An adaptive path tracking method based on fusion of inertial sensor-wheel speed encoder information

Abstract

This invention discloses an adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion, relating to the field of wall-climbing robot technology. The method includes: establishing a cylindrical coordinate model of the surface to be worked on; transforming surface path planning into a combined control problem of tangential angle and axial displacement; and generating a bow-shaped reciprocating coverage path. This invention approximates the surface to be worked on as a cylindrical geometric model, establishes a combined control strategy of axial displacement and tangential angle in cylindrical coordinates, and achieves accurate transformation of the surface full-coverage path planning problem. By employing a bow-shaped reciprocating coverage strategy, combined with real-time correction of encoder calibration coefficients, it effectively eliminates lateral translation deviations caused by tire friction variations or surface unevenness, ensures uniform spacing between adjacent paths, avoids path omissions or overlaps, improves work coverage and consistency, and meets the requirements of high-precision automated operations.

Description

Technical Field

[0001] This invention relates to the field of wall-climbing robot technology, specifically to an adaptive path tracking method based on the fusion of information from inertial sensors and wheel speed encoders. Background Technology

[0002] With the improvement of industrial automation, curved surface climbing robots are increasingly used in the maintenance of large pressure equipment and the finishing of tunnel construction. These robots need to complete automated operations on cylindrical or near-cylindrical curved surfaces in complex environments without fixed reference points. The core requirements are to achieve accurate positioning, stable path planning, and high coverage. In existing technologies, the positioning and path planning of planar climbing robots are relatively mature, often relying on laser SLAM, visual VSLAM, or GNSS+IMU fusion positioning technologies.

[0003] Existing technologies have shortcomings in curved surface operations: 1. High dependence on environment: Existing technologies mostly rely on lidar, visual sensors or GPS, which are prone to failure in dusty, low light or curved surface scenes without fixed reference objects, resulting in a significant decrease in positioning accuracy; 2. Insufficient scene adaptability: Traditional path planning methods are difficult to adapt to the curvature changes of cylinders or quasi-cylindrical surfaces, lack the ability to adapt to non-standard sized work surfaces, and are prone to path omissions or uneven coverage. 3. Poor operational stability: In curved surface operations, robots are prone to posture deviation and slippage due to curvature changes. Existing technologies lack effective real-time detection and correction mechanisms, which can easily lead to path deviation and operation failure. 4. High system cost: Solutions that rely on equipment such as LiDAR and high-definition vision cameras have high hardware costs and complex maintenance, making them difficult to apply on a large scale. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention is implemented through the following technical solution: an adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion, applicable to cylindrical or near-cylindrical curved surface operations such as tunnel lining trolley grinding and cylindrical tank inspection, including the following steps: Step 1: Establish a cylindrical coordinate model of the surface to be worked on, transform the surface path planning into a combined control problem of tangential angle and axial displacement, and generate a bow-shaped round-trip covering path to achieve automatic generation of full-coverage path for complex surfaces. Step 2: Based on the inertial sensors and wheel speed encoders mounted on the robot body, a fixed sampling frequency is set to collect robot posture, angular velocity and wheel speed data in real time, so as to achieve real-time accurate acquisition and time alignment of posture and displacement data; Step 3: Use inertial sensors to obtain pitch angle, calculate tangential position, measure travel distance through wheel speed encoder, estimate axial displacement, perform precise surface positioning, and fuse pitch angle and encoder data to achieve surface dead reckoning positioning without external reference. Step 4: Generate a parallel arc path based on the width of the work tool, and cover the work area using a bow-shaped strategy. Automatically stop moving based on the pitch angle change threshold and time conditions, control the robot to translate along the axis to the starting point of the adjacent path and switch directions, and automatically determine the path endpoints based on the pitch angle to achieve autonomous lane changing and continuous coverage of the bow-shaped path. Step 5: Monitor changes in roll and pitch angles during travel, detect attitude deviation or slippage, and perform correction, retry, or manual takeover operations to ensure path tracking stability. Step 6: The infrared sensor detects the surface boundary and triggers deceleration, stop or path adjustment control when it approaches the edge of the work area. After the task is completed, the trajectory map is output and prompts for grinding depth verification. The detection of the surface edge automatically triggers safety control and outputs the trajectory map to realize the visual traceability of work quality.

[0005] Preferably, step 1 specifically includes: The surface of the tunnel lining trolley or cylindrical tank to be operated is approximated as a cylindrical geometric model, and a cylindrical coordinate system is established with the surface axis as the horizontal axis and the tangential angle as the vertical axis. The path planning problem of full surface coverage is transformed into a combined control problem of axial displacement and tangential angle under cylindrical coordinates, realizing the equivalent transformation of complex surface motion to a planar coordinate system and reducing the solution complexity of path planning and control. Based on the preset grinding width d of the work tool as the spacing parameter between adjacent paths, the surface to be worked is divided into multiple parallel arc paths in the axial dimension of the cylindrical coordinate system, so that the spacing between each path in the axial projection is equal to or greater than the grinding width d, ensuring that the work area is evenly divided, and providing an accurate path topology for subsequent full-coverage path execution. By adopting a bow-shaped reciprocating coverage strategy, the robot is controlled to move back and forth along each arc path in sequence, so that the movement direction between adjacent paths is opposite. After completing the current path, it automatically switches to the next adjacent path until all the divided paths are covered in sequence, so as to achieve efficient and continuous coverage of the work area, avoid path omissions and reduce idle travel time.

[0006] Preferably, step 1 further includes: During the generation of the bow-shaped path, the stable tilt angle sensor data was recorded 30 seconds after the robot initially adhered to the working surface and remained stationary. The initial value of the X-axis pitch angle was denoted as... The initial value of the Y-axis roll angle is denoted as And the maximum pitch angle of the X-axis at the starting point of the trolley is recorded as This establishes a reliable baseline threshold for subsequent attitude determination and path endpoint identification; The entry direction of the initial grinding path is determined based on the left or right entry mode selected by the operator. The starting point and initial direction of the first arc path are determined based on the entry direction to ensure that the robot's initial movement trend matches the layout of the work surface and improve the adaptability of path planning. By conducting multiple translation experiments on the working surface, the encoder correction coefficient λ is calibrated to correct the deviation between the robot's lateral translation distance and the theoretical value (grinding width d) caused by tire friction or surface unevenness, thereby eliminating lateral translation error and ensuring axial positioning accuracy during path switching.

