velocity and attitude angle data of the UAV. The output of the visual sensor is also electrically connected to a motion compensation module, which is used to achieve motion blur compensation by optimizing the exposure time of the visual sensor and enabling HDR image processing. The output of the motion compensation module is electrically connected to a dynamic object filtering module. The dynamic object filtering module is used to distinguish between static environment and moving target through time series analysis or semantic segmentation algorithm, and to remove interference data containing moving target. The output of the dynamic object filtering module is electrically connected to the input of the data processing module, and is used to transmit the quality-optimized inspection environment data to the data processing module.

[0008] Preferably, the SLAM core processing module integrates a data optimization component, which includes at least a data calibration and fusion module, a feature extraction module, a pose estimation module, and a map building module that are electrically connected in sequence. The data calibration and fusion module is used to perform time synchronization in the SLAM coordinate system, Kalman filter noise filtering, and multi-sensor data fusion on the pre-processed inspection environment data to generate a unified data input. The feature extraction module is used to extract key feature points from visual data and structural features from point cloud data. The pose estimation module is used to match feature data of adjacent frames and combine IMU motion data to calculate the real-time position and attitude of the UAV. The map building module is used to generate sparse point cloud maps or dense raster maps based on multi-frame pose data and feature point clouds.

[0009] Preferably, the data optimization component further includes an auxiliary positioning module and an auxiliary repair module; The auxiliary positioning module is connected in series with the map building module and is used to merge the newly collected observation data in real time with the existing map and dynamically update the local map. The auxiliary repair module is connected in series with the auxiliary positioning module. It is used to identify areas repeatedly traversed by the drone through bag-of-words model loop closure, and to correct accumulated errors by combining bundle adjustment algorithm to optimize the global positioning. Figure 1 To the point of being responsive.

[0010] Preferably, the fault-tolerant module adopts a dual-MCU redundancy design, and its real-time monitoring of location reliability includes at least the following methods: Compare the pose data output by the SLAM core processing module with the pose data output by the redundant IMU, and calculate the variance of their deviation. Monitor the feature point matching success rate of the SLAM algorithm in the core SLAM processing module; The thresholds include a location reliability threshold of 80% and a feature point matching success rate threshold of 60%. When the deviation variance exceeds the preset range or the feature point matching success rate is lower than 60%, the location reliability is determined to be lower than the threshold, and the fault tolerance module outputs a switching command.

[0011] Preferably, the backup module is connected to the fault-tolerant module via a CAN bus and to the UAV control execution module via an SPI interface; The redundant sensor includes at least a backup IMU and a UWB positioning module, and the UWB positioning module communicates with a UWB base station pre-deployed at key nodes in the inspection environment. The logic for triggering safety commands by the backup module is as follows: if the distance between the current position of the UAV and the preset safety point is less than or equal to the preset threshold, a landing command is triggered; otherwise, a hovering command is triggered first, and a landing command is triggered after the UWB positioning is stable.

[0012] Preferably, the UAV control execution module integrates a flight control motherboard, which converts hovering or landing commands into flight control commands in the following ways: The hover command is converted into a PWM signal command to maintain the motor speed; The descent command is converted into a PWM signal command for uniform descent. The UAV control execution module converts PWM signal commands into motor drive signals through an electronic speed controller (ESC) to control the UAV rotor speed to achieve hovering or precise landing.

[0013] Preferably, the system further includes a path planning and obstacle avoidance module, the input of which is electrically connected to the SLAM core processing module, and the output of which is connected in series with an auxiliary module; The auxiliary module includes at least an environment perception module, a path planning module, and a trajectory tracking module connected in series. The path planning module also includes a global path module and a local obstacle avoidance module connected in parallel. The global path module is used to plan the globally optimal path, and the local obstacle avoidance module uses a dynamic window method to adjust the local obstacle avoidance trajectory. The trajectory tracking module is used to track the deviation between the UAV and the planned path in real time and output a correction signal.

[0014] Preferably, the system further includes a closed-loop monitoring module and a map optimization module; The input terminal of the closed-loop monitoring module is electrically connected to the UAV control execution module and the data acquisition module, and is used to compare the UAV execution command with the actual sensor data to correct the pose error and identify the area repeatedly passed by the UAV to correct the map drift. The map optimization module is electrically connected to the closed-loop monitoring module and is used to optimize the accuracy of the three-dimensional map and eliminate redundant map data through map optimization technology or bundle adjustment technology.

[0015] Preferably, the system further includes a backup module, which is electrically connected to the output of the standby module and is used to store the 3D map data generated by the SLAM core processing module, the raw inspection data collected by the data acquisition module, and the fault logs of the fault-tolerant module and the standby module.

