A multi-sensor fusion pylon climbing robot obstacle avoidance method and system

Abstract

The present application relates to the technical field of robot navigation and control, and particularly relates to a multi-sensor fusion pylon climbing robot obstacle avoidance method and system, comprising: synchronously collecting environment data through a laser radar, an ultrasonic sensor, an infrared sensor and an inertial measurement unit; filtering and denoising the data and performing validity discrimination, entering fusion judgment when at least two sensors are valid; comparing the valid data with a preset threshold value, preliminarily classifying obstacles; through laser-infrared and laser-ultrasonic cross verification, combining with IMU attitude angle verification, comprehensively determining the danger level of the obstacles and generating an obstacle avoidance instruction. The system comprises a collection module, a discrimination module, a judgment module and a decision module. Through the multi-source data fusion and cross verification mechanism, the present application effectively distinguishes the pylon structure and the surface attachments, solves the problem of easy misjudgment of single sensor detection, and simultaneously through the hierarchical obstacle avoidance strategy and sensor failure degradation processing, significantly improves the operation safety of the pylon climbing robot in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of robot navigation and control technology, and more specifically, to a method and system for obstacle avoidance of a climbing robot using multi-sensor fusion. Background Technology

[0002] With the increasing demand for intelligent inspection of concrete piers in bridges and high-rise buildings, pier-climbing robots have become mainstream equipment due to their vertical climbing capabilities. However, in actual working conditions, the surface of the pier is often covered with foreign objects such as moss, rust, and cables, or suffers structural deformation such as peeling and bulging due to long-term service. The robot needs to accurately identify and avoid obstacles during dynamic climbing, otherwise it is prone to overturning or falling accidents.

[0003] Existing obstacle avoidance solutions mostly use a single sensor (such as lidar) to detect distance. However, a single data source is susceptible to environmental interference: lidar struggles to distinguish between changes in the curvature of the obstacle itself and surface attachments, ultrasonic sensors are prone to false detections on rough surfaces, and infrared sensors are unreliable due to lighting conditions. Furthermore, the lack of a cross-validation mechanism for multi-sensor data leads to a high rate of misclassification of obstacle types, and the absence of degradation logic in case of sensor failure further reduces the reliability of obstacle avoidance.

[0004] Therefore, how to integrate multi-source sensor data to accurately distinguish obstacle types and maintain basic obstacle avoidance capabilities when sensors malfunction has become a core technical challenge for the safe operation of climbing robots. Summary of the Invention

[0005] In view of this, the present invention addresses the shortcomings of the prior art by proposing a multi-sensor fusion obstacle avoidance method and system for a climbing robot, aiming to solve at least one of the problems mentioned in the background art.

[0006] In a first aspect, the present invention provides a multi-sensor fusion obstacle avoidance method for a climbing robot, comprising the following steps: synchronously collecting environmental perception data during the operation of the robot using multiple sensors installed on the climbing robot; The collected sensor data is filtered and denoised, and the validity of each sensor data is judged in turn. Sensor data that exceeds its own effective range is marked as invalid data. When all sensor data is invalid, it is determined that the sensor is faulty and an emergency braking is performed. When at least two sensor data are valid, it is determined to be valid data. The valid data is compared with a preset distance threshold to preliminarily determine whether there is an obstacle in front of the robot, and the type of obstacle is preliminarily classified according to the location and distance range of the obstacle to obtain a preliminary judgment result; Based on the preliminary judgment results, cross-validation is performed. First, laser-infrared verification is conducted. If the laser ranging is less than the preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed. Otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed. Then, laser-ultrasonic verification is performed. If both detect the obstacle and the ranging difference is within the preset error range, the obstacle is confirmed and obstacle avoidance is performed. Otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. At the same time, the corresponding attitude adjustment is performed based on the attitude angle verification results of the inertial measurement unit. The obstacle danger level is determined by combining the above verification and validation results, and an obstacle avoidance command is generated.

[0007] In some embodiments, the step of synchronously collecting environmental perception data during the operation of the climbing robot using multiple sensors installed on the robot includes: The various sensors include at least lidar, ultrasonic sensors, infrared sensors, and inertial measurement units; The lidar is used to acquire the distance and contour point cloud of obstacles in front, the ultrasonic sensor is used to detect the distance of obstacles at medium and short distances, the infrared sensor is used to detect the contact distance between the robot chassis and the pier surface, and the inertial measurement unit is used to detect the attitude angle of the robot body.

[0008] In some embodiments, the process of filtering and denoising the collected sensor data, and sequentially determining the validity of each sensor data point, marking sensor data exceeding its effective range as invalid data, determining sensor malfunction and performing emergency braking when all sensor data are invalid, and determining valid data when at least two sensor data points are valid, includes: Determine whether the number of effective sensors for the target is greater than or equal to the preset minimum sensor count threshold; When the number of effective sensors is less than the preset minimum sensor number threshold, the system is determined to be in a failure state, the robot brakes urgently and sends out a fault alarm signal. When the number of effective target sensors is greater than or equal to the preset minimum sensor number threshold, the system is determined to be in a degraded operation state, and the fusion judgment is made based on the current effective sensor data.

[0009] In some embodiments, comparing the valid data with a preset distance threshold to preliminarily determine whether there is an obstacle in front of the robot, and preliminarily classifying the type of obstacle according to its location and distance range to obtain a preliminary determination result includes: Determine the relationship between the distance in front of the robot and a preset safety threshold; When the distance in front of the robot is greater than the preset safety threshold, it is determined that the path ahead is clear, and the robot continues to climb. When the forward distance value is less than or equal to the preset safety threshold, the relationship between the forward distance value and the preset near distance threshold is determined, and the relationship between the point cloud density and the preset density threshold is also determined. When the forward distance value is greater than the preset near distance threshold and the point cloud density is greater than the preset density threshold, it is initially determined to be a protruding obstacle; When the forward distance value is less than the preset close distance threshold and the rate of change of the fitting spacing is within the preset rate of change threshold, it is initially determined that the curvature of the pier itself is changing. Determine whether the change in lateral distance value within a preset sampling period is consistent with the change in forward distance value; When the change in the lateral distance value exceeds a preset change threshold and is inconsistent with the change in the forward distance value, it is initially determined to be a lateral obstacle.

[0010] In some embodiments, the cross-verification based on the preliminary judgment result first involves laser-infrared verification. If the laser ranging is less than a preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed. Otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed. Then, laser-ultrasonic verification is performed. If both methods detect an obstacle and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is performed. Otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, the corresponding attitude adjustment is performed based on the attitude angle verification result of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance instructions by combining the above verification and calibration results, the process includes: When laser-infrared verification determines that the pier itself is structurally sound and laser-ultrasonic verification does not identify any obstacle, it is classified as level zero. When laser-infrared verification determines that it is a surface attachment and laser-ultrasonic verification confirms an obstacle, but attitude verification does not require adjustment, it is classified as Level 1. When laser-ultrasonic verification confirms an obstacle and attitude verification results trigger attitude adjustment, it is classified as Level 2; When laser-ultrasonic verification confirms an obstacle and the attitude verification results show that the attitude angle exceeds the preset threshold by a continuously increasing amount, it is determined to be level three; The obstacle avoidance control commands are executed according to the determined danger level. Level 0 means ignoring the obstacle and maintaining the original speed. Level 1 means decelerating to a preset percentage of the original speed and implementing a wall-hugging bypass strategy. Level 2 means stopping and shifting a preset distance to the obstacle-free side before continuing to move forward. Level 3 means emergency braking, reporting obstacle information, and waiting for manual intervention.

[0011] In some embodiments, the cross-verification based on the preliminary judgment result includes: first, laser-infrared verification; if the laser ranging is less than a preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed; otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed; then, laser-ultrasonic verification is performed; if both detect obstacles and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is performed; otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance instructions by combining the above verification and validation results, the method further includes: Determine whether obstacle features have disappeared in the re-acquired sensor data; When the obstacle features disappear, the obstacle avoidance is considered successful, and the original speed is resumed to continue climbing; If the obstacle features still exist, then determine the relationship between the current number of attempts and the preset upper limit of the number of attempts; If the current number of attempts is less than the preset maximum number of attempts, then the obstacle avoidance strategy level is upgraded and the attempt is repeated. When the current number of attempts is greater than or equal to the preset maximum number of attempts, it is determined that the attempt cannot be completed independently and the system is reported and awaits manual instructions.

[0012] In some embodiments, the cross-verification based on the preliminary judgment results includes: first, laser-infrared verification; if the laser ranging is less than a preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed; otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed; then, laser-ultrasonic verification is performed; if both detect obstacles and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is performed; otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance instructions by combining the above verification and validation results, the method further includes: Determine whether the ultrasonic sensor detects a change rate of distance value exceeding a preset change rate threshold in both the front and side views; When both the laser sensor and the infrared sensor determine that it is a surface attachment, and the ultrasonic sensor detects a change rate of distance value exceeding a preset change rate threshold in both the front and the side, it is confirmed as a stubborn attachment and a secondary obstacle avoidance strategy is executed. When both the laser sensor and the infrared sensor determine that it is a surface attachment, but the ultrasonic sensor only detects that the rate of change of the distance value in front exceeds the preset rate of change threshold while it does not exceed the threshold in the side, it is confirmed as a local small attachment and a first-level wall-hugging bypass strategy is executed. Determine whether the ultrasonic sensor detects a continuous echo in front; When the laser sensor detects an obstacle, the infrared sensor detects the curvature of the pier itself, and the ultrasonic sensor detects a continuous echo in front, obstacle avoidance is performed based on the data from the ultrasonic sensor.