[0007] Preferably, step 2 specifically includes: A tilt sensor and a wheel speed encoder are installed on the robot body. The tilt sensor is used to collect the robot's X-axis pitch angle in real time. and Y-axis roll angle Data, wheel speed encoders are used to collect real-time rotational speed or rotational angle pulse data of robot drive wheels, providing raw data support for surface positioning and attitude calculation; A fixed data sampling frequency is set so that the tilt sensor and wheel speed encoder can collect data synchronously under the same time reference, so that the attitude data and displacement data are time-aligned and the time consistency of attitude and displacement data is guaranteed during dead reckoning. The host computer records and stores the collected three-axis attitude angle data, angular velocity data, and left and right wheel pulse counts in real time, providing a data foundation for path tracking accuracy analysis and anomaly tracing.

[0008] Preferably, step 3 specifically includes: Pitch angle acquired in real time using tilt sensor The data, combined with the geometric relationships of the curved surface, is used to calculate the robot's current position in the tangential direction. When the value is large, the robot is determined to be in a lower position on the arc surface. As the height gradually decreases, the robot is determined to be rising along the curved surface, enabling the robot to autonomously perceive the height and position of the curved surface in real time, providing accurate state criteria for path tracking; The robot's travel distance is calculated by collecting the number of drive wheel pulses from the wheel speed encoder, and the axial displacement is estimated by combining the encoder correction coefficient λ. This gives the robot's current position in the axial direction of the curved surface, effectively eliminating displacement deviations caused by friction and unevenness, and ensuring the accuracy and consistency of axial positioning. By fusing the estimated tangential angle position and axial displacement within a cylindrical coordinate system, a two-dimensional position coordinate system for the robot on the curved surface is formed. This enables dead reckoning and positioning on the curved surface without the need for an external positioning reference, thus constructing a complete curved surface position coordinate system and achieving autonomous positioning and path tracking without an external reference.

[0009] Preferably, step 4 specifically includes: During the grinding process as the robot moves along the curved path, the pitch angle acquired by the tilt sensor is monitored in real time. Change, when When the angle is less than 5° and the state is maintained continuously for more than or equal to 1 second, it is determined that the robot has reached the highest point of the current arc path, ensuring that the robot accurately triggers path switching at the vertex position of the curved surface, avoiding overshoot or premature lane change. Once the highest point is reached, the robot immediately stops its current forward movement and, based on the encoder correction coefficient λ, controls the robot to translate along the axial direction by a distance d, i.e., the grinding width, from the highest point of the current path to the starting point of the adjacent path. This ensures that the robot accurately reaches the starting point of the next path and eliminates the cumulative error of lateral displacement. After the translation is completed, the robot switches to the backward grinding state and moves in the opposite direction along the adjacent arc path to continue the grinding operation, achieving continuous back-and-forth coverage of the bow-shaped path, improving work efficiency and path integrity.

[0010] Preferably, step 4 further includes: During the robot's backward grinding process, the pitch angle is monitored in real time. Change, when Greater than or equal to And the current state is maintained continuously for a time greater than or equal to 1 second, or Greater than or equal to Furthermore, when the state is maintained continuously for more than or equal to 1 second, or when the trigger signal of the bottom infrared sensor is detected, it is determined that the robot has reached the lowest point of the current arc path, ensuring the reliability of the lowest point identification and multi-source redundancy, and avoiding path switching errors. Once the lowest point is reached, the robot immediately stops its current backward movement and, based on the encoder correction coefficient λ, controls the robot to translate a distance d along the axis from the lowest point of the current path to the starting point of the next adjacent path, ensuring accurate path switching and maintaining consistency in the spacing between adjacent paths. After the translation is completed, the robot switches to forward grinding mode and moves forward along the next adjacent arc path. This cycle repeats to achieve continuous coverage of the bow-shaped path, realizing full coverage and automated looping of the work area, improving work efficiency and path integrity.

[0011] Preferably, step 5 specifically includes: Real-time monitoring of roll angle during robot movement With initial value The deviation, when If the robot is found to have a posture deviation, it should immediately stop its current movement to avoid uneven polishing or equipment instability caused by abnormal posture. After stopping, the control system performs a micro-rotation correction operation in place, causing the robot to rotate around the vertical axis until the roll angle is reached. Return to the allowed range This ensures that the robot regains a stable working posture. After the roll angle correction is completed, the control system restarts the travel motion and continues the interrupted grinding operation to ensure the continuity of the operation and the accuracy of path tracking.

[0012] Preferably, step 5 further includes: Real-time monitoring of pitch angle during robot movement The rate of change, when detected If the change is less than 1° within 10 seconds, the robot is judged to have slipped based on its current motion state, and its current movement is stopped immediately to effectively prevent path deviation caused by slipping and ensure accurate operation. Once the slippage detection is established, the robot is controlled to retreat to the bottom of the path. The criterion for retreating to the correct position is... Greater than or equal to Furthermore, if this state is maintained continuously for a period of 1 second or more, or if the trigger signal from the bottom infrared sensor is detected, the robot will be able to re-establish a stable starting reference for the operation and eliminate pose uncertainty. After retracing to the bottom of the path, perform in-situ micro-rotation correction to bring the roll angle back to the allowable range, and start grinding forward again. If slippage occurs repeatedly and cannot be eliminated by automatic correction, the system enters protection mode and prompts the operator to intervene manually. This provides safety redundancy for anomalies that cannot be automatically recovered, ensuring that the equipment and operation process are safe and controllable.

[0013] Preferably, step 6 specifically includes: During the process of the robot performing the bow-shaped path coverage, the curved surface edge is detected in real time by infrared sensors set on the side of the robot. When the infrared sensors detect the edge signal, it is determined that the robot has entered the last path, realizing autonomous edge recognition and ensuring the accuracy and reliability of the triggering in the closing stage. Once the determination of entering the finishing mode is established, the current movement is immediately stopped, and the robot is controlled to move laterally by 0.1m in the opposite direction of the current movement, so that the robot can get away from the edge danger zone, effectively preventing the robot from crossing the boundary and falling, and ensuring the safety of equipment and operation; After translating to a safe area, continue grinding the last arc. Once the task is completed, stop driving, upload the actual motion trajectory and the ideal trajectory curve to the host computer, and prompt for grinding depth verification in areas with excessive trajectory deviation. This enables visual traceability of the operation process and provides data support for quality control and path optimization.