[0016] This invention also provides a positioning and navigation method for autonomous inspection by unmanned aerial vehicles, comprising: Step S1: Collect environmental data within the inspection area using the data acquisition module; Step S2: The collected environmental data is preliminarily processed by the data processing module electrically connected to the data acquisition module. Step S3: The SLAM core processing module, which is electrically connected to the data processing module, calculates the real-time pose of the UAV and constructs a three-dimensional map of the inspection area based on the pre-processed environmental data. Step S4: The fault-tolerant module connected to the SLAM core processing module monitors the positioning confidence level output by the SLAM core processing module in real time. If the positioning confidence level is lower than a preset threshold, a switching command is output. Step S5: After receiving the switching command, the backup module, which is connected to the fault-tolerant module and the UAV control execution module respectively, enables the redundant sensor or backup positioning strategy and transmits the hovering command or the preset safe point landing command to the UAV control execution module. Step S6: The drone control execution module converts the transmitted hovering command or preset safe point landing command into flight control commands, driving the drone to perform the safety action of hovering or landing at the preset safe point.

[0017] The present invention also provides a computer program product, characterized in that it includes computer instructions, which instruct a computer device to perform operations corresponding to the positioning and navigation method for the autonomous inspection of the unmanned aerial vehicle.

[0018] The present invention offers the following advantages: In extreme environments with weak tunnel lighting and multipath reflection, it provides UAVs with accurate anomaly detection, instantaneous redundancy switching, and safe hovering / return-to-home capabilities, completely overcoming the major drawbacks of traditional solutions where SLAM failure leads to crashes and irretrievable losses. The data optimization component performs multi-sensor time synchronization, Kalman noise filtering, and anomaly removal at the front end, ensuring clean and reliable input data. Feature extraction and pose estimation achieve high frame rate and low latency pose output. The auxiliary repair module continuously corrects accumulated errors through loop closure detection. The map building module generates centimeter-level dense point clouds or raster maps. The closed-loop monitoring module compares new and old observations in real time to dynamically eliminate map drift. The map optimization module further optimizes the global map. This invention significantly improves long-term positioning consistency and map accuracy. The path planning and obstacle avoidance module uses the environmental perception module to scan obstacles and wall structures in real time, achieving dual path planning and online trajectory adjustment to ensure safe obstacle avoidance during high-speed flight. The trajectory tracking module has a built-in independent locator supporting remote monitoring. The motion compensation module suppresses motion blur through exposure adaptation and HDR processing. The dynamic object filtering module uses semantic segmentation to remove moving targets, ensuring map purity. The fault-tolerant module deploys dual IMUs and redundant sensors to continuously monitor variance mutations. Once the positioning confidence falls below a threshold, the backup module is immediately activated, switching to the backup sensor or downgrading the positioning strategy, triggering hovering or returning to a preset safe point. The backup module simultaneously completes the original data write-to-disk, providing a complete record for post-event analysis. This invention achieves a closed-loop perception-decision-control system, significantly reducing equipment losses, manual intervention, and inspection costs, achieving the goal of high safety, high reliability, and high efficiency in autonomous tunnel UAV inspection. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the positioning and navigation system for autonomous inspection of unmanned aerial vehicles according to Embodiment 1 of the present invention.

[0021] Figure 2 This is a schematic diagram of the connection structure between the data acquisition module and the data processing module in an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram of the connection structure between the SLAM core processing module and the data optimization component in an embodiment of the present invention.

[0023] Figure 4 This is a schematic diagram of the connection structure between the path planning and obstacle avoidance module and the auxiliary module in an embodiment of the present invention.

[0024] Figure 5 This is a flowchart illustrating a positioning and navigation method for autonomous inspection by a drone according to Embodiment 2 of the present invention. Detailed Implementation

[0025] The following description of the embodiments is taken with reference to the accompanying drawings, which illustrate specific embodiments in which the invention can be implemented.

[0026] Please refer to Figure 1 As shown, Embodiment 1 of the present invention provides a positioning and navigation system for autonomous inspection of unmanned aerial vehicles (UAVs), including a data acquisition module 1, a data processing module 2, a SLAM core processing module 3, a UAV control and execution module 4, a fault-tolerant module 5, and a backup module 6, wherein: The data acquisition module 1 is used to collect inspection environment data within the inspection area; The data processing module 2 is electrically connected to the data acquisition module 1 and is used to perform preliminary processing on the inspection environment data acquired by the data acquisition module 1. The SLAM core processing module 3 is electrically connected to the data processing module 2 and is used to calculate the UAV pose and construct a three-dimensional map based on the pre-processed inspection environment data. The fault-tolerant module 5 is connected to the SLAM core processing module 3 and is used to monitor the location confidence in real time and output a switching command when the location confidence is lower than the threshold. The backup module 6 is connected to the fault-tolerant module 5 and the UAV control execution module 4, and is used to immediately activate the redundant sensor or backup positioning strategy after receiving the switching command, and trigger the UAV to hover or land to a preset safe point. The UAV control execution module 4 is used to convert hovering or landing commands into flight control commands to drive the UAV to complete safe actions.

