A method for autonomous navigation positioning and punching operation of a crawler-type punching robot

Abstract

This invention discloses an autonomous navigation, positioning, and drilling method for a tracked drilling robot, comprising: a global initialization and registration step, which registers BIM point clouds with real-time laser point clouds to obtain an initial pose; a multi-sensor fusion SLAM step, which fuses data from LiDAR, visual depth cameras, IMUs, and encoders to generate a dense environmental map in real time and output the pose; a BIM-constrained path planning step, which constructs a cost map fused with BIM structural constraints and plans a global path and local trajectory; a parallel mechanism positioning step, which drives the parallel mechanism to the target drilling point through inverse kinematics solving; a vision-force closed-loop drilling step, which integrates visual servoing and force / torque monitoring to achieve safe and accurate drilling; and a global error correction step, which re-registers the point cloud to eliminate accumulated errors. This invention deeply integrates BIM prior information with multi-source perception to achieve full-process automation from navigation to drilling, significantly improving positioning accuracy, safety, and construction efficiency.

Description

Technical Field

[0001] This invention relates to the field of construction robot technology, specifically to an autonomous navigation, positioning, and drilling operation method for a tracked drilling robot, and more particularly to an intelligent operation method that integrates Building Information Modeling (BIM) and multi-sensor information fusion technology to achieve autonomous navigation of the tracked chassis, precise positioning of the parallel mechanism, and vision-force closed-loop safe drilling. Background Technology

[0002] Indoor drilling is a crucial part of building construction, widely used in pipeline installation, bracket fixing, and ceiling construction. Traditional indoor drilling mainly relies on manual measurement and handheld electric drills. Workers must repeatedly confirm hole locations, avoid embedded pipelines, and judge concrete thickness on-site. This process is not only inefficient but also often results in significant hole position errors, failing to meet the precision requirements of modern industrialized construction. Furthermore, manual operation faces safety hazards such as working at heights, in confined spaces, and in cluttered, dusty, and noisy environments. It is also difficult to detect and stop obstacles such as reinforcing bars and pipes in time, potentially leading to structural damage or personnel injury.

[0003] With the widespread application of Building Information Modeling (BIM) technology in the construction field, BIM provides three-dimensional global structural data for construction, enabling the pre-annotation of hole coordinates, pipeline routes, and wall material properties during the design phase, thus providing fundamental data support for construction automation. However, in current technologies, BIM technology is mainly used for design briefings and construction management, and has not yet been effectively integrated with real-time on-site sensing and automated operation equipment, resulting in the inability of BIM's refined data to fully play its role in the construction execution phase.

[0004] Meanwhile, the development of Simultaneous Localization and Mapping (SLAM) technology enables mobile robots to build environmental maps in real time and achieve self-localization in unknown or dynamic environments. In particular, SLAM technology combining LiDAR and vision has been widely applied in fields such as material handling robots, cleaning robots, and inspection robots. For example, existing literature discloses a navigation method for indoor mobile robots based on LiDAR and visual SLAM, which can achieve autonomous movement and obstacle avoidance in complex environments. However, these mobile robots are mostly used for tasks such as material handling and environmental cleaning, and their end effectors are usually grasping or cleaning mechanisms. They lack dedicated actuators and force-sensing closed-loop control for high-precision drilling operations, making it difficult to meet the stringent requirements of building construction for hole position accuracy (usually within ±5mm), drilling depth control, and real-time impact force monitoring.

[0005] In the field of automated drilling operations, existing technologies include some fixed or rail-mounted drilling equipment, such as drilling rigs used in subway tunnels and utility tunnel construction. These devices typically use mechanical guide rails for positioning, which can ensure a certain level of accuracy, but they are large, inconvenient to move, and difficult to adapt to the variable construction environment and dispersed drilling locations in indoor buildings. A few studies have attempted to mount industrial robotic arms on mobile platforms for drilling operations, but the combined effect of the positioning error of the mobile platform and the absolute positioning error of the robotic arm makes it difficult to guarantee end-effector accuracy; moreover, existing solutions mostly use single vision positioning or teach-and-playback methods, which cannot perceive changes in concrete resistance and steel bar collisions in real time during drilling, posing a risk of damaging the drill bit or the structure.