[0013] In some embodiments, the cross-verification based on the preliminary judgment result includes: first, laser-infrared verification; if the laser ranging is less than a preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed; otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed; then, laser-ultrasonic verification is performed; if both detect obstacles and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is performed; otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance instructions by combining the above verification and validation results, the method further includes: During the execution of the wall-hugging bypass strategy, it is determined whether the real-time contact distance between the robot and the surface of the pier is within the preset contact range; When the real-time bonding distance exceeds the preset bonding range, the robot's motion trajectory is adjusted. Determine whether the robot has bypassed the attachments on the surface of the pier; If it fails to bypass the obstacle, it will continue to bypass the obstacle along an arc path and continuously monitor the contact gap of the infrared sensor. Once the detour has been completed, the wall-hugging detour strategy is discontinued, and the original driving path is restored. In the attitude constraint verification, it is determined whether the pitch angle is greater than a preset pitch angle threshold. When the pitch angle is greater than the preset pitch angle threshold, it is determined to be a slope obstacle and the speed is adjusted to low-speed climbing. Determine if the roll angle is greater than the preset roll angle threshold; When the roll angle is greater than the preset roll angle threshold, it is determined to be a roll risk and a leveling action is performed first. Determine whether the yaw angle deviation is greater than the preset yaw deviation threshold; When the yaw angle deviation is greater than the preset yaw deviation threshold, it is determined to be a path deviation and the course is corrected before obstacle avoidance is performed.

[0014] In some embodiments, the process of filtering and denoising the collected sensor data, and sequentially determining the validity of each sensor data point, marking sensor data exceeding its effective range as invalid data, and determining sensor malfunction and emergency braking when all sensor data are invalid, further includes determining valid data when at least two sensor data points are valid. When a sensor is marked as invalid data, determine whether the data of that sensor recovers to normal within a subsequent preset number of sampling periods; When the sensor data returns to normal, it is reintroduced into the fusion judgment. If the sensor data is not recovered, determine whether the duration of the sensor failure exceeds a preset time window; If the failure duration does not exceed the preset time window, the failure flag of the sensor is maintained and monitoring continues. If the duration of the failure exceeds the preset time window, the sensor is confirmed to have failed.

[0015] Secondly, the present invention provides a multi-sensor fusion obstacle avoidance system for a climbing robot, comprising: The data acquisition module is used to synchronously acquire environmental perception data during the operation of the robot through multiple sensors installed on the climbing robot; The discrimination module is used to filter and denoise the collected sensor data, and sequentially judge the validity of each sensor data. Sensor data that exceeds its own effective range is marked as invalid data. When all sensor data are invalid, it is determined that the sensor is faulty and an emergency braking is performed. When at least two sensor data are valid, it is determined to be valid data. The determination module is used to compare the valid data with a preset distance threshold, preliminarily determine whether there is an obstacle in front of the robot, and preliminarily classify the type of obstacle according to the location and distance range of the obstacle to obtain a preliminary determination result; The decision module is used to perform cross-validation based on the preliminary judgment results. First, laser-infrared verification is performed. If the laser ranging is less than the preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed. Otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed. Then, laser-ultrasonic verification is performed. If both detect the obstacle and the ranging difference is within the preset error range, the obstacle is confirmed and obstacle avoidance is performed. Otherwise, the obstacle level is reduced or obstacle avoidance is temporarily not performed. At the same time, the corresponding attitude adjustment is performed based on the attitude angle verification results of the inertial measurement unit. The obstacle danger level is determined by combining the above verification and validation results and obstacle avoidance instructions are generated.

[0016] Compared with existing technologies, the beneficial effects of this invention are mainly reflected in the following three aspects: First, by using multi-sensor data fusion and cross-validation logic from lidar, ultrasonic, infrared, and IMU, the accuracy of obstacle recognition is significantly improved. It can effectively distinguish between changes in the curvature of the obstacle itself and surface attachments, solving the problem of misjudgment caused by environmental interference from a single sensor. Second, a hierarchical obstacle avoidance decision-making mechanism is constructed, matching differentiated response strategies according to the obstacle's danger level (level zero to three), progressing step-by-step from ignoring obstacles to emergency braking. This ensures operational safety, avoids operational interruptions due to overreaction, and improves climbing efficiency. Finally, sensor failure monitoring and degradation operation logic are introduced. When some sensor data is invalid, the system can automatically switch to degradation mode to maintain basic obstacle avoidance functions, only applying emergency braking when all sensors fail. This greatly enhances the robot's environmental adaptability and system robustness under complex working conditions.

[0017] The above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure.

[0018] Other features and aspects of this disclosure will become clearer from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

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

[0020] Fig. 1 This is a flowchart of a multi-sensor fusion obstacle avoidance method for a climbing robot provided in an embodiment of the present invention; Fig. 2 This is a functional block diagram of a multi-sensor fusion obstacle avoidance system for a climbing robot provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0022] See Figs. 1-2As shown in the first embodiment, a multi-sensor fusion obstacle avoidance method for a climbing robot according to an embodiment of this application includes the following steps: Environmental perception data during the robot's operation is collected synchronously by multiple sensors installed on the climbing robot. The collected sensor data is filtered and denoised, and the validity of each sensor data is judged in turn. Sensor data that exceeds its own effective range is marked as invalid data. When all sensor data is invalid, it is determined that the sensor is faulty and an emergency braking is performed. When at least two sensor data are valid, it is determined to be valid data. The valid data is compared with a preset distance threshold to preliminarily determine whether there is an obstacle in front of the robot, and the type of obstacle is preliminarily classified according to the location and distance range of the obstacle to obtain a preliminary judgment result; Based on the preliminary judgment results, cross-validation is performed. First, laser-infrared verification is conducted. If the laser ranging is less than the preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed. Otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed. Then, laser-ultrasonic verification is performed. If both detect the obstacle and the ranging difference is within the preset error range, the obstacle is confirmed and obstacle avoidance is performed. Otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. At the same time, the corresponding attitude adjustment is performed based on the attitude angle verification results of the inertial measurement unit. The obstacle danger level is determined by combining the above verification and validation results, and an obstacle avoidance command is generated.

[0023] It should be understood that this method synchronously collects environmental perception data during the robot's operation using multiple sensors installed on the climbing robot, and performs comprehensive processing and fusion judgment on the collected data to ultimately achieve autonomous obstacle avoidance. Specifically, the multiple sensors include at least a lidar, an ultrasonic sensor, an infrared sensor, and an inertial measurement unit (IMU). The lidar is used to acquire distance information and contour point cloud data of obstacles in front, the ultrasonic sensor is used to detect the distance of obstacles in the short to medium range, the infrared sensor is used to detect the contact distance between the robot chassis and the surface of the pier, and the IMU is used to detect the robot's attitude angles, including pitch angle, roll angle, and yaw angle. The collected sensor data is first filtered and denoised to eliminate the influence of environmental noise and sensor noise on measurement accuracy. Then, the system sequentially judges the validity of each sensor data, marking sensor data that exceeds its effective range as invalid data. When all sensor data is invalid, the system determines that the sensor is faulty and performs emergency braking. When at least two sensor data are valid, the data is determined to be valid. Next, the system compares the valid data with a preset distance threshold to preliminarily determine whether there are obstacles in front of the robot. It then preliminarily classifies the obstacles based on their location and distance range, obtaining a preliminary judgment result. Based on this, the system performs cross-validation based on the preliminary judgment result. First, it performs laser-infrared verification. If the laser distance measurement is less than a preset safety threshold and the infrared contact distance changes abruptly, it is determined to be a surface attachment and the robot performs obstacle avoidance. Otherwise, it is determined to be the structure of the pier itself and obstacle avoidance is not performed. Then, it performs laser-ultrasonic verification. If both methods detect obstacles and the distance difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is performed. Otherwise, the obstacle level is lowered or obstacle avoidance is temporarily suspended. Simultaneously, the robot performs corresponding attitude adjustments based on the attitude angle verification results from the inertial measurement unit. Finally, the system combines the above verification and validation results to determine the obstacle's hazard level and generates an obstacle avoidance command.

[0024] Regarding data validity assessment, the system presets an effective measurement range for each sensor. For example, the effective range for lidar can be set to 0.1 meters to 30 meters, for ultrasonic sensors to 0.02 meters to 5 meters, for infrared sensors to 0.01 meters to 0.5 meters, and for inertial measurement units to measure attitude angles within ±180°. When data collected by a sensor exceeds its preset effective range, the data is marked as invalid. When all sensor data is invalid, it means the robot may be completely "blind," and emergency braking is the optimal choice to ensure safety. When at least two sensor data are valid, the system enters a fusion judgment mode. The preset minimum number of sensors can be set to two based on actual safety requirements. For preliminary judgment and classification, the system compares the distance values ​​in the valid data with a preset safety threshold. This threshold can be determined based on factors such as the robot's size, speed, and braking distance; for example, it can be set to 0.5 meters. If the distance ahead is greater than the safety threshold, it is considered unobstructed; if it is less than or equal to the safety threshold, further analysis is performed.

[0025] The system comprehensively considers multiple dimensions, including forward distance, point cloud density, rate of change of contact spacing, and lateral distance change. Through a combination of preset proximity thresholds, density thresholds, and rate of change thresholds, it can initially distinguish between different types of obstacles, such as protruding obstacles, changes in the curvature of the pier itself, and side obstacles. For cross-validation, laser-infrared verification utilizes two sensors with different principles—laser radar ranging and infrared contact spacing—for mutual verification. When the laser ranging is less than a safety threshold and the infrared contact spacing shows a sudden change, it is determined to be a surface attachment; conversely, if the infrared contact spacing changes steadily, it is determined to be a change in the curvature of the pier itself. Laser-ultrasonic verification uses two sensors with different ranging principles to cross-check distance values. If both detect an obstacle and the ranging difference is within a preset error range, the obstacle is confirmed to exist. If the ranging difference exceeds the error range, it may be due to interference with one sensor; in this case, the obstacle level is lowered or obstacle avoidance is temporarily suspended.