[0027] As can be seen from the above settings, the embodiments of the present invention monitor the confidence level in real time during the SLAM positioning process and immediately activate the backup module when it falls below the threshold, automatically triggering hovering or return to base. This avoids crashes and equipment loss without manual intervention, while ensuring the continuity of inspection tasks and data integrity, significantly improving the safety, reliability, and operational efficiency of tunnel UAV operation and maintenance.

[0028] Specifically, in this embodiment of the invention, the data acquisition module 1 is a multi-sensor integrated unit, specifically equipped with: LiDAR is used to collect depth information and 3D point cloud data of the environment inside the tunnel, and can identify static structures such as the concavity and convexity of the tunnel wall and the location of pipelines. Visual sensors (such as cameras) are used to collect visual feature data inside the tunnel, such as wall markings and obstacle outlines; The IMU (Inertial Measurement Unit) is used to collect data on the UAV's motion acceleration, angular velocity, and attitude angles (roll, pitch, and yaw) to provide a dynamic reference for attitude calculation.

[0029] It is understandable that uneven lighting inside the tunnel (such as strong light at the entrance and weak light in the middle section) can easily lead to blurred images from visual sensors, and dynamic targets in dusty environments (such as construction workers and temporary tools) can easily interfere with SLAM positioning. Figure 2 As shown in this embodiment of the invention, the output of the visual sensor (camera) of the data acquisition module 1 is electrically connected to the motion compensation module 11. The motion compensation module 11 is used to address visual imaging problems caused by uneven lighting in the tunnel and the movement of drones. It optimizes the exposure time of the visual sensor and enables HDR image processing to achieve motion blur compensation, thereby improving the clarity of visual feature data. The output of the motion compensation module 11 is electrically connected to the dynamic object filtering module 12. The dynamic object filtering module 12 receives the motion-compensated visual data and the LiDAR point cloud data. Through time-series analysis (e.g., comparing the positional changes of three consecutive frames of point cloud data) or semantic segmentation algorithms (identifying the contours of moving targets), it distinguishes between static environments (such as walls and fixed pipelines) and moving targets (such as construction workers and temporary equipment) in the tunnel, eliminating interference data containing moving targets to ensure the noise-free output data. The output of the dynamic object filtering module 12 is electrically connected to the input of the data processing module 2, transmitting the quality-optimized inspection environment data (including LiDAR depth information, optimized visual features, and IMU motion acceleration) to the data processing module 2.

[0030] Data processing module 2 uses an embedded processor, and its main function is to perform general processing on the high-quality environmental data transmitted by dynamic object filtering module 12, specifically including: The lidar point cloud (.pcd format), visual images (.jpg format), and IMU data (.txt format) are uniformly converted into binary format to adapt to the data interface of downstream modules; The threshold filtering method is used to remove obviously abnormal data (such as the distance value of the ultra-tunnel length collected by lidar) to achieve preliminary data purification.

[0031] The output of data processing module 2 is electrically connected to SLAM core processing module 3, transmitting the processed general data to SLAM core processing module 3. For example... Figure 3 As shown, the SLAM core processing module 3 is electrically connected to the data processing module 2. Its functionality relies on the internally integrated data optimization component 30 (including a data calibration and fusion module 31, a feature extraction module 32, a pose estimation module 33, and a map building module 34), as detailed below: The data calibration and fusion module 31 calibrates the extrinsic parameters (relative position, angle) of the LiDAR and camera using a calibration matrix, and then fuses the dynamic data of the IMU to generate multi-source fused data in a unified coordinate system. Specifically, the data calibration and fusion module 31 is electrically connected to the data processing module 2 and is used to accurately process the initially processed inspection environment data based on the requirements of the SLAM algorithm: First, it achieves time synchronization in the SLAM coordinate system—using the timestamp of the IMU data as a reference, it aligns the time axis of the LiDAR point cloud and visual feature data with the operation time axis of the SLAM algorithm to ensure the time consistency of multi-sensor data in pose calculation; second, it performs SLAM-specific noise filtering on the IMU motion data through the Kalman filter algorithm to correct the dynamic cumulative error of the IMU; third, it fuses the depth data of the LiDAR, the feature data of the visual sensor, and the motion data of the IMU in the SLAM coordinate system to generate unified data input adapted to subsequent pose estimation.

[0032] The feature extraction module 32 uses the ORB (Oriented for Fast Feature Point Extraction and Description) algorithm to extract key feature points from the camera image and the voxel grid downsampling algorithm to extract structural features from the lidar point cloud to ensure the robustness of the feature points.

[0033] The pose estimation module 33 uses the extracted feature points and the ICP (Iterative Closest Point) algorithm to match adjacent frame data. It also combines the motion acceleration integral results of the IMU to calculate the real-time position (X / Y / Z coordinates) and attitude (yaw / pitch / roll angle) of the UAV.

[0034] The map building module 34 generates a sparse point cloud map (for routine inspection scenarios) or a dense grid map (for high-precision detection scenarios) within the tunnel based on multi-frame pose data and feature point clouds. The map data is stored in real time in the system's local storage unit (such as a 128GB SSD).