[0014] This invention provides an adaptive path tracking method based on the fusion of information from inertial sensors and wheel speed encoders. It has the following beneficial effects: (I) This adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion, by approximating the surface to be worked as a cylindrical geometric model, establishes a combined control strategy of axial displacement and tangential angle in cylindrical coordinate system, realizes the accurate transformation of the full-coverage path planning problem of the surface, adopts a bow-shaped reciprocating coverage strategy, combined with real-time correction of encoder correction coefficient, effectively eliminates the lateral translation deviation caused by tire friction changes or surface unevenness, ensures uniform spacing between adjacent paths, avoids path omission or overlap, improves work coverage and consistency, and meets the requirements of high-precision automated operation.

[0015] (II) This adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion monitors the changes in roll angle and pitch angle in real time during robot movement. It determines the attitude deviation or slippage state through preset thresholds, interrupts movement in time, and performs in-situ micro-rotation correction or backtracking retry operation. It can automatically identify and eliminate attitude abnormalities caused by factors such as curvature changes and uneven friction, restore the robot to a stable working state, and avoid path deviation and operation failure. For abnormal situations where multiple corrections fail, the system enters a protection state and prompts manual intervention to ensure equipment safety and operation continuity.

[0016] (III) This adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion uses infrared sensors set on the side of the robot to detect the curved surface boundary in real time. When entering the last path, the robot automatically triggers the end-of-path mode, immediately stops the current motion and controls the robot to move laterally to a safe area to avoid the risk of falling off the boundary. After completing the last path operation, the system automatically stops driving and uploads the trajectory data to the host computer, realizing orderly control from edge detection, safe retreat to complete path execution, and ensuring the safe and efficient operation of the robot in the curved surface edge area. Attached Figure Description