[0006] In summary, current technologies lack a complete automated drilling method that can organically combine BIM global data, multi-sensor fusion SLAM real-time perception, high-precision parallel mechanism positioning, and vision-force closed-loop control to achieve a complete automated drilling operation, from global path planning to precise local positioning and safe drilling. Overcoming the shortcomings of traditional manual drilling—low efficiency, poor accuracy, and significant safety hazards—and resolving the technical problems of disconnect between BIM data and on-site perception, insufficient end-point positioning accuracy, and lack of real-time monitoring of the drilling process in existing automated solutions, are urgent technical challenges that need to be addressed by those skilled in the art. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention provides a method for autonomous navigation, positioning, and drilling operations of a tracked drilling robot.

[0008] This invention is implemented as follows: a method for autonomous navigation, positioning, and drilling operations of a tracked drilling robot, comprising the following steps:

[0009] Global initialization and registration steps: acquire the first point cloud data of the Building Information Model (BIM) and extract structural features to build a BIM structural feature library; after the robot starts, collect real-time second point cloud data, register the second point cloud data with the first point cloud data, and obtain the robot's initial pose in the global coordinate system; Multi-sensor fusion SLAM steps: Fusion of data from LiDAR, visual depth camera, inertial measurement unit (IMU) and encoder; back-end fusion of LiDAR and visual odometry based on extended Kalman filter (EKF); graph optimization combined with loop closure detection; real-time generation of dense grid map of environment; and calculation of robot's real-time pose. BIM constraint path planning steps: Overlay structural constraints from the BIM structural feature library onto the dense grid map of the environment to construct a cost map containing obstacle costs and BIM structural constraint costs; On the cost map, plan the global path and local trajectory from the real-time pose to the target borehole point marked in the BIM. Parallel mechanism positioning steps: Based on the robot's current base posture and the target drilling point, determine the target end pose of the parallel drilling mechanism; by solving the inverse kinematics of the parallel drilling mechanism, drive the end of the parallel mechanism to move to the target end pose; Vision-force closed-loop drilling steps: During the drilling process, the actual drilling point coordinates are extracted from images captured in real time by a depth camera, the positional error between the actual drilling point and the target drilling point is calculated, and visual servo positioning is performed; at the same time, the pressure sensor data and torque sensor data of the drill bit feed mechanism are monitored in real time to build a vision-force closed-loop control system; when the monitored pressure value exceeds the first preset threshold, the machine is controlled to decelerate or stop; when the monitored torque value exceeds the second preset threshold, it is determined to be a collision with the reinforcing bar and the drill bit is controlled to retract. Global error correction steps: After completing a single drilling, the current third point cloud data is collected, the third point cloud data is re-registered with the BIM structural feature library, the global pose correction value of the robot is calculated, and the current pose of the robot is updated using the global pose correction value to eliminate accumulated errors.

[0010] Further preferably, the global initialization and registration step specifically includes: converting the BIM model into the first point cloud data in IFC format, and performing voxel filtering downsampling on the first point cloud data and the second point cloud data respectively, with the voxel size set to 5mm; extracting the normal vector and curvature features of the filtered point cloud to form the BIM structural feature library; and using the iterative nearest point (ICP) algorithm to register the second point cloud data with the first point cloud data to output the initial pose. .

[0011] Further preferably, the multi-sensor fusion SLAM step specifically includes: providing initial pose estimation through IMU pre-integration; performing line / surface feature matching on the LiDAR point cloud to obtain the laser pose increment; and performing ORB feature matching on the depth camera image to obtain the visual pose increment; using the laser pose increment and visual pose increment as observation vector inputs to the EKF for state prediction and update, with the fusion formula being:

[0012]

[0013]

[0014]

[0015]

[0016] in: for kThe prior state prediction vector at time t; for k The posterior state prediction vector at time 1; for k The posterior state prediction vector at time t; For the observation matrix, This is the state transition matrix; To control the input matrix; The process noise covariance matrix; To observe the noise covariance matrix; The Kalman gain matrix; for k The prior estimate of the covariance matrix at time t; for k The posterior estimated covariance matrix at time 1; for k The posterior estimated covariance matrix at time t; To control the input vector; For the observation vectors of laser and vision; A dual retrieval method using Scan-Context and Bag-of-Words is employed for loop closure detection, triggering global graph optimization to correct the dense grid map of the environment and the real-time pose.

[0017] More preferably, the total cost function of the cost map constructed in the BIM constrained path planning step is:

[0018] in: To plan a specific coordinate position for the robot in space; Let the barrier cost function be used. The cost function for BIM structural constraints; The weighting coefficient for the obstacle cost; The constraint cost weighting coefficient; the weighting coefficient is taken as follows: =0.7, =0.3.