[0035] The fault-tolerant module 5 adopts a dual MCU (microcontroller unit) redundancy design and establishes a data connection with the SLAM core processing module 3 through a UART serial port. Its function is to monitor the positioning accuracy in real time and trigger switching commands. The specific implementation method is as follows: Position confidence monitoring: By comparing the pose data output by the SLAM core processing module 3 with the pose calculation results of the redundant IMU (which is the same model as the main IMU and has an independent power supply), the variance of the two is calculated; at the same time, the feature point matching success rate of the SLAM algorithm is monitored (for example, when the matching success rate is <60%, it is judged as a decrease in confidence). Threshold setting and command output: As an example, the preset location reliability threshold is 80% (that is, when the deviation variance is greater than the preset threshold or the matching success rate is less than 60%, the location reliability is determined to be lower than the threshold). At this time, the fault tolerance module 5 immediately outputs an electrical signal command to the backup module 6 to switch to the backup mode.

[0036] The backup module 6 is electrically connected to the fault-tolerant module 5 (connected via CAN bus) and the UAV control execution module 4 (connected via SPI interface). Its functions are divided into two parts: backup positioning activation and safety command triggering. Backup positioning activated: Upon receiving the switching command from fault-tolerant module 5, the redundant sensors (including backup IMU, UWB positioning module, and UWB base station pre-deployed at tunnel entrance and key nodes) are immediately activated. The backup strategy of backup IMU + UWB coarse positioning is adopted to replace the faulty SLAM positioning data and provide basic position reference for the safe operation of the UAV. Safety command trigger: Simultaneously transmit hovering command or preset safety point landing command to the UAV control execution module. If the current position of the UAV is ≤10m away from the preset safety point (such as a flat area set every 50m in the tunnel), the landing command is triggered; if the distance is >10m, the hovering command is triggered first, and the landing command is triggered after the backup positioning is stable.

[0037] The UAV control execution module 4 integrates the flight control motherboard, which is connected to the backup module 6 and the UAV power system (motor, ESC). Its main functions are command conversion and execution drive. Command conversion: Convert the hovering / landing commands transmitted by the backup module 6 into PWM signal commands that the flight controller can recognize (such as motor speed maintenance commands during hovering, uniform descent commands during landing, descent speed, etc.). Execution drive: The ESC converts PWM commands into motor drive signals to control the rotor speed of the drone, enabling stable hovering (unlimited hovering time until manual recall) or precise landing at a preset safe point.

[0038] To further improve inspection efficiency and environmental adaptability, this embodiment of the invention also integrates the following auxiliary modules, which work in conjunction with the core module: 1. Path planning and obstacle avoidance module 7 The input terminal of the path planning and obstacle avoidance module 7 is electrically connected to the SLAM core processing module 3, and is used to receive the real-time UAV pose data and 3D map data output by the SLAM core processing module 3; its output terminal is electrically connected in series with the auxiliary module 70, which serves as the execution auxiliary unit for the path planning and obstacle avoidance module 7, and further includes an environmental perception module 71, a path planning module 72, and a trajectory tracking module 73 connected in series. Please refer to... Figure 4 As shown, the environmental perception module 71 receives real-time tunnel environment data transmitted by lidar and camera. By scanning the tunnel interior space in real time, it extracts key environmental information such as obstacle distribution, wall structure outline, and pipeline direction to ensure that the path planning can fully adapt to the actual tunnel environment. The output of the environmental perception module 71 is electrically connected to the path planning module 72. The path planning module 72 combines the preset inspection task requirements (such as inspection along the tunnel central axis, round-trip inspection of key areas, etc.) and the real-time map provided by the SLAM core processing module 3 to first generate a preliminary flight path that meets the inspection coverage requirements. At the same time, it uses a deep learning target detection algorithm (such as YOLO algorithm) to identify obstacle data (such as temporary construction equipment, fallen tunnel lining fragments) transmitted by the environmental perception module 71 in real time and dynamically adjusts the preliminary path to avoid the path from overlapping with obstacles. In addition to being connected in series with the trajectory tracking module 73, the path planning module 72 also has another output terminal. This output terminal is electrically connected in parallel with the global path module 74 and the local obstacle avoidance module 75, forming a dual path guarantee mechanism: the global path module 74, based on the complete 3D map constructed by the SLAM core processing module 3, uses the A* algorithm or RRT algorithm to plan the globally optimal path from the tunnel entrance to the inspection endpoint. This path takes into account both inspection efficiency and tunnel structural characteristics (such as avoiding areas with large tunnel curvature); the local obstacle avoidance module 75 focuses on real-time sudden obstacles, receiving data from the data acquisition module... The high-frequency sensor data (such as 10Hz point cloud data from lidar) is used to quickly calculate the obstacle avoidance trajectory within a local area using the dynamic window method. When the UAV encounters temporary obstacles that are not mapped in advance on the global map (such as suddenly appearing staff), the local obstacle avoidance module 75 can output an adjusted local trajectory to ensure that the UAV avoids the obstacle in time. The output data of the global path module 74 and the local obstacle avoidance module 75 are fed back to the path planning module 72. The path planning module 72 integrates the macroscopic nature of the global path and the real-time nature of the local path to generate the final execution path. The input of the trajectory tracking module 73 is electrically connected to the path planning module 72. It integrates an independent locator (such as a UWB locator, which communicates with a UWB base station pre-deployed in the tunnel) to track the deviation of the UAV from the final execution path in real time. When the UAV is detected to deviate from the path due to airflow interference or sensor error, the trajectory tracking module 73 will output a deviation correction signal to the path planning and obstacle avoidance module 7, which will then fine-tune the path parameters to ensure that the UAV always flies along the planned path.