[0017] Figure 1 This is a flowchart of the robot localization and autonomous path planning scheme of the present invention; Figure 2 This is a diagram showing the curved "bow-shaped" path planning of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1, please refer to Figure 1 , Figure 2 This invention provides a technical solution: an adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion, comprising the following steps: Step 1: Establish a cylindrical coordinate model of the surface to be worked on. Transform the surface path planning into a combined control problem of tangential angle and axial displacement, and generate a bow-shaped reciprocating coverage path to achieve automatic generation of a full-coverage path for complex surfaces. Approximate the surface of the tunnel lining trolley or cylindrical tank to be worked on as a cylindrical geometric model, and establish a cylindrical coordinate system with the surface axis as the horizontal axis and the tangential angle as the vertical axis. Transform the full-coverage path planning problem of the surface into a combined control problem of axial displacement and tangential angle in cylindrical coordinates, achieving an equivalent transformation of complex surface motion to a planar coordinate system, reducing the computational complexity of path planning and control, and based on the preset of the work tool... The grinding width d is used as the spacing parameter between adjacent paths. In the axial dimension of the cylindrical coordinate system, the surface to be worked is divided into multiple parallel arc paths, so that the spacing between each path in the axial projection is equal to or greater than the grinding width d. This ensures that the work area is evenly divided and provides an accurate path topology for subsequent full-coverage path execution. A bow-shaped reciprocating coverage strategy is adopted to control the robot to move back and forth along each arc path in sequence, so that the movement direction between adjacent paths is opposite. After completing the current path, it automatically switches to the next adjacent path until all the divided paths are covered in sequence, so as to achieve efficient and continuous coverage of the work area, avoid path omissions and reduce idle travel time. It should be noted that when establishing the cylindrical coordinate model of the surface to be worked on, the surface of the tunnel lining trolley or cylindrical tank is first approximated as a cylindrical geometric model, and a cylindrical coordinate system is established with the surface axis as the horizontal axis and the tangential angle as the vertical axis. After the robot is powered on, the operator inputs the preset grinding width d of the working tool through the host computer. The system automatically uses this parameter as the axial spacing between adjacent paths. Based on this, the system divides the surface to be worked on into multiple parallel arc paths along the axis, ensuring that the spacing of each path in the axial projection is equal to or slightly greater than the grinding width d. This transforms the full-coverage path planning problem of the surface into a combined control problem of axial displacement and tangential angle in cylindrical coordinates. During path generation and execution, a bow-shaped reciprocating coverage strategy is adopted to control the robot to move back and forth along each arc path in sequence. When the robot completes the current arc path, the system automatically controls it to stop the current movement and, based on encoder feedback... The displacement information controls the robot to translate along the axial direction by a preset grinding width d, allowing the robot to switch to the starting position of the adjacent arc path. At the same time, according to the bow-shaped strategy, the direction of travel is automatically switched so that the directions of travel between adjacent paths are opposite, achieving efficient and continuous coverage of the work area until all divided paths are covered in sequence. In the actual operation, the robot first adheres to the work surface and remains stationary for 30 seconds. The system records the initial tilt angle sensor data as the attitude reference. The operator selects the left or right entry mode to determine the entry direction of the initial path. The system determines the starting point and initial direction of travel of the first arc path based on this selection. Through multiple translation experiments on the work surface, the system pre-calibrates the encoder correction coefficient λ to correct the lateral translation deviation caused by tire friction or surface unevenness, ensuring that the robot can accurately reach the predetermined path position during axial translation, ensuring the accuracy of path coverage and the quality of work. Furthermore, step 1 also includes: during the generation of the bow-shaped path, recording the stable tilt angle sensor data obtained after the robot initially adheres to the working surface and remains stationary for 30 seconds, and recording the initial value of the X-axis pitch angle as... The initial value of the Y-axis roll angle is denoted as And the maximum pitch angle of the X-axis at the starting point of the trolley is recorded as To establish a reliable benchmark threshold for subsequent attitude judgment and path endpoint recognition, the entry direction of the initial grinding path is determined according to the left or right entry mode selected by the operator. Based on the entry direction, the starting point position and initial travel direction of the first arc path are determined to ensure that the robot's initial movement trend matches the layout of the work surface, thereby improving the adaptability of path planning. The encoder correction coefficient λ is calibrated by conducting multiple translation experiments on the work surface to correct the deviation between the robot's lateral translation distance and the theoretical value (grinding width d) caused by tire friction or uneven surface, thereby eliminating lateral translation error and ensuring axial positioning accuracy during path switching. It should be noted that after the robot adheres to the work surface, it is first kept still for 30 seconds to eliminate the impact of disturbances during start-up and shutdown, and to obtain stable tilt sensor data as the attitude reference for subsequent control. During this time, the initial value of the X-axis pitch angle is automatically recorded. Initial value of Y-axis roll angle And measure the maximum pitch angle of the X-axis at the starting point of the trolley. The acquired parameters are used as benchmark thresholds for subsequent robot posture judgment and path endpoint recognition, and are used to monitor the robot's position on the curved surface in real time, ensuring accurate switching of motion control at each stage and timely identification of abnormal states. Based on the layout and starting position of the work surface, the operator selects either a left-entry or right-entry mode via the host computer, thereby determining the starting point and initial direction of the first arc path. This setting determines the robot's initial upward or downward movement trend along the arc surface and coordinates with the reversing logic of the subsequent bow-shaped path to ensure that the robot can sequentially cover each arc in a predetermined order. Path; Due to factors such as changes in friction between the tires and the wall and unevenness of the surface when the robot travels on curved surfaces, there is a tendency for deviations to occur between the actual lateral translation distance and the theoretically set grinding width d. To solve this problem, multiple translation experiments are conducted on the working surface. By comparing the measured displacement with the theoretical displacement, the encoder correction coefficient λ is pre-calibrated. In actual operation, the displacement data fed back by the encoder is corrected in real time based on the encoder correction coefficient λ to ensure the axial translation accuracy of the robot when switching paths, ensure uniform spacing of each arc path, and improve the work coverage and consistency. Step 