[0026] In terms of attitude verification and hazard level determination, the system presets pitch angle threshold, roll angle threshold and yaw angle deviation threshold. When the attitude angle exceeds these thresholds, the system needs to prioritize attitude adjustment. Combining the laser-infrared verification results, laser-ultrasonic verification results and attitude verification results, the system classifies the obstacle hazard level into several levels and generates corresponding obstacle avoidance control commands.

[0027] In a specific application scenario, a climbing robot is climbing up the surface of a power tower. Assuming the robot moves at a constant speed of 0.2 meters per second, LiDAR, ultrasonic sensors, infrared sensors, and inertial measurement units collect data synchronously at a sampling frequency of 50 Hz. In the first scenario, a protruding metal object is present on the surface of the pier. The lidar detects an obstacle 0.25 meters ahead, which is less than the preset safety threshold of 0.5 meters. The infrared sensor detects a jump in the contact distance from 0.03 meters to 0.08 meters, a jump of 0.05 meters, which is greater than the preset jump threshold of 0.02 meters. This is initially identified as a surface object. The lidar-infrared verification confirms the identification of the surface object. Subsequently, the ultrasonic sensor also detects an obstacle at 0.23 meters. The difference in distance between the two is 0.02 meters, which is less than the preset error threshold of 0.05 meters, confirming the existence of the obstacle. At this time, the inertial measurement unit detects a pitch angle of 8°, which is less than the threshold of 15°, and a roll angle of 3°, which is less than the threshold of 10°. The attitude is normal and no adjustment is required. The overall assessment determines the danger level to be Level 1. The system reduces the speed to 50% of the original speed, i.e., 0.1 meters per second, and executes a wall-hugging detour strategy.

[0028] In the second scenario, the surface of the pier is a natural arc-shaped protrusion. The lidar detected an obstacle 0.35 meters ahead, but the infrared sensor detected that the contact distance changed smoothly from 0.03 meters to 0.06 meters. The rate of change was 5% per second, which was less than the preset rate of change threshold of 10% per second. The lidar-infrared verification determined that it was the pier's own structure and no obstacle avoidance was performed. The robot continued to move forward at the original speed.

[0029] In the third scenario, there is a large obstacle in front and the robot's posture is tilted. Both the lidar and ultrasonic sensors detect that there is an obstacle 0.2 meters ahead. At the same time, the inertial measurement unit detects that the roll angle reaches 12°, which is greater than the preset threshold of 10°. The system determines the danger level to be level two, stops moving forward, shifts to the unobstructed side by a preset distance, such as 0.3 meters, and then continues to move forward. At the same time, the leveling action is performed first.

[0030] This implementation achieves significant technical benefits in several aspects through multi-sensor synchronous data acquisition and hierarchical fusion judgment. By setting a sensor validity discrimination mechanism and a minimum effective sensor number threshold, the system avoids the problem of a single sensor failure causing the entire obstacle avoidance system to collapse, significantly improving the system's fault tolerance and reliability. When all sensors fail, it can promptly apply emergency braking to prevent collisions. Through a laser-infrared verification mechanism, it effectively distinguishes between surface attachments and curvature changes of the pier itself, avoiding misjudging the pier's natural geometric features as obstacles and causing unnecessary obstacle avoidance actions, thus improving the accuracy of obstacle avoidance decisions and the robot's operational efficiency.

[0031] By employing a laser-ultrasonic verification mechanism, cross-validation is achieved using sensors with two different ranging principles. This effectively eliminates false alarms caused by environmental interference from a single sensor, such as light, color, or transparent objects, thus improving the reliability of obstacle identification. Attitude angle calibration via the inertial measurement unit ensures the robot maintains a stable posture during obstacle avoidance maneuvers, preventing tipping or loss of control due to abnormal posture and enhancing the safety of climbing operations. By comprehensively verifying and calibrating the results, the obstacle hazard level is determined, and obstacle avoidance commands are executed in stages. This achieves a gradient response strategy, from ignoring obstacles to decelerating and bypassing them, then stopping to avoid them, and finally emergency braking, balancing operational efficiency and safety.

[0032] In some specific embodiments, the step of synchronously collecting environmental perception data during the operation of the climbing robot using multiple sensors installed on the robot includes: The various sensors include at least lidar, ultrasonic sensors, infrared sensors, and inertial measurement units; The lidar is used to acquire the distance and contour point cloud of obstacles in front, the ultrasonic sensor is used to detect the distance of obstacles at medium and short distances, the infrared sensor is used to detect the contact distance between the robot chassis and the pier surface, and the inertial measurement unit is used to detect the attitude angle of the robot body.

[0033] It should be understood that the sensor configuration of the climbing robot is as follows: The lidar adopts a 16-line rotating lidar, installed on the top front of the robot, with a horizontal field of view of 360°, a vertical field of view of 30°, a ranging range of 0.1 meters to 30 meters, an accuracy of ±2 centimeters, and a scanning frequency of 10 Hz; the ultrasonic sensor adopts 4 ultrasonic transducers, installed in the front, back, left, and right directions of the robot, with a ranging range of 0.02 meters to 5 meters, an accuracy of ±1 centimeter, and an operating frequency of 40 kHz; the infrared sensor adopts an 8-point array infrared distance sensor, installed on the bottom of the robot chassis in a ring distribution, with a ranging range of 0.01 meters to 0.5 meters, an accuracy of ±1 millimeter, used to detect the contact distance between each point of the chassis and the surface of the pier; the inertial measurement unit adopts a 6-axis inertial measurement unit chip, including a 3-axis accelerometer and a 3-axis gyroscope, installed at the center of the robot body, with an acceleration range of ±16g, an angular velocity range of ±2000° / s, and an output frequency of 100 Hz. Four sensors are connected to the robot’s main control chip via a controller area network bus or a serial peripheral interface bus to achieve synchronous data acquisition. The main control chip is a microcontroller with an ARM Cortex-M7 core, which has sufficient computing power to complete tasks such as filtering, validity discrimination and fusion judgment.

[0034] This implementation achieves significant technical benefits in several aspects by defining the specific types and functional positioning of four complementary sensors. LiDAR provides high-precision distance measurement and contour point cloud information, offering a reliable data foundation for accurate obstacle localization and shape recognition. Ultrasonic sensors compensate for LiDAR's shortcomings in detecting transparent objects and in strong light environments; the two work together to achieve full-coverage detection of different types of obstacles. Infrared sensors are specifically used to detect the contact distance between the chassis and the pier surface, providing a unique basis for distinguishing between surface attachments and the pier's own structure—a key feature that differentiates this method from general obstacle avoidance schemes. An inertial measurement unit (IMU) monitors the robot's posture in real time, providing information on the robot's own motion state for obstacle avoidance decisions, ensuring that obstacle avoidance actions are executed under controllable posture. These four sensors complement each other functionally and corroborate each other's data, forming a multi-layered perception system: long-range laser detection, short-to-medium-range ultrasonic supplementation, short-range infrared precision measurement, and posture IMU monitoring. This significantly enhances the system's environmental perception capabilities in complex pier environments.

[0035] In some specific embodiments, the process of filtering and denoising the collected sensor data, and sequentially determining the validity of each sensor data point, marks sensor data exceeding its effective range as invalid data, determines a sensor malfunction and executes emergency braking when all sensor data are invalid, and determines the data as valid when at least two sensor data points are valid, includes: Determine whether the number of effective sensors for the target is greater than or equal to the preset minimum sensor count threshold; When the number of effective sensors is less than the preset minimum sensor number threshold, the system is determined to be in a failure state, the robot brakes urgently and sends out a fault alarm signal. When the number of effective target sensors is greater than or equal to the preset minimum sensor number threshold, the system is determined to be in a degraded operation state, and the fusion judgment is made based on the current effective sensor data.

[0036] It should be understood that this implementation establishes a system state classification mechanism based on the number of effective sensors. The preset minimum sensor number threshold is the minimum number of effective sensors required for normal system operation. Considering that this method uses four types of sensors—LiDAR, ultrasonic sensors, infrared sensors, and an inertial measurement unit—this threshold can be set to two. This means that as long as at least two sensors can provide valid data, the system can continue to operate, even if other sensors malfunction or provide invalid data. When the number of effective sensors is less than two, the system is determined to be in a failed state. In this case, the robot may face a situation where only one sensor is effective and cross-validation is impossible, or all sensors are ineffective and the robot is completely "blind." Continuing to operate in either of these situations could lead to collisions or falls. Therefore, the system performs an emergency stop, immediately halting all robot movements, and simultaneously sends a fault alarm signal via audible and visual alarms or wireless communication to notify the operator for inspection and intervention.

[0037] When the number of valid sensors is greater than or equal to two, the system is determined to be in a degraded operation state. Degraded operation means that although the system can continue to run, its perception capability and decision-making accuracy may be lower than when all sensors are working normally due to the failure of some sensors. In this state, the system relies on the currently valid sensor data for fusion judgment, but the confidence level of obstacle avoidance decisions may be correspondingly reduced. For example, if an infrared sensor fails, the system will be unable to perform laser-infrared verification. In this case, it may be necessary to lower the obstacle hazard level judgment standard and adopt a more conservative obstacle avoidance strategy. Furthermore, when a sensor is marked as invalid data, the system continues to monitor whether the sensor's data recovers within a preset number of sampling periods. The preset number can be set to 10 consecutive sampling periods. If the data recovers, it is reinstated into the fusion judgment. If the data does not recover, it is determined whether the duration of the sensor failure exceeds a preset time window, such as 5 seconds. If it does not exceed this window, the failure mark is maintained and monitoring continues. If it exceeds this window, the sensor is confirmed to be faulty, the system enters a degraded operation state, and a failure log is recorded.