[0019] More preferably, the BIM-constrained path planning step specifically includes: using the A* algorithm to search for a discrete global path from the robot's current position to the target drilling point on the cost map; employing the Dynamic Window Method (DWA) during the execution phase, combined with real-time obstacle point clouds, to generate feasible local trajectories; and performing cubic spline interpolation on the local trajectories to obtain continuous speed control commands. The motion controller of the tracked chassis is sent to the motion controller, where: Instantaneous linear velocity, It represents the instantaneous angular velocity.

[0020] More preferably, the positioning step of the parallel mechanism specifically includes: Based on the target drilling points marked in the BIM and the current base posture of the robot, the target end-effector pose of the parallel mechanism is obtained by combining them; based on the kinematic model of the parallel mechanism, the Levenberg-Marquardt algorithm is used to solve for the iteratively updated joint angle vectors as follows:

[0021] in: For the first k Joint angle vector at the next iteration; For the first k Jacobian matrix at the next iteration; For the first k Damping factor at the next iteration; It is the identity matrix; It is a positive kinematic function; The desired end-target pose.

[0022] More preferably, the vision-force closed-loop drilling step specifically includes: The coordinates of the actual borehole points are obtained by performing Hough line transform detection on the borehole markings using a depth camera. ; Calculate the coordinates of the actual borehole point and the coordinates of the target borehole point. Position error:

[0023] in: actual borehole coordinates The coordinates of the target borehole point are output by inverse kinematics; The error is adjusted by the PID controller to regulate the joint commands of the parallel mechanism, achieving sub-millimeter-level positioning; Establish a pressure sensor model ; This refers to the drill bit feed rate. , For experimental calibration constants; when The system will automatically slow down or stop. Establish a torque sensor model: ; For cutting resistance, Where is the drill bit radius; when If the sudden rise exceeds 12 N·m, it is determined to be a steel bar collision, triggering a retraction and repositioning.

[0024] More preferably, the global error correction step specifically includes: after completing one drilling operation, performing ICP registration again on the currently acquired third point cloud data and the BIM structural feature library to minimize the objective function:

[0025] Based on ICP, a normal distribution transformation NDT is introduced to form the cost function:

[0026] in: It is a rotation matrix; It is a translation vector; The first point cloud in the source cloud i One point; For the target point cloud i One corresponding point; The first point cloud in the source cloud j One point; The mean vector of the target raster is used; the optimal pose correction value is obtained by solving the Levenberg-Marquardt iterative method. ; The optimal pose correction value As the observation input EKF, it completes the smooth update of global pose and controls the cumulative error within 5mm.

[0027] A further preferred option includes a task completion determination step: the system counts the number of boreholes that have been drilled. Compared with the preset total number of boreholes in BIM ;like When the operation is completed, a safety shutdown procedure is executed, including shutting down the drill motor, stopping the track drive, and disconnecting the power supply to the lidar and depth camera; the point cloud data, pose data, and force / torque logs of this operation are saved to the local server, and the control system is reset to standby mode.

[0028] The present invention also relates to a computer program product, comprising a computer program that, when executed by a processor, implements the above-described method for autonomous navigation, positioning, and drilling operations of a tracked drilling robot.

[0029] The advantages and technical effects of this invention are as follows: This invention proposes an autonomous navigation, positioning, and drilling operation method for an adaptive concrete slab drilling robot. It combines BIM global information with multi-sensor fusion SLAM real-time depth perception, achieving a complete automated process from autonomous navigation to precise drilling through six core steps. Compared with existing technologies, this invention has the following significant advantages: High positioning accuracy: Integrating BIM priors and multi-source SLAM, the cumulative error after global error correction is ≤5mm, and the single hole positioning reaches the sub-millimeter level.

[0030] Safe and reliable operation: Vision-force closed-loop control monitors drilling resistance in real time and automatically stops when encountering steel bars or abnormalities to avoid structural damage.

[0031] Highly adaptable to the environment: Laser-visual SLAM provides real-time perception of dynamic obstacles, while BIM-constrained path planning ensures avoidance of structural components, adapting to complex construction sites.

[0032] Efficiency has been greatly improved: the operation time for a single hole has been reduced from 10 minutes by manual labor to 3 minutes by robot, and the overall construction cycle has been reduced by about 80%.