2: Based on the inertial sensors and wheel speed encoders mounted on the robot body, a fixed sampling frequency is set to collect robot attitude, angular velocity, and wheel speed data in real time, achieving real-time and accurate acquisition and time alignment of attitude and displacement data. A tilt sensor and wheel speed encoder are installed on the robot body, where the tilt sensor is used to collect the robot's X-axis pitch angle in real time. and Y-axis roll angle The wheel speed encoder is used to collect the rotational speed or rotational angle pulse data of the robot's drive wheels in real time, providing raw data support for surface positioning and attitude calculation. A fixed data sampling frequency is set so that the tilt sensor and wheel speed encoder can collect data synchronously under the same time reference, so that the attitude data and displacement data are time-aligned, ensuring the time consistency of attitude and displacement data during dead reckoning. The collected three-axis attitude angle data, angular velocity data and left and right wheel pulse counts are recorded and stored in real time by the host computer, providing a data foundation for path tracking accuracy analysis and anomaly tracing. It should be noted that a tilt sensor and a wheel speed encoder are installed on the robot body. The tilt sensor is fixed at the geometric center of the robot and is used to collect the robot's X-axis pitch angle in real time. and Y-axis roll angle The data reflects the robot's tangential position and lateral attitude on the curved surface. The wheel speed encoder is installed at the end of the drive wheel axle to collect the rotational speed and angular pulse data of the robot's left and right drive wheels in real time, serving as the basic input for axial displacement calculation. The sensor selection meets the protection level requirements of the curved surface working environment, ensuring long-term stable operation under dusty and humid conditions. A fixed data sampling frequency is set so that the tilt sensor and wheel speed encoder can collect data synchronously under the same time reference. A unified clock source is used to trigger the sampling command, ensuring that the attitude data and displacement data at each moment correspond one-to-one in time, eliminating the data misalignment problem caused by the difference in sampling timing, and ensuring that the tangential position and axial displacement can be combined and calculated in the same time coordinate system during subsequent dead reckoning, improving positioning accuracy and real-time response. The host computer records and stores the collected three-axis attitude angle data, angular velocity data, and left and right wheel pulse counts in real time. The data recording adopts a combination of cyclic caching and key event triggered storage. The cache is continuously updated under normal operation. When key events such as path endpoints, attitude abnormalities, or slippage are detected, the data of the preceding and following time periods are automatically persisted for subsequent analysis of path tracking accuracy and the cause of the abnormality. Step 3: Use inertial sensors to obtain the pitch angle, calculate the tangential position, and measure the travel distance using wheel speed encoders to estimate the axial displacement for precise surface positioning. Fuse the pitch angle and encoder data to achieve surface dead reckoning positioning without external references. The pitch angle is obtained in real-time using tilt sensors. The data, combined with the geometric relationships of the curved surface, is used to calculate the robot's current position in the tangential direction. When the value is large, the robot is determined to be in a lower position on the arc surface. As the distance gradually decreases, the robot is determined to be rising along the curved surface, enabling real-time autonomous perception of the robot's height and position on the curved surface. This provides accurate state criteria for path tracking. The robot's travel distance is calculated by collecting the number of drive wheel pulses from the wheel speed encoder, and the axial displacement is estimated by combining the encoder correction coefficient λ. This yields the robot's current position in the axial direction of the curved surface, effectively eliminating displacement deviations caused by friction and unevenness, and ensuring the accuracy and consistency of axial positioning. The estimated tangential angle position and axial displacement are fused in a cylindrical coordinate system to form the robot's two-dimensional position coordinates on the curved surface. This enables dead reckoning and positioning on the curved surface without the need for an external positioning reference, constructing a complete curved surface position coordinate system, and achieving autonomous positioning and path tracking without an external reference. It should be noted that the pitch angle is obtained in real time using a tilt sensor. The data, combined with the cylindrical geometric model, allows for real-time calculation of the robot's current position in the tangential direction, and continuous monitoring during actual operations. Numerical change, when detected When the value is large, it is determined that the robot is currently at a lower position on the arc surface; when As the value gradually decreases, it is determined that the robot is rising along the arc surface. This tangential position information serves as the main state criterion for the robot's movement along the arc path, used for path endpoint identification and motion direction switching control, ensuring that the robot can accurately perceive its real-time position in the height direction of the curved surface. The rotational speed and rotational angle pulse data of the left and right drive wheels are collected by the wheel speed encoder to calculate the robot's travel distance in real time. Combined with the pre-calibrated encoder correction coefficient λ, the axial displacement is estimated. In actual operation, the encoder correction coefficient λ is used to correct the lateral translation deviation caused by factors such as tire friction changes and surface unevenness, ensuring the accuracy of the axial displacement calculation. The current position of the robot in the axial direction of the curved surface is calculated in real time based on the encoder pulse accumulation, providing accurate displacement feedback for translation control during path switching and ensuring uniform spacing between each arc path. The tangential angle position and the estimated axial displacement are fused in a cylindrical coordinate system. A unified coordinate system is constructed with the axial direction as the horizontal axis and the tangential angle as the vertical axis. The tangential position calculated by the pitch angle and the axial displacement corrected by the encoder are combined to form the robot's two-dimensional position coordinates on the curved surface. Step 4: Generate a parallel arc path based on the width of the work tool, and cover the work area using a bow-shaped strategy. Automatically stop moving based on the pitch angle change threshold and time conditions, control the robot to translate along the axis to the starting point of the adjacent path and switch directions, and automatically determine the path endpoints based on the pitch angle to achieve autonomous lane changing and continuous coverage of the bow-shaped path. Step 5: Monitor roll and pitch angle changes during travel, detect attitude deviation or slippage, and perform correction, retry or manual takeover operations to ensure path tracking stability. Monitor attitude changes in real time and automatically perform correction and retry to ensure stable and reliable path tracking. Step 6: The infrared sensor detects the surface boundary and triggers deceleration, stop or path adjustment control when it approaches the edge of the work area. After the task is completed, the trajectory map is output and prompts for grinding depth verification. The detection of the surface edge automatically triggers safety control and outputs the trajectory map to realize the visual traceability of work quality.