[0038] In one specific embodiment, during the climbing process, the ultrasonic sensor of the robot continuously output data exceeding its effective range by 0.02 meters to 5 meters due to strong wind interference. The system marked this sensor's data as invalid. At this time, the data from the lidar, infrared sensor, and inertial measurement unit were all normal, with a valid sensor count of 3, exceeding the preset minimum sensor count threshold of 2. The system determined that it was in a degraded operation state. The system continued to monitor the ultrasonic sensor data. After 5 consecutive sampling cycles (0.1 seconds), the ultrasonic sensor data returned to normal, and the system re-included it in the fusion judgment. Throughout the entire process, the robot continued to operate relying on the data from the lidar, infrared sensor, and inertial measurement unit without interruption. In another scenario, the infrared sensor continuously outputs invalid data due to hardware failure, failing to recover after 20 consecutive sampling cycles. The failure duration reaches 1 second, exceeding the preset time window of 0.5 seconds. The system confirms the infrared sensor failure and enters a degraded operation state. At this time, the effective sensors are the lidar, ultrasonic sensor, and inertial measurement unit, totaling 3. Simultaneously, a fault alarm signal indicating the infrared sensor failure is sent to the remote control terminal via wireless communication. Since there are still 3 effective sensors and more than the threshold of 2, the robot continues to perform obstacle avoidance tasks, but skips the laser-infrared verification step and relies solely on laser-ultrasonic verification to confirm obstacles.

[0039] This implementation achieves several significant technical benefits by establishing a system state hierarchy mechanism based on the number of effective sensors. By setting a preset minimum sensor count threshold, the minimum conditions for normal system operation are clearly defined, avoiding safety risks caused by blind operation with insufficient sensing capabilities. Emergency braking and fault alarms in system failure states ensure the robot can stop promptly when a large number of sensors fail, preventing collisions or falls and protecting the safety of the equipment and surrounding environment. The setting of degraded operating states allows the system to continue completing tasks even with partial sensor failure, improving the system's fault tolerance and task completion rate, and preventing the entire task from being interrupted due to a single sensor failure. The dynamic recovery monitoring and re-inclusion mechanism for sensors enables automatic management of sensor states and adaptive adjustment of the system, automatically restoring full functionality without manual intervention once sensors recover. Setting a failure duration threshold prevents sensors from being prematurely identified as failed due to occasional interference, improving the accuracy of system state determination.

[0040] In some specific embodiments, the step of comparing the valid data with a preset distance threshold to preliminarily determine whether there is an obstacle in front of the robot, and preliminarily classifying the type of obstacle according to its location and distance range, to obtain a preliminary determination result, includes: Determine the relationship between the distance in front of the robot and a preset safety threshold; When the distance in front of the robot is greater than the preset safety threshold, it is determined that the path ahead is clear, and the robot continues to climb. When the forward distance value is less than or equal to the preset safety threshold, the relationship between the forward distance value and the preset near distance threshold is determined, and the relationship between the point cloud density and the preset density threshold is also determined. When the forward distance value is greater than the preset near distance threshold and the point cloud density is greater than the preset density threshold, it is initially determined to be a protruding obstacle; When the forward distance value is less than the preset close distance threshold and the rate of change of the fitting spacing is within the preset rate of change threshold, it is initially determined that the curvature of the pier itself is changing. Determine whether the change in lateral distance value within a preset sampling period is consistent with the change in forward distance value; When the change in the lateral distance value exceeds a preset change threshold and is inconsistent with the change in the forward distance value, it is initially determined to be a lateral obstacle.

[0041] It should be understood that this implementation establishes a preliminary obstacle classification mechanism based on multi-dimensional parameters. A preset safety threshold is the basic threshold for determining whether an obstacle exists in front of the robot. This threshold can be determined based on factors such as the robot's size, speed, and braking distance. For example, for a robot climbing at a speed of 0.2 meters per second, it can be set to 0.5 meters. When the distance in front is greater than 0.5 meters, the path ahead is considered clear and the robot can continue moving forward; when it is less than or equal to 0.5 meters, further analysis is required. A preset proximity threshold is used to distinguish between distant obstacles and closely spaced obstacles; for example, it can be set to 0.2 meters. A preset density threshold is used to determine the density of the point cloud data, reflecting the geometric features of the obstacle surface; for example, it can be set to 50 points per square meter.

[0042] When the distance in front is greater than a preset proximity threshold (e.g., 0.25 meters) and the point cloud density is greater than a preset density threshold (e.g., 120 points per square meter), it indicates the presence of a moderately sized and clearly defined object in front, initially identified as a protruding obstacle. When the distance in front is less than a preset proximity threshold (e.g., 0.15 meters) and the rate of change of the contact gap is within a preset rate of change threshold (i.e., the change is gradual), it indicates the robot is very close to the pier surface, but the contact gap is changing steadily, initially identified as a change in the curvature of the pier itself, meaning the robot is climbing the curved surface of the pier rather than encountering an obstacle. Simultaneously, the system also monitors the lateral distances, i.e., the distances to the left and right sides of the robot. Within a preset sampling period (e.g., 0.5 seconds), it calculates the changes in lateral and forward distances. If the change in lateral distance exceeds a preset change threshold (e.g., 0.05 meters) and is inconsistent with the change in forward distance, it indicates the presence of an abnormal object to the side of the robot, such as a lateral protrusion on the pier surface or a side obstacle, initially identified as a side obstacle. If the changes in both are consistent, it indicates the robot is moving towards or away from the pier surface as a whole, rather than encountering a side obstacle.

[0043] In one specific embodiment, the climbing robot climbs the surface of the pier, and the system collects data every 0.1 seconds. In the first case, the lidar measures a distance of 0.28 meters in front, which is less than the preset safety threshold of 0.5 meters but greater than the preset close-range threshold of 0.2 meters. The point cloud density is 120 points per square meter, which is greater than the preset density threshold of 50 points per square meter. The system initially determines this to be a protruding obstacle and proceeds to the subsequent cross-validation process. In the second case, the lidar measures a distance of 0.18 meters in front, which is less than the preset close-range threshold of 0.2 meters. The infrared sensor detects that the contact gap changes smoothly from 0.03 meters to 0.05 meters, with a change rate of 4% per second, which is less than the preset change rate threshold of 10% per second. The system initially determines this to be a change in the curvature of the pier itself and does not perform obstacle avoidance. In the third case, within a 0.5-second sampling period, the distance value in front changes from 0.35 meters to 0.33 meters, a change of 0.02 meters, while the distance value on the right side changes from 0.15 meters to 0.22 meters, a change of 0.07 meters, which is greater than the preset change threshold of 0.05 meters. Furthermore, the trends of the two changes are inconsistent, and the system initially determines that it is a side obstacle.

[0044] This implementation achieves significant technical benefits in several aspects by establishing a preliminary classification mechanism based on multi-dimensional parameters. By comparing preset safety thresholds, it enables rapid preliminary judgment of obstacle presence, improving the system's response speed. By comparing the forward distance value with a preset near-distance threshold and the point cloud density with a preset density threshold, it can initially distinguish between protruding obstacles and changes in the curvature of the pier itself, providing valuable pre-classification information for subsequent cross-validation. Through consistency analysis of lateral distance changes and forward distance changes, it achieves effective identification of lateral obstacles, expanding the perception dimension of the obstacle avoidance system and enabling the robot to not only cope with forward obstacles but also perceive lateral risks.

[0045] By setting a preset threshold for the rate of change, the criteria for judging the smooth change of the mating distance are quantified, making the judgment of the curvature change of the pier itself repeatable and objective, and avoiding the uncertainty caused by subjective judgment. The entire preliminary judgment process does not rely on data from a single sensor, but integrates multiple parameters such as distance value, point cloud density, mating distance change rate, and lateral change, thereby improving the accuracy and robustness of the preliminary judgment.

[0046] In some specific embodiments, the cross-verification based on the preliminary judgment result involves first performing laser-infrared verification. If the laser ranging is less than a preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is executed. Otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed. Then, laser-ultrasonic verification is performed. If both methods detect obstacles and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is executed. Otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance instructions by combining the above verification and validation results, the process includes: When laser-infrared verification determines that the pier itself is structurally sound and laser-ultrasonic verification does not identify any obstacle, it is classified as level zero. When laser-infrared verification determines that it is a surface attachment and laser-ultrasonic verification confirms an obstacle, but attitude verification does not require adjustment, it is classified as Level 1. When laser-ultrasonic verification confirms an obstacle and attitude verification results trigger attitude adjustment, it is classified as Level 2; When laser-ultrasonic verification confirms an obstacle and the attitude verification results show that the attitude angle exceeds the preset threshold by a continuously increasing amount, it is determined to be level three; The obstacle avoidance control commands are executed according to the determined danger level. Level 0 means ignoring the obstacle and maintaining the original speed. Level 1 means decelerating to a preset percentage of the original speed and implementing a wall-hugging bypass strategy. Level 2 means stopping and shifting a preset distance to the obstacle-free side before continuing to move forward. Level 3 means emergency braking, reporting obstacle information, and waiting for manual intervention.

[0047] It should be understood that this implementation establishes a multi-factor comprehensive assessment-based obstacle hazard level classification system. Level zero is the lowest hazard level, corresponding to the situation where laser-infrared verification determines that the pier itself is not an obstacle and laser-ultrasonic verification does not confirm an obstacle. In this case, the comprehensive judgment is that there is no object in front that needs to be avoided, and the robot should ignore the obstacle and continue to move forward at the original speed.