[0033] High degree of automation: It is fully autonomous from start to finish, with automatic data archiving and no need for human intervention. Attached Figure Description

[0034] Figure 1 This is a flowchart of the invention; Figure 2 This is a schematic diagram of the tracked drilling robot for concrete floor slabs according to the present invention.

[0035] In the diagram: 1. Drilling mechanism; 2.1. LiDAR; 2.2. Binocular depth camera; 2.3. Pressure sensor; 2.4. Torque sensor; 2.5. Laser displacement sensor; 3. Platform lifting mechanism; 4. Tracked chassis. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0037] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment uses an adaptive tracked drilling robot for concrete slabs as an example, used for high-precision automated drilling operations on indoor concrete ceilings. The overall structure of the robot is as follows: Figure 2 As shown, the system includes a tracked chassis 4, a platform lifting mechanism 3 mounted on the chassis, and a parallel six-degree-of-freedom drilling mechanism 1. The sensor system includes: a lidar 2.1 (located at the front of the chassis), a binocular depth camera 2.2 (located at the top of the platform), an IMU, and an encoder (built into the chassis). The force sensing system includes a pressure sensor 2.3 and a torque sensor 2.4 mounted on the drill bit feed mechanism; the pressure sensor 2.3 is mounted on the bottom of the drill bit; the torque sensor 2.4 and the laser displacement sensor 2.5 are mounted on the upper part of the drilling rig. The control unit (not shown) communicates with each component and executes the following steps.

[0038] Please refer to the details. Figure 1 A method for autonomous navigation, positioning, and drilling operations of a tracked drilling robot, step 1: global initialization and registration. After the robot starts, it first performs global initialization and registration. It exports the first point cloud data of the ceiling structure from the Building Information Model (BIM), including all preset hole locations, beam and column positions, and wall outlines. The BIM model is converted to IFC format point cloud and voxel filtering downsampling is performed, with the voxel size set to 5mm to reduce data volume while preserving key geometric features. Simultaneously, the normal vector and curvature of the filtered point cloud are calculated, and key structural points (such as beam-column intersections and wall corners) are extracted to construct a BIM structural feature library. The robot acquires real-time second point cloud data of the surrounding environment using LiDAR. The Iterative Closest Point (ICP) algorithm is used to register the second point cloud with the first point cloud in the BIM, minimizing the distance between the point clouds, thereby obtaining the robot's initial pose (including position and orientation) in the global coordinate system. This step provides a global reference for subsequent navigation, avoiding cumulative errors caused by unknown initial position.

[0039] By using 5mm voxel filtering and ICP registration, the robot is quickly positioned in the BIM global coordinate system with an initial pose accuracy of centimeters, laying the foundation for subsequent high-precision navigation.

[0040] Step 2: Multi-sensor SLAM Fusion During movement, the robot continuously fuses data from LiDAR, binocular depth camera, IMU, and encoder to achieve Simultaneous Localization and Mapping (SLAM). The specific process is as follows: IMU pre-integration provides an initial pose estimate within a short timeframe to compensate for motion distortions from the LiDAR and camera. LiDAR odometry extracts line and surface features from the point cloud and matches them with the previous frame to obtain the LiDAR pose increment. Visual odometry extracts ORB features from the depth camera image and obtains the visual pose increment through feature matching. The LiDAR pose increment and visual pose increment are used as observation vectors and input into an Extended Kalman Filter (EKF) for backend fusion. The EKF prediction and update formulas are as follows:

[0041]

[0042]

[0043]

[0044]

[0045] in: fork The prior state prediction vector at time t; for k The posterior state prediction vector at time 1; for k The posterior state prediction vector at time t; For the observation matrix, This is the state transition matrix; To control the input matrix; The process noise covariance matrix; To observe the noise covariance matrix; The Kalman gain matrix; for k The prior estimate of the covariance matrix at time t; for k The posterior estimated covariance matrix at time 1; for k The posterior estimated covariance matrix at time t; To control the input vector; For the observation vectors of laser and vision; The fused pose is used to update the environment map. Simultaneously, a dual retrieval method combining Scan-Context and Bag-of-Words is employed for loop closure detection. Once a loop closure is detected, global graph optimization is triggered to correct accumulated drift, generate a consistent dense grid map, and output the robot's real-time pose.