[0020] Example 2, as Figure 1 , Figure 2 As shown, based on Embodiment 1, the present invention provides a technical solution: Step 4 specifically includes: during the grinding process of the robot moving along the arc path, real-time monitoring of the pitch angle obtained by the tilt sensor. Change, when When the angle is less than 5° and the state is maintained for more than or equal to 1 second, it is determined that the robot has reached the highest point of the current arc path. This ensures that the robot accurately triggers path switching at the vertex of the curved surface, avoiding overshoot or premature lane change. After the determination of reaching the highest point is established, the robot is controlled to immediately stop its current forward movement. Based on the encoder correction coefficient λ, the robot is controlled to translate along the axial direction by a distance d, i.e., the grinding width, from the highest point of the current path to the starting point of the adjacent path. This ensures that the robot accurately reaches the starting point of the next path, eliminating the cumulative error of lateral displacement. After the translation is completed, the robot is controlled to switch to the backward grinding state and move in the opposite direction along the adjacent arc path to continue the grinding operation. This achieves continuous back-and-forth coverage of the bow-shaped path, improving work efficiency and path integrity. It should be noted that during the grinding process, the system monitors the pitch angle collected by the tilt sensor in real time. The numerical change in pitch angle reflects the robot's height position on the tangential direction of the curved surface, and the control system reads this value at a fixed frequency. The data is collected and compared with a preset threshold of 5°, while a timer continuously records the data. The duration below this threshold, when When the angle is less than 5° and the state is maintained continuously for 1 second or more, it is determined that the robot has reached the highest point of the current arc path. The path endpoint processing program is then triggered to prepare for the path switching operation. After the determination of reaching the highest point is confirmed, the control system immediately sends a stop command to the drive unit, causing the robot to end the current forward grinding state. Subsequently, based on the pre-calibrated encoder correction coefficient λ, the displacement data fed back by the encoder is corrected in real time, and the robot is controlled to translate along the axial direction of the curved surface by a preset grinding width d. This translation process uses the corrected encoder pulse accumulation as feedback to ensure that the robot moves accurately from the highest point of the current path to the starting position of the adjacent arc path, ensuring that the spacing between adjacent paths is uniform and avoiding cumulative errors caused by tire friction or uneven surface. After translating to the correct position, the robot automatically switches the direction of travel according to the bow-shaped path coverage strategy, controls the robot to enter the backward grinding state, and moves in the opposite direction along the adjacent arc path to continue the grinding operation. Throughout the switching process, the attitude sensor data is continuously monitored to ensure that the robot's attitude is stable and the path is accurate. This cycle repeats, and the robot completes the grinding task of each arc path in sequence. In addition, step 4 also includes: real-time monitoring of the pitch angle during the robot's backward grinding process. Change, when Greater than or equal to And the current state is maintained continuously for a time greater than or equal to 1 second, or Greater than or equal to Furthermore, when the state is maintained continuously for more than or equal to 1 second, or when the trigger signal of the bottom infrared sensor is detected, it is determined that the robot has reached the lowest point of the current arc path, ensuring the reliability of the lowest point identification and multi-source redundancy, avoiding path switching errors. After the lowest point determination is established, the robot is controlled to immediately stop the current backward movement, and according to the encoder correction coefficient λ, the robot is controlled to translate along the axial direction by a distance d, from the lowest point of the current path to the starting point of the next adjacent path, ensuring accurate path switching position and maintaining the consistency of the spacing between adjacent paths. After the translation is completed, the robot is controlled to switch to the forward grinding state and move forward along the next adjacent arc path. This cycle is repeated to achieve continuous coverage of the bow-shaped path, realize full coverage of the work area and automated cycle, and improve work efficiency and path integrity. It should be noted that during the robot's backward grinding process along the arc path, the control system reads the pitch angle collected by the tilt sensor in real time at a fixed sampling frequency. The value is then compared with a preset threshold for determining the lowest point of the path. When a value is detected... Greater than or equal to And if this state is maintained continuously for 1 second or more, it is determined that the robot has reached the top return point of the current path; or when Greater than or equal to If the position is maintained for 1 second or more, the robot is determined to have returned to its initial height. If the bottom infrared sensor is triggered during movement, the robot is directly determined to have reached the lowest point of the path. If any condition is met, the lowest point arrival determination is immediately triggered, and a stop command is sent to the drive unit, causing the robot to end its current backward grinding state and prepare to perform a path switching operation. After the lowest point arrival determination is successful, the control system corrects the displacement data fed back by the encoder in real time according to the pre-calibrated encoder correction coefficient λ, and controls the robot to translate along the axial direction of the curved surface by a preset grinding width d. This translation process uses the corrected encoder pulse accumulation as the displacement feedback basis to ensure that the robot moves accurately from the current lowest point of the path to the adjacent arc. The starting position of the line path is effectively compensated in real time by the encoder correction coefficient to eliminate lateral translation deviation caused by tire friction changes or surface unevenness, ensuring uniform spacing between adjacent paths and providing accurate path starting point positioning for subsequent grinding operations. After translation into position, the control system automatically switches the direction of travel according to the bow-shaped path coverage strategy, controlling the robot to enter the forward grinding state and move forward along the next adjacent arc path to continue grinding operations. Throughout the switching process, the attitude sensor data is continuously monitored. This cycle repeats, and the robot sequentially completes the forward and backward grinding tasks of each arc path, achieving automated continuous coverage of the curved surface work area, ensuring the orderly progress of the work process and the effective execution of path planning. Step 5 specifically includes: real-time monitoring of the roll angle during the robot's movement. With initial value The deviation, when When a robot is detected to have deviated from its posture, it immediately stops its current movement to prevent uneven polishing or equipment instability caused by abnormal posture. After stopping, the control system performs a micro-rotation correction operation in place, causing the robot to rotate around the vertical axis until the roll angle is reached. Return to the allowed range This ensures that the robot regains a stable working posture. After the roll angle correction is completed, the control system restarts the movement and continues to perform the interrupted grinding operation, ensuring the continuity of the operation and the accuracy of path tracking. It should be noted that during the actual operation of the robot, the tilt sensor collects the roll angle in real time. The data control system compares this value with a pre-recorded initial value at a fixed frequency. Perform a comparison, when detected If the robot's current lateral posture is determined to be outside the allowable range and poses a risk of tilting, a stop command is immediately sent to the drive unit to interrupt the current movement and prevent uneven grinding or equipment instability caused by abnormal posture. After stopping, the control system automatically enters the posture correction program and performs a micro-rotation operation in place. The robot slowly rotates around the vertical axis, with the tilt sensor providing continuous real-time feedback. Value, until detected Once the posture has been determined to have returned to a safe and permissible range, the correction process is complete, ensuring that the robot has regained a stable working posture. After the roll angle correction is completed, the control system resumes the operation of the drive unit, and the robot continues to perform the interrupted grinding operation along the original path. Throughout the correction and recovery process, the system maintains continuous monitoring of the posture data to ensure stable correction results and guarantee the continuity and safety of the operation. Furthermore, step 5 also includes: real-time monitoring of the pitch angle during the robot's movement. The rate of change, when detected If the change is less than 1° within 10 seconds, and the robot is judged to have slipped based on its current motion state, it immediately stops moving to effectively prevent path deviation caused by slippage and ensure accurate work position. Once slippage is confirmed, the robot is controlled to retreat to the bottom of the path. The criterion for retreating to the correct position is... Greater than or equal to Furthermore, if this state is maintained for a continuous time of 1 second or more, or if the trigger signal of the bottom infrared sensor is detected, the robot will re-establish a stable starting reference for the operation, eliminate positional uncertainty, and after retreating to the bottom of the path, perform in-situ micro-rotation correction to bring the roll angle back to the allowable range and restart the forward grinding. If the slippage phenomenon occurs repeatedly and cannot be eliminated by automatic correction, the system will enter a protection state and prompt the operator to intervene manually, providing safety redundancy for abnormalities that cannot be automatically recovered, and ensuring that the equipment and operation process are safe and controllable. It should be noted that during the actual grinding operation of the robot, the control system reads the pitch angle collected by the tilt sensor in real time at a fixed sampling frequency. Data, and continuously monitor its rate of change, when detected If the change in angle does not exceed 1° within 10 consecutive seconds, and considering that the robot is currently moving, a slippage is determined to have occurred. At this point, the control system immediately sends a stop command to the drive unit, interrupting the current forward or backward movement to prevent path deviation or loss of work position due to slippage, ensuring the robot can respond to abnormal conditions immediately. Once slippage is confirmed, the system automatically enters a reversal and correction procedure. The robot first performs a reversal operation, controlling its movement in the opposite direction along the original path until it reaches the bottom of the path. The criteria for determining when the robot has reached the correct position employ a dual-protection mechanism: one is the pitch angle... Greater than or equal to Furthermore, this state must be maintained continuously for 1 second or longer, and secondly, the bottom infrared sensor must detect a trigger signal. Meeting either condition is considered a successful retraction. This retraction strategy ensures the robot can return to a stable starting position, eliminating positional uncertainty caused by slippage and establishing a reliable reference point for restarting the grinding operation. After retracting to the bottom of the path, the system executes a micro-rotation correction program to adjust the roll angle. Regression to initial value The deviation is within the allowable range of 2° to ensure that the robot's lateral posture is restored to stability. After the correction is completed, the robot resumes the forward grinding operation. If the slippage phenomenon occurs repeatedly in subsequent operations and the automatic correction program cannot effectively eliminate it, the system will enter the protection state, suspend automatic operation and prompt the operator to make manual intervention. While ensuring the continuity of operation, it provides safety redundancy for abnormal situations that cannot be automatically recovered, ensuring the safety and controllability of equipment and operation process. Step 6 specifically includes: During the process of the robot performing the bow-shaped path coverage, the curved surface edge is detected in real time by an infrared sensor set on the side of the robot. When the infrared sensor detects an edge signal, it is determined that the robot has entered the last path, realizing autonomous edge recognition and ensuring the accuracy and reliability of the triggering in the finishing stage. After the determination of entering the finishing mode is established, the current movement is stopped immediately, and the robot is controlled to move laterally by 0.1m in the opposite direction of the current movement direction, so that the robot can get away from the edge danger area, effectively avoiding the robot from falling out of the boundary and ensuring the safety of equipment and operation. After moving to the safe area, the grinding task of the last arc is completed. After the task is completed, the drive is stopped, and the actual movement trajectory and the ideal trajectory curve are uploaded to the host computer. In areas where the trajectory deviation is too large, a prompt is made to check the grinding depth, realizing the visual traceability of the operation process and providing data support for quality control and path optimization. It should be noted that during the robot's bow-shaped path coverage process, infrared sensors positioned on the robot's side continuously monitor the curved surface edges in real time. When the infrared sensors detect an edge signal, it determines that the robot has entered the last arc path, triggering the finishing mode. The control system immediately sends a stop command to the drive unit, interrupting the current forward or backward movement to prevent the robot from falling due to continued movement. Subsequently, the robot is controlled to move laterally 0.1m in the opposite direction to its current direction of movement, quickly moving it away from the dangerous edge area and into a safe working position, providing safety assurance for completing the grinding task of the last path. After moving to the safe area, according to the bow-shaped path coverage strategy, the grinding operation of the last arc path continues. During this process, the control system... The system continuously monitors the tilt sensor and encoder data to ensure the robot moves stably along the predetermined trajectory until the endpoint judgment conditions of the path are met. After the task is completed, the system automatically stops the drive unit and ends the automatic grinding operation. The entire finishing process realizes orderly control from edge detection and safe retreat to complete path execution, ensuring that the robot can safely and efficiently complete the remaining tasks in the curved surface edge area. After the task is completed, the actual motion trajectory and ideal trajectory curve are uploaded to the host computer for operators to view and analyze. In areas with excessive trajectory deviation, the system automatically marks and prompts for grinding depth verification, realizing the visual traceability of the operation process. This provides an objective basis for system debugging and path planning accuracy verification, further improving the automation level and quality control capability of curved surface grinding operations.