[0048] Level 1 corresponds to low-risk obstacles. The judgment conditions are that laser-infrared verification determines that it is a surface attachment, laser-ultrasonic verification confirms the existence of the obstacle, and the attitude verification result does not need to be adjusted. At this time, it is confirmed that there is an attachment-type obstacle in front, but the robot's attitude is normal and no additional adjustment is needed. The corresponding obstacle avoidance command is to decelerate to a preset ratio of the original speed, such as 50%, and execute the wall-hugging bypass strategy.

[0049] Level 2 corresponds to medium-risk obstacles. The judgment condition is that the obstacle is confirmed by laser-ultrasonic verification and the attitude verification result triggers the attitude adjustment requirement, such as the pitch angle or roll angle exceeding the threshold. At this time, not only is there an obstacle, but the robot's attitude is also abnormal. It is necessary to deal with two problems at the same time: obstacle avoidance and attitude adjustment. The corresponding obstacle avoidance command is to stop moving forward and shift to the unobstructed side by a preset distance, such as 0.3 meters, and then continue moving forward.

[0050] Level 3 corresponds to the highest risk obstacle. The judgment condition is that the obstacle is confirmed by laser-ultrasound verification and the attitude verification result shows that the attitude angle exceeds the preset threshold by a continuously increasing amount. This means that the robot has not only encountered an obstacle but its attitude is also continuously deteriorating, and there is a risk of tipping over or losing control. The corresponding obstacle avoidance command is to immediately stop all movement by emergency braking and report the obstacle information via wireless communication to wait for manual intervention.

[0051] In one specific embodiment, the system determines the hazard level and performs obstacle avoidance according to the following process. In the first scenario, laser-infrared verification determines it to be the structure of the pier itself, laser-ultrasonic verification does not identify an obstacle, and the attitude angles of the inertial measurement unit are all within the threshold range. The system determines it to be level zero, and the robot continues to move forward at its original speed of 0.2 meters per second.

[0052] In the second scenario, laser-infrared verification identifies the object as a surface attachment, and laser-ultrasonic verification confirms the presence of an obstacle 0.25 meters ahead. The inertial measurement unit detects that the pitch angle is less than the threshold of 15° (7°), the roll angle is less than the threshold of 10° (4°), and the yaw angle deviation is less than the threshold of 5° (2°). Since the attitude does not need to be adjusted, the system is determined to be at level one, and the speed is reduced to 0.1 meters per second, which is 50% of the original speed. The wall-hugging bypass strategy is executed, and the robot bypasses the attachment along the surface of the pier in an arc path.

[0053] In the third scenario, laser-ultrasonic verification confirms the presence of an obstacle 0.2 meters ahead. The inertial measurement unit detects a roll angle exceeding the 10° threshold by 12°, triggering an attitude adjustment requirement. The system is classified as Level 2, stops advancing, shifts 0.3 meters to the right (towards the obstacle-free side), and then continues advancing while simultaneously performing a leveling maneuver to restore the roll angle to within the threshold. In the fourth scenario, laser-ultrasonic verification confirms the presence of an obstacle 0.15 meters ahead. The inertial measurement unit detects a roll angle continuously increasing from 8° to 14°, 16°, and 18°, exceeding the 10° threshold by a continuously increasing margin. The system is classified as Level 3, immediately applies emergency braking, and transmits obstacle information, including location, type, and attitude angle data, to a remote control terminal via 4G or 5G communication, awaiting manual instructions.

[0054] This implementation achieves significant technical benefits in several aspects by establishing a four-level hazard classification system and corresponding graded obstacle avoidance commands. Level 0 judgment effectively filters out false obstacles such as those inherent to the pier's structure, avoiding unnecessary obstacle avoidance actions and improving the robot's operational efficiency and climbing smoothness. Level 1 judgment and a wall-hugging detour strategy enable precise handling of surface-attached obstacles, avoiding both overly aggressive stopping actions that impact efficiency and ignoring them that could lead to collisions. Level 2 judgment and a stop-off-continue strategy mitigate risks by actively stopping and deflecting when encountering medium-risk obstacles with abnormal posture, while also adjusting posture to ensure safety. Level 3 judgment and an emergency braking strategy decisively stop the robot in the highest-risk situation—when obstacles approach and posture continues to deteriorate—effectively preventing collisions and rollover accidents and ensuring the safety of equipment and personnel. The progressive relationship between the levels forms a gradient response mechanism from ignoring to deceleration and detour, then to stopping and deflecting, and finally to emergency braking, employing differentiated response strategies for different risk levels, balancing operational efficiency and safety.

[0055] In some specific embodiments, the cross-verification based on the preliminary judgment result includes: first, laser-infrared verification; if the laser ranging is less than a preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed; otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed; then, laser-ultrasonic verification is performed; if both detect obstacles and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is performed; otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance instructions by combining the above verification and validation results, the method further includes: Determine whether obstacle features have disappeared in the re-acquired sensor data; When the obstacle features disappear, the obstacle avoidance is considered successful, and the original speed is resumed to continue climbing; If the obstacle features still exist, then determine the relationship between the current number of attempts and the preset upper limit of the number of attempts; If the current number of attempts is less than the preset maximum number of attempts, then the obstacle avoidance strategy level is upgraded and the attempt is repeated. When the current number of attempts is greater than or equal to the preset maximum number of attempts, it is determined that the attempt cannot be completed independently and the system is reported and awaits manual instructions.

[0056] It should be understood that this implementation establishes a mechanism for verifying obstacle avoidance effectiveness and upgrading strategies. After executing obstacle avoidance commands such as wall-hugging or offset avoidance, the system re-collects sensor data to determine whether the previously detected obstacle features have disappeared. The criteria for determining whether obstacle features have disappeared may include the lidar ranging value recovering to above a preset safety threshold, the point cloud density dropping below a preset density threshold, and the infrared contact gap returning to a stable state. When these conditions are met, obstacle avoidance is determined to be successful, and the robot resumes its original speed to continue climbing.

[0057] If the obstacle features still exist, it means that the current obstacle avoidance strategy has failed to bypass the obstacle. At this time, the system judges the relationship between the current number of attempts and the preset upper limit of attempts, such as 3 times. If the current number of attempts is less than the upper limit, the obstacle avoidance strategy level is upgraded and the attempt is repeated, for example, from level 1 to level 2 or from level 2 to level 3. If the current number of attempts has reached or exceeded the upper limit, such as 3 attempts and still failed, it is determined that the robot cannot pass through the obstacle autonomously, and the obstacle information is reported and waiting for human instructions.

[0058] In one specific embodiment, after executing the first-level wall-hugging detour strategy, the climbing robot re-collects data to verify the obstacle avoidance effect. In a successful scenario, the robot performs wall-hugging detour at a speed of 0.1 meters per second, with an arc path radius of 0.3 meters. During the detour, the system continuously monitors the forward LiDAR data. After 2 seconds, the LiDAR measures the forward distance to be restored to 0.8 meters, which is greater than the safety threshold of 0.5 meters. The point cloud density drops to 10 points per square meter, which is less than the density threshold of 50 points per square meter. The infrared contact spacing returns to a stable level. The system determines that obstacle avoidance is successful, the speed is restored to 0.2 meters per second, and the robot continues to climb along the original path.

[0059] In the failure and upgrade scenario, after the robot executes the first-level wall-hugging avoidance strategy, the LiDAR still detects an obstacle 0.28 meters ahead, with a point cloud density of 100 points per square meter. The current number of attempts is 1, less than the maximum of 3. The system upgrades to a second-level strategy, stopping its forward movement and shifting 0.3 meters to the right before continuing. If the second-level strategy also fails, the number of attempts increases to 2, and the system upgrades to a third-level strategy, namely emergency braking and reporting the failure. In the scenario where the robot cannot pass autonomously, after three attempts—first-level wall-hugging avoidance, second-level offset avoidance, and third-level emergency braking—the obstacle features still exist. The current number of attempts has reached the maximum of 3. The system determines that it cannot pass autonomously and reports obstacle information, including location coordinates, obstacle type, and attempt records, via wireless communication, awaiting remote human intervention.

[0060] This implementation achieves several significant technical benefits by establishing an obstacle avoidance effectiveness verification and strategy upgrade mechanism. The obstacle avoidance success determination mechanism enables automatic verification of obstacle avoidance effectiveness, ensuring the robot can promptly resume normal operation after successfully bypassing obstacles, thus improving operational efficiency. The attempt limit and strategy upgrade mechanism enable adaptive adjustment of the obstacle avoidance strategy, gradually upgrading from low-risk to high-risk strategies, attempting autonomous passage as much as possible while ensuring safety. Setting a preset attempt limit prevents the robot from getting stuck in an infinite loop of attempts at insurmountable obstacles, ensuring task controllability. The reporting and waiting mechanism for failure to autonomously pass ensures timely human intervention when system capacity is insufficient, preventing equipment damage or safety accidents caused by blind attempts.