[0046] IMU pre-integration improves the robustness of motion estimation, EKF tightly couples laser and vision data, dual loop closure detection eliminates cumulative drift, map building accuracy reaches the centimeter level, and real-time pose output frequency is high, meeting the needs of dynamic navigation.

[0047] Step 3: BIM Constraint Path Planning: Based on the dense grid map generated by SLAM, structural constraints from the BIM structural feature library constructed in Step 1 are overlaid to build a fused cost map. The total cost function is defined as:

[0048] in: To plan a specific coordinate position for the robot in space; Let the barrier cost function be used. The cost function for BIM structural constraints; The weighting coefficient for the obstacle cost; The constraint cost weighting coefficient; the weighting coefficient is taken as follows: =0.7, =0.3. On the cost map, the A* algorithm is used to search for a discrete global path from the robot's current position to the target drilling point marked in the BIM. This path takes into account global structural information and avoids passing through fixed obstacles such as walls. During the execution phase, the Dynamic Window Method (DWA) combined with real-time obstacle point clouds is used to generate feasible trajectories within a local area to achieve dynamic obstacle avoidance. Finally, cubic spline interpolation is performed on the local trajectory to obtain continuous and smooth speed control commands. , where: Instantaneous linear velocity, The instantaneous linear velocity; the speed control command is sent to the motion controller of the tracked chassis, driving the robot to move smoothly to the vicinity of the target area.

[0049] Technical benefits: The A* algorithm ensures global optimization, DWA provides real-time obstacle avoidance, spline interpolation smooths the trajectory, the robot moves smoothly, the positioning accuracy is not affected by sudden stops and starts, and the path always follows the BIM structure to avoid collisions with walls.

[0050] Step 4: Positioning of parallel mechanisms After the robot reaches the vicinity of the target drilling point, the end effector (drill bit) of the parallel drilling mechanism needs to be precisely aligned with the drilling point marked in the BIM. First, based on the robot's current base pose (provided by SLAM) and the coordinates of the target drilling point in the BIM, the pose (including position and orientation) of the target end effector of the parallel mechanism is obtained through coordinate transformation. The parallel mechanism is a six-degree-of-freedom platform, and its inverse kinematics solution uses the Levenberg-Marquardt (LM) iterative algorithm. The iterative formula is as follows:

[0051] in: For the first k Joint angle vector at the next iteration; For the first k Jacobian matrix at the next iteration; For the first k Damping factor at the next iteration; It is the identity matrix; It is a positive kinematic function; The desired end-target pose.

[0052] Through iterative calculations, the joint angles required to drive each joint of the parallel mechanism are obtained, and the servo motor is controlled to drive the platform to move so that the drill bit end reaches the predetermined position.

[0053] The LM algorithm efficiently solves nonlinear inverse kinematics, achieving sub-millimeter-level positioning at the end of the parallel mechanism, ensuring that the drill bit is accurately aligned with the hole position marked by the BIM, and providing a precise initial position for subsequent drilling.

[0054] Step 5: Vision-Force Closed-Loop Drilling During the drilling process, a closed-loop control system combining visual servoing and force / torque monitoring is employed. First, images of the drilling area are captured in real-time using a binocular depth camera. Then, Hough line transform is performed on the drill line to obtain the coordinates of the actual drill points. ; Calculate the coordinates of the actual borehole point and the target borehole point. Position error:

[0055] in: actual borehole coordinates The coordinates of the target borehole point are output by inverse kinematics; The error is adjusted by the PID controller to regulate the joint commands of the parallel mechanism, achieving sub-millimeter-level positioning; The error is input to the PID controller, which generates fine-tuning instructions for the joints of the parallel mechanism, achieving sub-millimeter-level visual servo positioning and compensating for deviations caused by minor shaking of the robot chassis or deformation of the mechanism.

[0056] Simultaneously, pressure sensor data and torque sensor data of the drill bit feed mechanism are monitored in real time. A pressure sensor model is established. ; This refers to the drill bit feed rate. , For experimental calibration constants; when The system will automatically slow down or stop. When the monitored pressure value F exceeds the first preset threshold of 150N, it indicates that abnormal resistance has been encountered. The system will automatically control the deceleration or stop immediately to prevent damage to the drill bit or structure. Establish a torque sensor model: ; For cutting resistance, Drill bit radius (in this embodiment) =10mm); when The sudden increase exceeded 12 N·m; When torque value When the rise exceeds the second preset threshold of 12 N·m, it is determined to be a collision with the reinforcing bar. The system immediately triggers the drill bit to retract and reposition, and attempts to drill again at the offset position.