[0021] Example 3, as Figure 1As shown, based on embodiments 1-2, the present invention provides a technical solution: the overall control logic of the robot is based on the "attitude perception - state judgment - phased motion control - anomaly correction" line. Based on the robot's attitude information on the working surface and encoder displacement feedback, it realizes the automation, orderliness and safety control of the entire grinding operation process. The system divides the operation process into an initialization stage, a normal grinding stage, an attitude anomaly handling stage and a finishing stage by setting multi-level thresholds and duration criteria, thereby ensuring stable grinding trajectory, uniform working force and adaptive capability to slippage and edge conditions.

[0022] I. Initialization Phase Before the operation begins, the robot first adheres to the work surface and remains stationary for 30 seconds to eliminate momentary disturbances and acquire stable initial data from the attitude sensors. During this stage, the grinding width is preset. Encoder correction coefficient Record the initial tilt sensor angle. The operator can select left-in / right-in mode to determine the entry direction of the initial grinding path. After initialization, the system enters automatic grinding mode. The axial distance between two adjacent grinding paths is denoted as the grinding width. ; In actual operation, due to tire friction, uneven surfaces, or control errors, the robot's lateral translation distance may differ from the theoretical value. There is a deviation, so an encoder correction factor is set. Used to correct for lateral translation distance It can be calibrated through multiple translation experiments on a horizontal or vertical surface; Tilt sensor The pitch angle of the axis is denoted as A larger value indicates that the robot is at a lower position on the arc surface, while a gradually decreasing value indicates that the robot is rising along the arc surface. Tilt sensor The roll angle of the shaft is denoted as Ideally, the roll angle of a robot is approximately 0 when it moves along an arc. Record the initial tilt angle sensor readings after the robot adheres to the work surface and before automatic cleaning begins. The axis angle is ; Record the initial tilt angle sensor readings after the robot adheres to the work surface and before automatic cleaning begins. The axis angle is .

[0023] II. Normal Polishing Stage In normal mode, the robot moves along a preset path while performing forward grinding, and real-time monitoring is conducted. Change, when Furthermore, if this state is maintained continuously for more than or equal to 1 second, it is determined that the highest point of the path has been reached. Once the criterion is met, the path switching state is switched, and the encoder controls the translation distance. It moves to the next starting point on the path, and then enters the reverse polishing state. And the duration of this state is greater than or equal to 1 second or Furthermore, if this state is maintained continuously for more than or equal to 1 second, or if a bottom infrared trigger signal is detected, it is determined that the lowest point of the path has been reached. At this time, the encoder controls the translation distance. The process is repeated until the next path begins, and then the polishing process continues.

[0024] After the robot adheres to the work surface and before automatic cleaning begins, the starting point of the trolley is... The maximum pitch angle of the axis is denoted as .

[0025] III. Attitude Anomaly Handling Stage The system monitors and The rate and magnitude of change are used to determine abnormal attitude states: when an abnormal attitude state occurs... or The change within 10 seconds does not exceed 1. If slippage is detected, stop moving forward and retreat to the bottom of the path. And this state is maintained continuously for more than or equal to 1 second (or triggered by bottom infrared), the system performs in-situ micro-rotation correction, so that... Return to the allowable range ( The system restarts the grinding process. If slippage occurs repeatedly and cannot be eliminated automatically, the system enters a protection state, prompting the operator for manual intervention. During the horizontal translation path transition, if the current... The deviation from the initial value exceeds the set threshold ( If so, in-situ micro-rotation correction will be performed first, until... Return to the permissible range to avoid uneven sanding or slippage risks caused by initial tilting.

[0026] IV. Final Stage During the process of the robot completing the preset bow-shaped path coverage, if the lateral infrared sensor detects an edge, it can be determined that the robot has entered the last path. The system enters the finishing mode, at which point the current forward or backward movement is immediately stopped, and the encoder controls the robot to move laterally by 0.1m in the opposite direction, so that it returns to the safe area and completes the grinding task of the last arc. After the task is completed, the drive stops, the path data is uploaded to the host computer, and the machine stops.

[0027] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0028] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive path tracking method based on fusion of inertial sensor-wheel speed encoder information, characterized in that, Includes the following steps: Step 1: Establish a cylindrical coordinate model of the surface to be worked on, transform the surface path planning into a combined control problem of tangential angle and axial displacement, and generate a bow-shaped round-trip covering path; Step 2: Based on the inertial sensors and wheel speed encoders mounted on the robot body, a fixed sampling frequency is set to collect robot posture, angular velocity and wheel speed data in real time; Step 3: Use an inertial sensor to obtain the pitch angle, calculate the tangential position, and use a wheel speed encoder to measure the travel distance, estimate the axial displacement, and perform precise positioning of the curved surface. Step 4: Generate a parallel arc path based on the width of the work tool, use a bow-shaped strategy to cover the work area, automatically stop moving according to the pitch angle change threshold and time conditions, control the robot to translate along the axis to the starting point of the adjacent path and switch directions; Step 5: Monitor changes in roll and pitch angles during travel, detect attitude deviation or slippage, and perform correction, retry, or manual takeover operations. Step 6: Detect the curved surface boundary using an infrared sensor. When the surface approaches the edge of the work area, trigger deceleration, stop, or path adjustment control. After the task is completed, output a trajectory diagram and prompt for grinding depth verification.