[0061] In some specific embodiments, the cross-verification based on the preliminary judgment result includes: first, laser-infrared verification; if the laser ranging is less than a preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed; otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed; then, laser-ultrasonic verification is performed; if both detect obstacles and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is performed; otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance instructions by combining the above verification and validation results, the method further includes: Determine whether the ultrasonic sensor detects a change rate of distance value exceeding a preset change rate threshold in both the front and side views; When both the laser sensor and the infrared sensor determine that it is a surface attachment, and the ultrasonic sensor detects a change rate of distance value exceeding the preset change rate threshold in both the front and the side, it is confirmed as a stubborn attachment and a secondary obstacle avoidance strategy is executed. When both the laser sensor and the infrared sensor determine that it is a surface attachment, but the ultrasonic sensor only detects that the rate of change of the distance value in front exceeds the preset rate of change threshold while it does not exceed the threshold in the side, it is confirmed as a local small attachment and a first-level wall-hugging bypass strategy is executed. Determine whether the ultrasonic sensor detects a continuous echo in front; When the laser sensor detects an obstacle, the infrared sensor detects the curvature of the pier itself, and the ultrasonic sensor detects a continuous echo in front, obstacle avoidance is performed based on the data from the ultrasonic sensor.

[0062] It should be understood that this embodiment utilizes the multi-directional detection capability of ultrasonic sensors for refined classification of attachments and establishes a resolution mechanism in case of sensor conflicts. The ultrasonic sensors are positioned in both the front and side directions, enabling simultaneous detection of distance changes in both directions. A preset change rate threshold, such as 0.1 meters per second, is used to determine if the distance value has changed rapidly. When both the laser and infrared sensors identify a surface attachment, the ultrasonic data is further analyzed: if the ultrasonic sensors detect a distance change rate exceeding the preset change rate threshold in both the front and side directions (i.e., the distances in both directions are changing rapidly), it indicates that the attachment has a large extension range in both the front-back and left-right directions, confirming it as a stubborn attachment—that is, a large attachment that is difficult to avoid by simple detours. A secondary obstacle avoidance strategy is implemented, i.e., stopping forward and shifting to the unobstructed side. If the ultrasonic sensors only detect a distance change rate exceeding the threshold in the front and not in the side, it indicates that the attachment exists only in a localized area in front, with sufficient passage space in the side, confirming it as a small, localized attachment. A primary wall-hugging detour strategy is implemented to bypass it.

[0063] Regarding sensor conflict resolution, when the laser sensor determines an obstacle (i.e., the distance measured is less than the safety threshold), the infrared sensor determines it is the curvature of the pier itself (i.e., the contact gap changes gently), and the ultrasonic sensor detects a continuous echo in front (i.e., the echo signal is detected in multiple consecutive sampling cycles), the data from the ultrasonic sensor is used to execute obstacle avoidance. This resolution rule is based on the fact that the laser sensor may be affected by transparent objects or mirror reflections, leading to misjudgments. The infrared sensor can only detect the contact situation near the chassis and is not sensitive to obstacles a little further ahead. The ultrasonic sensor is not affected by light and color and has good detection capabilities for transparent objects. Therefore, in such conflict scenarios, the data from the ultrasonic sensor has higher reliability.

[0064] In one specific embodiment, in a scenario with persistent attachments, the laser sensor detects an obstacle 0.22 meters ahead, and the infrared sensor detects a jump in the contact distance from 0.03 meters to 0.07 meters. This jump of 0.04 meters is greater than the threshold of 0.02 meters, and together they are determined to be surface attachments. The ultrasonic sensor data shows that the distance value in front changes from 0.30 meters to 0.20 meters within 0.5 seconds, with a change rate of 0.2 meters per second, which is greater than the threshold of 0.1 meters per second. The lateral distance value changes from 0.25 meters to 0.18 meters, with a change rate of 0.14 meters per second, which is greater than the threshold of 0.1 meters per second. The system confirms this as persistent attachment and executes a secondary obstacle avoidance strategy, i.e., stopping forward and shifting 0.3 meters to the left (the unobstructed side) before continuing forward. In the case of localized small attachments, both laser and infrared sensors jointly identify them as surface attachments. Ultrasonic data shows that the rate of change of the forward distance value is 0.15 meters per second, which is greater than the threshold of 0.1 meters per second, but the rate of change of the lateral distance value is only 0.03 meters per second, which is less than the threshold of 0.1 meters per second. The system confirms that it is a localized small attachment and executes a first-level wall-hugging detour strategy, which means decelerating to 0.1 meters per second and bypassing it in an arc path.

[0065] In a sensor conflict scenario, the laser sensor detects an obstacle 0.20 meters ahead and identifies it as an obstacle, while the infrared sensor detects a gradual change in the contact distance from 0.03 meters to 0.05 meters and identifies it as the curvature of the pier itself. The two judgments conflict. At this time, the ultrasonic sensor detects an echo signal for 10 consecutive sampling cycles in front, i.e., continuous echo. The system uses the ultrasonic data to confirm the existence of the obstacle and performs obstacle avoidance.

[0066] This implementation achieves significant technical benefits in several aspects through multi-directional detection using ultrasonic sensors and a sensor conflict resolution mechanism. By comprehensively analyzing the rate of change of ultrasonic waves from the front and sides, it enables precise differentiation of the size of surface attachments. Stubborn attachments are handled with higher-level obstacle avoidance strategies, while small, locally located attachments are handled with lower-level wall-hugging and bypassing strategies, improving the targeting of obstacle avoidance strategies. By establishing a resolution rule based on ultrasonic waves in sensor conflict situations, it effectively solves the decision-making dilemma when different sensors provide inconsistent results, ensuring the certainty of obstacle avoidance decisions in complex environments. It fully utilizes the characteristic of ultrasonic sensors being unaffected by light and color, providing reliable third-party verification in cases of conflict between laser and infrared sensors, improving the system's adaptability to complex lighting and surface conditions.

[0067] In some specific embodiments, the cross-verification based on the preliminary judgment result includes: first, laser-infrared verification; if the laser ranging is less than a preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed; otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed; then, laser-ultrasonic verification is performed; if both detect obstacles and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is performed; otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance instructions by combining the above verification and validation results, the method further includes: During the execution of the wall-hugging bypass strategy, it is determined whether the real-time contact distance between the robot and the surface of the pier is within the preset contact range; When the real-time bonding distance exceeds the preset bonding range, the robot's motion trajectory is adjusted. Determine whether the robot has bypassed the attachments on the surface of the pier; If it fails to bypass the obstacle, it will continue to bypass the obstacle along an arc path and continuously monitor the contact gap of the infrared sensor. Once the detour has been completed, the wall-hugging detour strategy is discontinued, and the original driving path is restored. In the attitude constraint verification, it is determined whether the pitch angle is greater than a preset pitch angle threshold. When the pitch angle is greater than the preset pitch angle threshold, it is determined to be a slope obstacle and the speed is adjusted to low-speed climbing. Determine if the roll angle is greater than the preset roll angle threshold; When the roll angle is greater than the preset roll angle threshold, it is determined to be a roll risk and a leveling action is performed first. Determine whether the yaw angle deviation is greater than the preset yaw deviation threshold; When the yaw angle deviation is greater than the preset yaw deviation threshold, it is determined to be a path deviation and the course is corrected before obstacle avoidance is performed.

[0068] It should be understood that this embodiment establishes a dynamic control and attitude constraint verification mechanism for wall-hugging maneuvering. Regarding the control of the contact distance during wall-hugging maneuvering, when executing the wall-hugging maneuvering strategy, the robot moves along the surface of the pier in an arc path. The system continuously monitors the real-time contact distance detected by the infrared sensor and determines whether it is within a preset contact range, such as 0.02 meters to 0.06 meters. If the real-time contact distance exceeds this range, such as being less than 0.02 meters, it means the robot is too close to the pier surface and there is a risk of collision; if it is greater than 0.06 meters, it means the robot has deviated too far from the pier surface and may lose its grip. In this case, the robot's trajectory is adjusted, for example, by adjusting the turning angle or speed, to restore the contact distance to the preset range.

[0069] Regarding bypass judgment and strategy exit, the system determines whether the robot has bypassed the surface attachments of the pier through comprehensive judgment. The judgment criteria may include the lidar ranging recovering to a level greater than the safety threshold, the infrared contact gap returning to a stable level, and the ultrasonic sensor no longer detecting echoes. If the obstacle has not been bypassed, the robot continues to travel along an arc path and continuously monitors the contact gap of the infrared sensor. If the obstacle has been bypassed, the robot exits the wall-hugging bypass strategy and resumes the original travel path. Regarding attitude constraint verification, attitude constraint verification includes three independent judgment branches: pitch angle verification determines whether the pitch angle is greater than a preset pitch angle threshold, such as 15°. If it is greater, it is judged as a slope obstacle and the robot is adjusted to a low-speed climbing mode, such as reducing the speed to 0.05 meters per second; roll angle verification determines whether the roll angle is greater than a preset roll angle threshold, such as 10°. If it is greater, it is judged as a roll risk and leveling actions are performed first, such as adjusting the robot's support legs or center of gravity position; yaw angle deviation verification determines whether the yaw angle deviation is greater than a preset yaw deviation threshold, such as 5°. If it is greater, it is judged as a path deviation and the robot first corrects its course before performing obstacle avoidance.

[0070] In one specific embodiment, in the wall-hugging detour scenario, the robot executes the wall-hugging detour using a primary strategy. The radius of the arc path is 0.3 meters. Infrared sensors monitor the contact gap in real time, with a preset contact range of 0.02 meters to 0.06 meters. During the detour, the contact gap gradually increases to 0.07 meters, exceeding the range. The system adjusts the motion trajectory by increasing the turning angle to bring the robot closer to the pier surface. After the contact gap returns to 0.04 meters, the robot continues to detour along the arc path. After 3 seconds, the lidar measures that the distance in front has returned to 0.7 meters, which is greater than the safety threshold of 0.5 meters. The system determines that it has bypassed the attachment and exits the wall-hugging detour strategy to resume the original straight path. In attitude verification scenarios, when the robot's inertial measurement unit detects a pitch angle of 18°, which is greater than the threshold of 15°, during climbing, the system determines it to be a slope obstacle and reduces its speed from 0.2 meters per second to a low-speed climbing mode of 0.05 meters per second to ensure sufficient adhesion and stability on the slope. Once the pitch angle recovers to 12°, the speed returns to 0.2 meters per second. If the inertial measurement unit detects a roll angle of 12°, which is greater than the threshold of 10°, the system prioritizes leveling, i.e., adjusting the support forces on both sides to restore the roll angle to within 10°, before performing obstacle avoidance. If the yaw angle deviation reaches 7°, which is greater than the threshold of 5°, the system first corrects the heading, i.e., adjusts the steering to bring the yaw angle deviation to zero, before performing obstacle avoidance.