[0039] 2. Closed-loop monitoring module and map optimization module The closed-loop monitoring module and the map optimization module work together to maintain positioning accuracy. By dynamically correcting pose errors and optimizing map quality, they provide reliable pose and map data support for the path planning and obstacle avoidance modules. The specific implementation is as follows: The input of the closed-loop monitoring module is electrically connected to the UAV control execution module 4. It receives execution status data such as motor speed and flight attitude (roll angle, pitch angle) from the UAV control execution module 4, and simultaneously receives real-time sensor data (such as IMU motion acceleration data and camera visual feature data) transmitted from the data acquisition module 1. The main function of the closed-loop monitoring module is to achieve closed-loop feedback of perception-decision-control: on the one hand, by comparing the execution commands from the UAV control execution module 4 with the actual sensor data, it dynamically corrects the UAV's pose error (e.g., when the IMU detects a 0.5° pitch angle deviation in the UAV, the closed-loop monitoring module outputs a correction signal to the UAV control execution module 4 to adjust the motor speed to eliminate the deviation); on the other hand, it identifies areas repeatedly traversed by the UAV using a feature point matching algorithm (e.g., the middle section of a tunnel traversed during a return inspection), compares the visual features and point cloud data collected in the area before and after the inspection, and corrects map drift caused by accumulated errors in the SLAM algorithm (e.g., correcting the offset tunnel wall point cloud coordinates in the map to their actual positions). The output of the closed-loop monitoring module is electrically connected to the map optimization module. The map optimization module receives the pose data and map drift correction data corrected by the closed-loop monitoring module, and uses graph SLAM or bundle adjustment techniques to optimize the 3D map generated by the SLAM core processing module 3: by constructing a graph model composed of pose nodes and observation edges, the global pose error is minimized, and the positional accuracy of structural features (such as tunnel pipeline interfaces and wall markings) in the map is improved; at the same time, redundant point cloud data (such as repeatedly collected point clouds of the same wall) in the map is removed, the map storage capacity is compressed, and the consistency of map data is ensured, providing a higher accuracy map input for the path planning and obstacle avoidance module 7.

[0040] 3. Backup module The backup module is electrically connected to the output of the standby module 6. The backup module integrates a mechanical hard drive (such as a 1TB industrial-grade hard drive) to store key data during system operation: including 3D map data generated by the SLAM core processing module 3, raw inspection environment data collected by the data acquisition module 1, and fault logs of the fault tolerance module 5 and the standby module 6 (such as fault occurrence time, fault type, and switching command content). Even if the drone lands in the tunnel due to a fault, the staff can still read historical data through the backup module after recovering the drone, avoiding the loss of inspection data and providing a basis for fault analysis and system optimization.