2. The adaptive path tracking method based on information fusion of inertial sensor-wheel encoder according to claim 1, characterized in that: Step 1 specifically includes: The surface of the tunnel lining trolley or cylindrical tank to be operated is approximated as a cylindrical geometric model, and a cylindrical coordinate system is established with the surface axis as the horizontal axis and the tangential angle as the vertical axis. The path planning problem of full surface coverage is transformed into a combined control problem of axial displacement and tangential angle in cylindrical coordinates. Based on the preset grinding width d of the working tool as the spacing parameter between adjacent paths, the surface to be worked is divided into multiple parallel arc paths in the axial dimension of the cylindrical coordinate system, so that the spacing between each path in the axial projection is equal to or greater than the grinding width d. The robot adopts a bow-shaped reciprocating coverage strategy, controlling the robot to move back and forth along each arc path in sequence, so that the movement direction between adjacent paths is opposite, and automatically switches to the next adjacent path after completing the current path, until all the divided paths are covered in sequence.

3. The adaptive path tracking method based on information fusion of inertial sensor and wheel speed encoder according to claim 2, characterized in that: Step 1 further includes: In the arch-shaped path generation process, the stable inclination sensor data obtained after the robot is initially adsorbed to the working surface for 30s is recorded, the initial value of the X-axis pitch angle is recorded as , the initial value of the Y-axis roll angle is recorded as , and the maximum pitch angle of the X-axis at the starting point of the trolley is recorded as ; The entry direction of the initial grinding path is determined based on the left or right entry mode selected by the operator, and the starting point and initial direction of the first arc path are determined based on the entry direction. The encoder correction coefficient λ was calibrated by conducting multiple translation experiments on the working surface to correct the deviation between the robot's lateral translation distance and the theoretical value caused by tire friction or surface unevenness.

4. The adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion according to claim 1, characterized in that: Step 2 specifically includes: A tilt sensor and a wheel speed encoder are installed on the robot body. The tilt sensor is used to collect the robot's X-axis pitch angle in real time. and Y-axis roll angle Data: Wheel speed encoders are used to collect real-time pulse data of the rotational speed or rotational angle of the robot's drive wheels; Set a fixed data sampling frequency so that the tilt sensor and wheel speed encoder can collect data synchronously under the same time reference, so that the attitude data and displacement data are time-aligned; The host computer records and stores the collected three-axis attitude angle data, angular velocity data, and pulse counts of the left and right wheels in real time.

5. The adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion according to claim 1, characterized in that: Step 3 specifically includes: Pitch angle acquired in real time using tilt sensor The data, combined with the geometric relationships of the curved surface, is used to calculate the robot's current position in the tangential direction. When the value is large, the robot is determined to be in a lower position on the arc surface. The robot is determined to be rising along the curved surface as the temperature gradually decreases. The robot's travel distance is calculated by collecting the number of drive wheel pulses from the wheel speed encoder, and the axial displacement is estimated by combining the encoder correction coefficient λ to obtain the robot's current position in the axial direction of the curved surface. The estimated tangential angle position and axial displacement are fused within a cylindrical coordinate system to form the robot's two-dimensional position coordinates on the curved surface.

6. The adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion according to claim 1, characterized in that: Step 4 specifically includes: During the grinding process as the robot moves along the curved path, the pitch angle acquired by the tilt sensor is monitored in real time. Change, when If the angle is less than 5° and the state is maintained continuously for more than or equal to 1 second, it is determined that the robot has reached the highest point of the current arc path. Once the highest point is reached, the robot immediately stops its current forward movement and, based on the encoder correction coefficient λ, controls the robot to translate along the axial direction by a distance d, i.e., the grinding width, from the highest point of the current path to the starting point of the adjacent path. After the translation is completed, the robot switches to the backward grinding state and moves in the opposite direction along the adjacent arc path to continue the grinding operation.

7. The adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion according to claim 6, characterized in that: Step 4 also includes: During the robot's backward grinding process, the pitch angle is monitored in real time. Change, when Greater than or equal to And the current state is maintained continuously for a time greater than or equal to 1 second, or Greater than or equal to Furthermore, if this state is maintained for a continuous time of 1 second or more, or if the trigger signal of the bottom infrared sensor is detected, it is determined that the robot has reached the lowest point of the current arc path. Once the lowest point is reached, the robot immediately stops its current backward movement and, based on the encoder correction coefficient λ, controls the robot to translate a distance d along the axis from the lowest point of the current path to the starting point of the next adjacent path. After the translation is completed, the robot is controlled to switch to forward grinding mode and move forward along the next adjacent arc path. This process is repeated to achieve continuous coverage of the bow-shaped path.

8. The adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion according to claim 1, characterized in that: Step 5 specifically includes: Real-time monitoring of roll angle during robot movement With initial value The deviation, when If the robot is detected to have deviated from its posture, it should immediately stop its current movement. After stopping, the control system performs a micro-rotation correction operation in place, causing the robot to rotate around the vertical axis until the roll angle is reached. Return to the allowed range ; After the roll angle correction is completed, the control system restarts the travel motion and continues the interrupted grinding operation.

9. The adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion according to claim 8, characterized in that: Step 5 further includes: Real-time monitoring of pitch angle during robot movement The rate of change, when detected If the change is less than 1° within 10 seconds, the robot is determined to be slipping based on its current motion state, and its current movement is immediately stopped. Once the slippage detection is established, the robot is controlled to retreat to the bottom of the path. The criterion for retreating to the correct position is... Greater than or equal to And this state is maintained continuously for a period of 1 second or more, or the trigger signal of the bottom infrared sensor is detected; After retracing to the bottom of the path, perform in-situ micro-rotation correction to bring the roll angle back to the allowable range, and start grinding forward again. If the slippage phenomenon occurs repeatedly and cannot be eliminated by automatic correction, the system enters protection mode and prompts the operator to intervene manually.

10. The adaptive path tracking method based on inertial sensor-wheel speed encoder information fusion according to claim 1, characterized in that: Step 6 specifically includes: During the process of the robot performing the bow-shaped path coverage, the edge of the curved surface is detected in real time by infrared sensors set on the side of the robot. When the infrared sensors detect the edge signal, it is determined that the robot has entered the last path. Once the determination of entering the finishing mode is established, immediately stop the current movement and control the robot to move laterally by 0.1m in the opposite direction to the current movement direction, so that the robot can get away from the edge danger zone; After translating to a safe area, continue to complete the grinding task of the last arc. After the task is completed, stop driving, upload the actual motion trajectory and the ideal trajectory curve to the host computer, and prompt for grinding depth verification in areas with excessive trajectory deviation.

Images