[0071] This implementation achieves several significant technical benefits through dynamic control and attitude constraint verification of wall-hugging maneuvering. Real-time monitoring of the contact distance and adjustment of the motion trajectory ensures the robot maintains an appropriate distance from the pier surface during wall-hugging maneuvering, avoiding collision risks and preventing loss of grip due to detachment from the pier surface. An bypass determination mechanism enables automatic exit from the wall-hugging maneuvering strategy and path recovery, avoiding unnecessary detour time and improving operational efficiency. Three-dimensional attitude constraint verification using pitch, roll, and yaw angles enables differentiated handling of different types of attitude anomalies: low-speed climbing for slope obstacles, leveling actions for roll risks, and heading corrections for path deviations, ensuring the robot takes the most appropriate measures under various attitude anomaly conditions. The organic combination of attitude constraint verification and obstacle avoidance decision-making ensures the robot remains within a controllable attitude range when performing obstacle avoidance actions, avoiding the risk of attitude loss of control caused by the obstacle avoidance actions themselves.

[0072] In some specific embodiments, the process of filtering and denoising the collected sensor data, and sequentially determining the validity of each sensor data point, marking sensor data exceeding its effective range as invalid data, and determining sensor malfunction and emergency braking when all sensor data is invalid, further includes determining valid data when at least two sensor data points are valid. When a sensor is marked as invalid data, determine whether the data of that sensor recovers to normal within a subsequent preset number of sampling periods; When the sensor data returns to normal, it is reintroduced into the fusion judgment. If the sensor data is not recovered, determine whether the duration of the sensor failure exceeds a preset time window; If the failure duration does not exceed the preset time window, the failure flag of the sensor is maintained and monitoring continues. If the duration of the failure exceeds the preset time window, the sensor is confirmed to have failed.

[0073] It should be understood that this implementation establishes a three-stage management mechanism for sensor dynamic monitoring, recovery confirmation, and failure confirmation. During the recovery monitoring phase, when a sensor is marked as having invalid data, the system enters the recovery monitoring phase, continues to collect data from that sensor, and determines whether the data has returned to normal within a preset number of sampling periods. The preset number can be set to 10 sampling periods, which corresponds to 0.2 seconds if the sampling frequency is 50 Hz. The criterion for determining that the data has returned to normal is that the sensor continuously outputs data within its effective measurement range.

[0074] During the failure duration monitoring phase, if the data fails to recover to normal within a preset number of sampling periods, the system enters the failure duration monitoring phase to determine whether the failure duration of the sensor exceeds a preset time window, such as 5 seconds. If it does not exceed the window, the failure flag of the sensor is maintained and its data status continues to be monitored. During the failure confirmation phase, if the failure duration exceeds the preset time window, such as failing to recover to normal for more than 5 consecutive seconds, the system confirms that the sensor has failed. After confirming the failure, the sensor is removed from the list of valid sensors, enters a degraded operation state, and a failure log is recorded and a fault alarm is issued.

[0075] In one specific embodiment, the sampling frequency is 50 Hz, i.e., each sampling period is 0.02 seconds, the preset number is 10 sampling periods (0.2 seconds), and the preset time window is 5 seconds. In the case of recovery after occasional interference, if the ultrasonic sensor outputs data exceeding its effective range at t=0 (e.g., 8 meters exceeding the range by 0.02 meters to 5 meters), it is marked as invalid. Within the subsequent 5 sampling periods (0.1 seconds), the data recovers to normal, outputting 0.35 meters. Since the recovery occurs within the preset number of 10 sampling periods, the system re-includes the ultrasonic sensor in the fusion judgment. In the case of recovery after continuous interference, if the infrared sensor continuously outputs invalid data due to strong light interference, the data recovers to normal after 15 sampling periods (0.3 seconds), still within the preset number of 10 sampling periods. The system re-includes it in the fusion judgment, although the recovery time is slightly longer, it is still within the acceptable range.

[0076] In a sensor failure scenario, the lidar continuously outputs invalid data due to hardware malfunction. If the data does not recover after 10 sampling cycles (0.2 seconds), the system starts timing the failure duration. If the data still does not recover after 5 seconds (250 sampling cycles), the failure duration exceeds the preset time window of 5 seconds. The system confirms the lidar failure, removes it from the list of valid sensors, and enters a degraded operation state. At this time, the valid sensors are the ultrasonic sensor, infrared sensor, and inertial measurement unit, totaling 3. The system also sends a fault alarm signal for lidar failure via wireless communication.

[0077] This implementation achieves significant technical benefits in several aspects by establishing a three-stage sensor management mechanism. The recovery monitoring stage, i.e., setting a preset sampling period, allows the sensor to automatically recover after experiencing brief, intermittent interference, preventing premature confirmation of sensor failure due to transient interference. The failure duration monitoring stage, i.e., setting a preset time window, provides sufficient recovery time for the sensor while preventing indefinite waiting, ensuring timely confirmation of failure when the sensor truly cannot recover. Degraded operation and fault alarms after failure confirmation ensure that the system can continue operating with reduced sensing capabilities after sensor failure, while simultaneously notifying operators for inspection and repair. The entire management process is fully automated, requiring no manual intervention, thus improving the system's autonomy and maintainability.

[0078] A second embodiment of the multi-sensor fusion obstacle avoidance system for a climbing robot according to an embodiment of this application includes: The data acquisition module is used to synchronously acquire environmental perception data during the operation of the robot through multiple sensors installed on the climbing robot; The discrimination module is used to filter and denoise the collected sensor data, and sequentially judge the validity of each sensor data. Sensor data that exceeds its own effective range is marked as invalid data. When all sensor data are invalid, it is determined that the sensor is faulty and an emergency braking is performed. When at least two sensor data are valid, it is determined to be valid data. The determination module is used to compare the valid data with a preset distance threshold, preliminarily determine whether there is an obstacle in front of the robot, and preliminarily classify the type of obstacle according to the location and distance range of the obstacle to obtain a preliminary determination result; The decision module is used to perform cross-validation based on the preliminary judgment results. First, laser-infrared verification is performed. If the laser ranging is less than the preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed. Otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed. Then, laser-ultrasonic verification is performed. If both detect the obstacle and the ranging difference is within the preset error range, the obstacle is confirmed and obstacle avoidance is performed. Otherwise, the obstacle level is reduced or obstacle avoidance is temporarily not performed. At the same time, the corresponding attitude adjustment is performed based on the attitude angle verification results of the inertial measurement unit. The obstacle danger level is determined by combining the above verification and validation results and obstacle avoidance instructions are generated.

[0079] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A multi-sensor fusion method for obstacle avoidance in a climbing robot, characterized in that, Includes the following steps: Environmental perception data during the robot's operation is collected synchronously by multiple sensors installed on the climbing robot. The collected sensor data is filtered and denoised, and the validity of each sensor data is judged in turn. Sensor data that exceeds its own effective range is marked as invalid data. When all sensor data is invalid, it is determined that the sensor is faulty and an emergency braking is performed. When at least two sensor data are valid, it is determined to be valid data. The valid data is compared with a preset distance threshold to preliminarily determine whether there is an obstacle in front of the robot, and the type of obstacle is preliminarily classified according to the location and distance range of the obstacle to obtain a preliminary judgment result; Based on the preliminary judgment results, cross-validation is performed. First, laser-infrared verification is conducted. If the laser ranging is less than the preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed. Otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed. Then, laser-ultrasonic verification is performed. If both detect the obstacle and the ranging difference is within the preset error range, the obstacle is confirmed and obstacle avoidance is performed. Otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. At the same time, the corresponding attitude adjustment is performed based on the attitude angle verification results of the inertial measurement unit. The obstacle danger level is determined by combining the above verification and validation results, and an obstacle avoidance command is generated.

2. The obstacle avoidance method for a climbing robot based on multi-sensor fusion according to claim 1, characterized in that, When the environmental perception data during the operation of the climbing robot is collected synchronously by multiple sensors installed on the robot, it includes: The various sensors include at least lidar, ultrasonic sensors, infrared sensors, and inertial measurement units; The lidar is used to acquire the distance and contour point cloud of obstacles in front, the ultrasonic sensor is used to detect the distance of obstacles at medium and short distances, the infrared sensor is used to detect the contact distance between the robot chassis and the pier surface, and the inertial measurement unit is used to detect the attitude angle of the robot body.

3. The obstacle avoidance method for a climbing robot based on multi-sensor fusion according to claim 2, characterized in that, The process involves filtering and denoising the collected sensor data, and sequentially determining the validity of each sensor data point. Sensor data exceeding its effective range is marked as invalid. When all sensor data is invalid, a sensor malfunction is identified and emergency braking is initiated. When at least two sensor data points are valid, the data is considered valid, including: Determine whether the number of effective sensors for the target is greater than or equal to the preset minimum sensor count threshold; When the number of effective sensors is less than the preset minimum sensor number threshold, the system is determined to be in a failure state, the robot brakes urgently and sends out a fault alarm signal. When the number of effective target sensors is greater than or equal to the preset minimum sensor number threshold, the system is determined to be in a degraded operation state, and the fusion judgment is made based on the current effective sensor data.