[0041] Furthermore, the data optimization component 30 also includes an auxiliary positioning module 35 and an auxiliary repair module 36. The auxiliary positioning module 35 is connected in series with the map building module 34 (its input end is electrically connected to the output end of the map building module), and simultaneously receives new observation data (including lidar point clouds and visual image frames collected in real time during UAV flight) from the data acquisition module 1 through a data interaction interface. This new observation data is used to fuse with the existing initial map, achieving dynamic updates of the local map and ensuring consistency between the map and the actual tunnel environment. The specific processing logic is as follows: The module first performs feature matching on the new observation data: extracting planar features (such as tunnel walls and pipeline surfaces) from the new lidar point cloud and corresponding features from the initial map, and removing outliers in the feature matching to ensure matching accuracy; Secondly, based on the feature matching results, the point cloud / visual features corresponding to the new observation data are updated to a local area of ​​the initial map (the update range is centered on the current position of the UAV). If there are new structures (such as temporary construction supports) in the local area that are not covered by the initial map, the point cloud data corresponding to the new structures are added to the map. Finally, the dynamically updated local map data is output. On the one hand, it is fed back to the map building module 34 to update the basic data of the initial map. On the other hand, it is transmitted to the auxiliary repair module 36 for subsequent global error correction. At the same time, the obstacle distribution information of the local map is output to the path planning and obstacle avoidance module 7 to provide the latest environmental reference for the UAV to avoid obstacles in real time. The auxiliary repair module 36 is connected in series with the auxiliary positioning module 35 (its input is electrically connected to the output of the auxiliary positioning module via a high-speed data bus). It also receives UAV trajectory data (including the pose sequence of the UAV's historical flight path) transmitted by the closed-loop monitoring module. This data is used to correct the accumulated errors of the SLAM system through closed-loop detection, optimize global map consistency, and avoid map drift caused by long-term flight. The specific implementation method is as follows: Loop closure detection execution: The module uses the bag-of-words (BoW) model to compare the local map features output by the auxiliary positioning module with the map features corresponding to the UAV's historical flight trajectory to identify tunnel areas repeatedly traversed by the UAV (such as the previously inspected road sections traversed during the return trip from inspection). When the feature matching similarity is ≥85%, it is determined that a loop closure has been detected, that is, the current position of the UAV and a certain position in the past are in the same tunnel area. Cumulative error calculation: Based on the closed-loop detection results, the deviation between the current pose of the UAV and the pose of the same area in history (including position deviation and attitude deviation) is compared. An error function is constructed through a graph optimization algorithm (using the GraphSLAM framework) to calculate the cumulative error of the SLAM system since the last closed loop. Error correction and map optimization: Based on the calculated cumulative error, the coordinates of the point cloud data in the global map are corrected, and local map areas with drift are adjusted (such as correcting the tunnel wall point cloud that has shifted due to cumulative error to the actual spatial position). At the same time, the historical pose sequence of the UAV is optimized to ensure that the pose data is consistent with the corrected map data. Optimization output: The corrected global map data is fed back to the map building module 34 to update the global map benchmark of the system. At the same time, error correction parameters are output to the pose estimation module 33 to improve the accuracy of subsequent UAV pose calculation, and finally realize a positive cycle of map optimization - pose accuracy improvement - further accurate map update.

[0042] The workflow of this invention is divided into four stages: routine inspection, fault triggering, fault-tolerant emergency response, and safe recovery, which are described below: Phase 1: Routine Inspection Data acquisition module 1 activates the lidar, camera, and IMU to collect tunnel environment data; Motion compensation module 11 optimizes image quality, dynamic object filtering module 12 filters dynamic targets, and data processing module 2 preprocesses the data and then transmits it to SLAM core processing module 3. SLA core processing module 3 calculates the UAV pose and builds a 3D map, path planning and obstacle avoidance module 7 generates a global path and avoids obstacles, and UAV control execution module 4 drives the UAV to inspect along the path. The closed-loop monitoring module and map optimization module correct pose errors and map drift in real time, while the auxiliary positioning module and auxiliary repair module update and optimize the map. Phase 2: Fault Trigger If the dust concentration inside the tunnel suddenly increases, the feature point matching success rate of the SLAM core processing module 3 will drop to 50% (below the 60% threshold). When the fault-tolerant module detects that the location confidence level drops to 75% (below the 80% threshold), it immediately sends a switching command to the backup module 6. Phase 3: Fault Tolerance and Emergency Response The backup module 6 receives the switching command, activates the backup IMU and UWB positioning, and uses UWB positioning to confirm the distance between the drone and the nearest preset safety point. The backup module 6 transmits landing commands to the UAV control execution module 4, and simultaneously transmits a pause signal to the path planning and obstacle avoidance module 7. Phase 4: Safe Recycling The UAV control execution module 4 parses the landing command and drives the UAV to descend at a constant speed to the preset safe point; The backup module automatically stores the collected inspection data, allowing staff to retrieve the drone on-site or restart the SLAM system to continue inspections after the dust dissipates.

[0043] Please refer to Figure 5 As shown, corresponding to the positioning and navigation system for autonomous UAV inspection described in Embodiment 1 of the present invention, Embodiment 2 of the present invention also provides a positioning and navigation method for autonomous UAV inspection, comprising: Step S1: Collect environmental data within the inspection area using the data acquisition module; Step S2: The collected environmental data is preliminarily processed by the data processing module electrically connected to the data acquisition module. Step S3: The SLAM core processing module, which is electrically connected to the data processing module, calculates the real-time pose of the UAV and constructs a three-dimensional map of the inspection area based on the pre-processed environmental data. Step S4: The fault-tolerant module connected to the SLAM core processing module monitors the positioning confidence level output by the SLAM core processing module in real time. If the positioning confidence level is lower than a preset threshold, a switching command is output. Step S5: After receiving the switching command, the backup module, which is connected to the fault-tolerant module and the UAV control execution module respectively, enables the redundant sensor or backup positioning strategy and transmits the hovering command or the preset safe point landing command to the UAV control execution module. Step S6: The drone control execution module converts the transmitted hovering command or preset safe point landing command into flight control commands, driving the drone to perform the safety action of hovering or landing at the preset safe point.

[0044] Corresponding to the positioning and navigation system for autonomous UAV inspection described in Embodiment 1 of the present invention, Embodiment 3 of the present invention also provides a computer program product, including computer instructions, which instruct a computer device to perform the operation corresponding to the positioning and navigation method for autonomous UAV inspection described in Embodiment 2 of the present invention.

[0045] Preferably, the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor can be any conventional processor. The processor is the control center of the device, connecting various parts of the device through various interfaces and lines.

[0046] The memory mainly includes a program storage area and a data storage area. The program storage area can store the operating system, applications required for at least one function, etc., while the data storage area can store related data, etc. Furthermore, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard drive, a SmartMedia Card (SMC), a Secure Digital (SD) card, and a Flash Card, or other volatile solid-state storage devices.

[0047] It should be noted that the above-mentioned devices may include, but are not limited to, processors and memory, as will be understood by those skilled in the art.

[0048] As can be seen from the above description, compared with the prior art, the beneficial effects of the present invention are as follows: In extreme environments with weak tunnel lighting and multipath reflection of signals, the present invention provides UAVs with accurate anomaly detection, instantaneous redundancy switching, and safe hovering / return-home capabilities, completely overcoming the major defects of traditional solutions that lead to crashes and irretrievable losses due to SLAM failure; the data optimization component completes multi-sensor time synchronization, Kalman noise filtering, and anomaly removal at the front end, ensuring clean and reliable input data; feature extraction and pose estimation achieve high frame rate and low latency pose output; the auxiliary repair module continuously corrects accumulated errors through loop closure detection; the map building module generates centimeter-level dense point clouds or raster maps; the closed-loop monitoring module compares new and old observations in real time to dynamically eliminate map drift; and the map optimization module further optimizes the entire... The local map is optimized, significantly improving long-term positioning consistency and map accuracy. The path planning and obstacle avoidance module scans obstacles and wall structures in real time through the environmental perception module, achieving dual path planning and online trajectory adjustment to ensure safe obstacle avoidance during high-speed flight. The trajectory tracking module has a built-in independent locator supporting remote monitoring. The motion compensation module suppresses motion blur through exposure adaptation and HDR processing. The dynamic object filtering module uses semantic segmentation to remove moving targets, ensuring map purity. The fault tolerance module deploys dual IMUs and redundant sensors to continuously monitor variance mutations. Once the positioning confidence falls below a threshold, the backup module is immediately activated, switching to the backup sensor or downgrading the positioning strategy, and triggering hovering or returning to a preset safe point. The backup module simultaneously completes the original data write-to-disk, providing a complete record for post-event analysis. This invention achieves a closed-loop perception-decision-control chain, significantly reducing equipment losses, manual intervention, and inspection costs, achieving the goal of high safety, high reliability, and high efficiency in autonomous tunnel UAV inspection.

[0049] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A positioning and navigation system for autonomous inspection by unmanned aerial vehicles (UAVs), characterized in that, It includes a data acquisition module, a data processing module, a SLAM core processing module, a UAV control and execution module, a fault-tolerant module, and a backup module, among which: The data acquisition module is used to collect inspection environment data within the inspection area; The data processing module is electrically connected to the data acquisition module and is used to perform preliminary processing on the inspection environment data acquired by the data acquisition module. The SLAM core processing module is electrically connected to the data processing module and is used to calculate the UAV pose and construct a 3D map based on the pre-processed inspection environment data. The fault-tolerant module is connected to the SLAM core processing module and is used to monitor the location confidence in real time and output a switching command when the location confidence is lower than the threshold. The backup module is connected to the fault-tolerant module and the UAV control execution module, and is used to immediately activate the redundant sensor or backup positioning strategy after receiving the switching command, and trigger the UAV to hover or land to a preset safe point. The UAV control execution module is used to convert hovering or landing commands into flight control commands to drive the UAV to complete safe actions.

2. The system according to claim 1, characterized in that, The data acquisition module is a multi-sensor integrated unit, including at least a lidar, a vision sensor, and an IMU inertial measurement unit; The lidar is used to collect depth information and 3D point cloud data of the inspection environment, the vision sensor is used to collect visual feature data of the inspection environment, and the IMU inertial measurement unit is used to collect the motion acceleration, angular velocity and attitude angle data of the UAV. The output of the visual sensor is also electrically connected to a motion compensation module, which is used to achieve motion blur compensation by optimizing the exposure time of the visual sensor and enabling HDR image processing. The output of the motion compensation module is electrically connected to a dynamic object filtering module. The dynamic object filtering module is used to distinguish between static environment and moving target through time series analysis or semantic segmentation algorithm, and to remove interference data containing moving target. The output of the dynamic object filtering module is electrically connected to the input of the data processing module, and is used to transmit the quality-optimized inspection environment data to the data processing module.

3. The system according to claim 1, characterized in that, The SLAM core processing module integrates a data optimization component, which includes at least a data calibration and fusion module, a feature extraction module, a pose estimation module, and a map construction module that are electrically connected in sequence. The data calibration and fusion module is used to perform time synchronization in the SLAM coordinate system, Kalman filter noise filtering, and multi-sensor data fusion on the pre-processed inspection environment data to generate a unified data input. The feature extraction module is used to extract key feature points from visual data and structural features from point cloud data. The pose estimation module is used to match feature data of adjacent frames and combine IMU motion data to calculate the real-time position and attitude of the UAV. The map building module is used to generate sparse point cloud maps or dense raster maps based on multi-frame pose data and feature point clouds.