4. The obstacle avoidance method for a climbing robot based on multi-sensor fusion according to claim 3, characterized in that, The step of comparing the valid data with a preset distance threshold to preliminarily determine whether there is an obstacle in front of the robot, and preliminarily classifying the type of obstacle according to its location and distance range, to obtain a preliminary determination result, includes: Determine the relationship between the distance in front of the robot and a preset safety threshold; When the distance in front of the robot is greater than the preset safety threshold, it is determined that the path ahead is clear, and the robot continues to climb. When the forward distance value is less than or equal to the preset safety threshold, the relationship between the forward distance value and the preset near distance threshold is determined, and the relationship between the point cloud density and the preset density threshold is also determined. When the forward distance value is greater than the preset near distance threshold and the point cloud density is greater than the preset density threshold, it is initially determined to be a protruding obstacle; When the forward distance value is less than the preset close distance threshold and the rate of change of the fitting spacing is within the preset rate of change threshold, it is initially determined that the curvature of the pier itself is changing. Determine whether the change in lateral distance value within a preset sampling period is consistent with the change in forward distance value; When the change in the lateral distance value exceeds a preset change threshold and is inconsistent with the change in the forward distance value, it is initially determined to be a lateral obstacle.

5. The obstacle avoidance method for a climbing robot based on multi-sensor fusion according to claim 4, characterized in that, The process of cross-verifying based on the preliminary judgment results involves first performing laser-infrared verification. If the laser ranging is less than a preset safety threshold and the infrared contact gap changes abruptly, it is determined to be a surface attachment and obstacle avoidance is executed. Otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed. Then, laser-ultrasonic verification is performed. If both methods detect an obstacle and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is executed. Otherwise, the obstacle level is lowered or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance commands by combining the above verification and validation results, the process includes: When laser-infrared verification determines that the pier itself is structurally sound and laser-ultrasonic verification does not identify any obstacle, it is classified as level zero. When laser-infrared verification determines that it is a surface attachment and laser-ultrasonic verification confirms an obstacle, but attitude verification does not require adjustment, it is classified as Level 1. When laser-ultrasonic verification confirms an obstacle and attitude verification results trigger attitude adjustment, it is classified as Level 2; When laser-ultrasonic verification confirms an obstacle and the attitude verification results show that the attitude angle exceeds the preset threshold by a continuously increasing amount, it is determined to be level three; The obstacle avoidance control commands are executed according to the determined danger level. Level 0 means ignoring the obstacle and maintaining the original speed. Level 1 means decelerating to a preset percentage of the original speed and implementing a wall-hugging bypass strategy. Level 2 means stopping and shifting a preset distance to the obstacle-free side before continuing to move forward. Level 3 means emergency braking, reporting obstacle information, and waiting for manual intervention.

6. The obstacle avoidance method for a climbing robot based on multi-sensor fusion according to claim 5, characterized in that, The process of cross-validating based on the preliminary judgment results includes: first, laser-infrared verification; if the laser ranging is less than a preset safety threshold and the infrared contact gap changes abruptly, it is determined to be a surface attachment and obstacle avoidance is executed; otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed; then, laser-ultrasonic verification is performed; if both detect obstacles and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is executed; otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance commands by combining the above verification and validation results, the process also includes: Determine whether obstacle features have disappeared in the re-acquired sensor data; When the obstacle features disappear, the obstacle avoidance is considered successful, and the original speed is resumed to continue climbing; If the obstacle features still exist, then determine the relationship between the current number of attempts and the preset upper limit of the number of attempts; If the current number of attempts is less than the preset maximum number of attempts, then the obstacle avoidance strategy level is upgraded and the attempt is repeated. When the current number of attempts is greater than or equal to the preset maximum number of attempts, it is determined that the attempt cannot be completed independently and the system is reported and awaits manual instructions.

7. The obstacle avoidance method for a climbing robot based on multi-sensor fusion according to claim 6, characterized in that, The process of cross-validating based on the preliminary judgment results includes: first, laser-infrared verification; if the laser ranging is less than a preset safety threshold and the infrared contact gap changes abruptly, it is determined to be a surface attachment and obstacle avoidance is executed; otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed; then, laser-ultrasonic verification is performed; if both detect obstacles and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is executed; otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance commands by combining the above verification and validation results, the process also includes: Determine whether the ultrasonic sensor detects a change rate of distance value exceeding a preset change rate threshold in both the front and side views; When both the laser sensor and the infrared sensor determine that it is a surface attachment, and the ultrasonic sensor detects a change rate of distance value exceeding the preset change rate threshold in both the front and the side, it is confirmed as a stubborn attachment and a secondary obstacle avoidance strategy is executed. When both the laser sensor and the infrared sensor determine that it is a surface attachment, but the ultrasonic sensor only detects that the rate of change of the distance value in front exceeds the preset rate of change threshold while it does not exceed the threshold in the side, it is confirmed as a local small attachment and a first-level wall-hugging bypass strategy is executed. Determine whether the ultrasonic sensor detects a continuous echo in front; When the laser sensor determines that it is an obstacle, the infrared sensor determines that it is the curvature of the pier itself, and the ultrasonic sensor detects a continuous echo in front, obstacle avoidance is performed based on the data from the ultrasonic sensor.

8. The obstacle avoidance method for a climbing robot based on multi-sensor fusion according to claim 7, characterized in that, The process of cross-validating based on the preliminary judgment results includes: first, laser-infrared verification; if the laser ranging is less than a preset safety threshold and the infrared contact gap changes abruptly, it is determined to be a surface attachment and obstacle avoidance is executed; otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed; then, laser-ultrasonic verification is performed; if both detect obstacles and the ranging difference is within a preset error range, the obstacle is confirmed and obstacle avoidance is executed; otherwise, the obstacle level is reduced or obstacle avoidance is temporarily suspended. Simultaneously, corresponding attitude adjustments are performed based on the attitude angle verification results of the inertial measurement unit. When determining the obstacle hazard level and generating obstacle avoidance commands by combining the above verification and validation results, the process also includes: During the execution of the wall-hugging bypass strategy, it is determined whether the real-time contact distance between the robot and the surface of the pier is within the preset contact range; When the real-time bonding distance exceeds the preset bonding range, the robot's motion trajectory is adjusted. Determine whether the robot has bypassed the attachments on the surface of the pier; If it fails to bypass the obstacle, it will continue to bypass the obstacle along an arc path and continuously monitor the contact gap of the infrared sensor. Once the detour has been completed, the wall-hugging detour strategy is discontinued, and the original driving path is restored. In the attitude constraint verification, it is determined whether the pitch angle is greater than a preset pitch angle threshold. When the pitch angle is greater than the preset pitch angle threshold, it is determined to be a slope obstacle and the speed is adjusted to low-speed climbing. Determine if the roll angle is greater than the preset roll angle threshold; When the roll angle is greater than the preset roll angle threshold, it is determined to be a roll risk and a leveling action is performed first. Determine whether the yaw angle deviation is greater than the preset yaw deviation threshold; When the yaw angle deviation is greater than the preset yaw deviation threshold, it is determined to be a path deviation and the course is corrected before obstacle avoidance is performed.

9. The obstacle avoidance method for a climbing robot based on multi-sensor fusion according to claim 8, characterized in that, The process involves filtering and denoising the collected sensor data, and sequentially determining the validity of each sensor data point. Sensor data exceeding its effective range is marked as invalid. When all sensor data is invalid, a sensor malfunction is identified and emergency braking is initiated. Determining data as valid when at least two sensor data points are valid further includes: When a sensor is marked as invalid data, determine whether the data of that sensor recovers to normal within a subsequent preset number of sampling periods; When the sensor data returns to normal, it is reintroduced into the fusion judgment. If the sensor data is not recovered, determine whether the duration of the sensor failure exceeds a preset time window; If the failure duration does not exceed the preset time window, the failure flag of the sensor is maintained and monitoring continues. If the duration of the failure exceeds the preset time window, the sensor is confirmed to have failed.

10. A multi-sensor fusion obstacle avoidance system for a climbing robot, characterized in that, An obstacle avoidance method for a climbing robot based on multi-sensor fusion, applicable to any one of claims 1 to 9, includes: The data acquisition module is used to synchronously acquire environmental perception data during the operation of the robot through multiple sensors installed on the climbing robot; The discrimination module is used to filter and denoise the collected sensor data, and sequentially judge the validity of each sensor data. Sensor data that exceeds its own effective range is marked as invalid data. When all sensor data are invalid, it is determined that the sensor is faulty and an emergency braking is performed. When at least two sensor data are valid, it is determined to be valid data. The determination module is used to compare the valid data with a preset distance threshold, preliminarily determine whether there is an obstacle in front of the robot, and preliminarily classify the type of obstacle according to the location and distance range of the obstacle to obtain a preliminary determination result; The decision module is used to perform cross-validation based on the preliminary judgment results. First, laser-infrared verification is performed. If the laser ranging is less than the preset safety threshold and the infrared contact spacing changes abruptly, it is determined to be a surface attachment and obstacle avoidance is performed. Otherwise, it is determined to be the pier's own structure and obstacle avoidance is not performed. Then, laser-ultrasonic verification is performed. If both detect the obstacle and the ranging difference is within the preset error range, the obstacle is confirmed and obstacle avoidance is performed. Otherwise, the obstacle level is reduced or obstacle avoidance is temporarily not performed. At the same time, the corresponding attitude adjustment is performed based on the attitude angle verification results of the inertial measurement unit. The obstacle danger level is determined by combining the above verification and validation results and obstacle avoidance instructions are generated.

Images