4. The system according to claim 3, characterized in that, The data optimization component also includes an auxiliary positioning module and an auxiliary repair module; The auxiliary positioning module is connected in series with the map building module and is used to merge the newly collected observation data in real time with the existing map and dynamically update the local map. The auxiliary repair module is connected in series with the auxiliary positioning module. It is used to identify areas repeatedly traversed by the UAV through bag-of-words model loop closure, and to correct accumulated errors by combining bundle adjustment algorithm to optimize global map consistency.

5. The system according to claim 1, characterized in that, The fault-tolerant module adopts a dual-MCU redundancy design, and its real-time monitoring of location reliability includes at least the following methods: Compare the pose data output by the SLAM core processing module with the pose data output by the redundant IMU, and calculate the variance of their deviation. Monitor the feature point matching success rate of the SLAM algorithm in the core SLAM processing module; The thresholds include a location reliability threshold of 80% and a feature point matching success rate threshold of 60%. When the deviation variance exceeds the preset range or the feature point matching success rate is lower than 60%, the location reliability is determined to be lower than the threshold, and the fault tolerance module outputs a switching command.

6. The system according to claim 1, characterized in that, The backup module is connected to the fault-tolerant module via a CAN bus and to the UAV control execution module via an SPI interface. The redundant sensor includes at least a backup IMU and a UWB positioning module, and the UWB positioning module communicates with a UWB base station pre-deployed at key nodes in the inspection environment. The logic for triggering safety commands by the backup module is as follows: if the distance between the current position of the UAV and the preset safety point is less than or equal to the preset threshold, a landing command is triggered; otherwise, a hovering command is triggered first, and a landing command is triggered after the UWB positioning is stable.

7. The system according to claim 1, characterized in that, The UAV control execution module integrates a flight control motherboard, which converts hovering or landing commands into flight control commands in the following ways: The hover command is converted into a PWM signal command to maintain the motor speed; The descent command is converted into a PWM signal command for uniform descent. The UAV control execution module converts PWM signal commands into motor drive signals through an electronic speed controller (ESC) to control the UAV rotor speed to achieve hovering or precise landing.

8. The system according to any one of claims 1-7, characterized in that, It also includes a path planning and obstacle avoidance module, the input of which is electrically connected to the SLAM core processing module, and the output of which is connected in series with an auxiliary module; The auxiliary module includes at least an environment perception module, a path planning module, and a trajectory tracking module connected in series. The path planning module also includes a global path module and a local obstacle avoidance module connected in parallel. The global path module is used to plan the globally optimal path, and the local obstacle avoidance module uses a dynamic window method to adjust the local obstacle avoidance trajectory. The trajectory tracking module is used to track the deviation between the UAV and the planned path in real time and output a correction signal.

9. The system according to any one of claims 1-7, characterized in that, It also includes a closed-loop monitoring module and a map optimization module; The input terminal of the closed-loop monitoring module is electrically connected to the UAV control execution module and the data acquisition module, and is used to compare the UAV execution command with the actual sensor data to correct the pose error and identify the area repeatedly passed by the UAV to correct the map drift. The map optimization module is electrically connected to the closed-loop monitoring module and is used to optimize the accuracy of the three-dimensional map and eliminate redundant map data through map optimization technology or bundle adjustment technology.

10. The system according to claim 1, characterized in that, It also includes a backup module, which is electrically connected to the output of the standby module and is used to store the 3D map data generated by the SLAM core processing module, the raw inspection data collected by the data acquisition module, and the fault logs of the fault tolerance module and the standby module.

11. A positioning and navigation method for autonomous inspection by unmanned aerial vehicles (UAVs), characterized in that, include: Step S1: Collect environmental data within the inspection area using the data acquisition module; Step S2: The collected environmental data is preliminarily processed by the data processing module electrically connected to the data acquisition module. Step S3: The SLAM core processing module, which is electrically connected to the data processing module, calculates the real-time pose of the UAV and constructs a three-dimensional map of the inspection area based on the pre-processed environmental data. Step S4: The fault-tolerant module connected to the SLAM core processing module monitors the positioning confidence level output by the SLAM core processing module in real time. If the positioning confidence level is lower than a preset threshold, a switching command is output. Step S5: After receiving the switching command, the backup module, which is connected to the fault-tolerant module and the UAV control execution module respectively, enables the redundant sensor or backup positioning strategy and transmits the hovering command or the preset safe point landing command to the UAV control execution module. Step S6: The drone control execution module converts the transmitted hovering command or preset safe point landing command into flight control commands, driving the drone to perform the safety action of hovering or landing at the preset safe point.

12. A computer program product, characterized in that, It includes computer instructions that instruct computer equipment to perform operations corresponding to the positioning and navigation method for autonomous UAV inspection as described in claim 11.

Images