[0057] Technical benefits: Visual servo correction reduces positioning errors to sub-millimeter level, ensuring accurate borehole positioning; real-time pressure / torque monitoring ensures drilling safety; immediate response to abnormalities such as steel reinforcement prevents structural damage; and repositioning improves the success rate of operations.

[0058] Step 6: Global Error Correction After completing a single drilling operation, the robot needs to eliminate accumulated errors to provide a precise reference for the next drilling point. At this point, third-point cloud data of the current environment is acquired using LiDAR. The third-point cloud is then registered again with the BIM structural feature library using ICP, minimizing the objective function.

[0059] Based on ICP, a normal distribution transformation NDT is introduced to form the cost function.

[0060] in: It is a rotation matrix; It is a translation vector; The first point cloud in the source cloud i One point; For the target point cloud i One corresponding point; The first point cloud in the source cloud j One point; The mean vector of the target raster; The optimal pose correction value was obtained by using the Levenberg-Marquardt iterative solution. ; To further improve registration accuracy, a normal distribution transformation (NDT) is introduced on top of ICP to form a cost function, and the optimal pose correction value is obtained by using LM iteration. The correction value is used as the observation input EKF to complete the smooth update of the global pose, strictly controlling the cumulative error within 5mm. The updated pose information is sent to the chassis motion controller and parallel mechanism drive to prepare for the positioning and operation of the next drilling point.

[0061] Technical effect: By using ICP+NDT multi-strategy point cloud registration and EKF smooth update, the cumulative error during movement and operation is effectively eliminated, ensuring continuous high precision in multi-hole operation with a cumulative error of ≤5mm.

[0062] Step 7: Assessing the Completion of the Task The system continuously counts the number of boreholes that have been drilled. And compared with the total number of boreholes preset in the BIM. Comparison. When Upon completion of the task, the system automatically enters the end-of-task process: First, it shuts down the drill motor, stops the track drive, and disconnects the power supply to the LiDAR and depth camera to ensure safety; then, it saves the point cloud data, pose data, and force / torque logs of this task to the local server for subsequent quality checks; finally, it restores the PLC, touch screen, and host computer to standby mode to prepare for the next task; achieving full automation without manual intervention; data archiving facilitates quality traceability, and system reset improves the continuity of operations.

[0063] After completing the above-described workflow, this invention is also implemented as a computer program product, including a computer program stored in a memory. When executed by a processor, the computer program implements the autonomous navigation, positioning, and drilling operation method for the tracked drilling robot. This program can adjust parameters (such as thresholds and weights) according to different construction scenarios, facilitating portability and upgrades.

[0064] Technical benefits: The method can be implemented in software, reducing hardware dependence and improving system flexibility and maintainability.

[0065] In summary, this invention aims to overcome the shortcomings of traditional indoor manual drilling operations, such as low efficiency, large positioning errors, unreliable obstacle avoidance, and difficulty in real-time monitoring of drilling safety. It provides a complete operational method for a tracked drilling robot based on BIM and multi-sensor fusion. By combining hole location and structural information from BIM with a dense 3D map constructed by LiDAR, the robot can perform global path planning and local obstacle avoidance in complex, confined indoor environments with potential obstacles such as rebar and pipelines. Precise drilling positioning is achieved using the inverse kinematics of a parallel six-degree-of-freedom mechanism and real-time line-marking detection via depth camera visual SLAM. Furthermore, a vision-force closed loop is constructed using pressure and torque sensors to monitor drilling resistance and rebar collisions in real time, automatically adjusting or stopping the robot to ensure drilling safety. The overall solution significantly improves operational accuracy and efficiency, providing a reliable automated drilling solution for various construction scenarios such as high-rise buildings, underground utility tunnels, and prefabricated housing.

[0066] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for autonomous navigation, positioning, and drilling operations of a tracked drilling robot, characterized in that, Includes the following steps: Global initialization and registration steps: acquire the first point cloud data of the Building Information Model (BIM) and extract structural features to build a BIM structural feature library; after the robot starts, collect real-time second point cloud data, register the second point cloud data with the first point cloud data, and obtain the robot's initial pose in the global coordinate system; Multi-sensor fusion SLAM steps: Fusion of data from LiDAR, visual depth camera, inertial measurement unit (IMU) and encoder; back-end fusion of LiDAR and visual odometry based on extended Kalman filter (EKF); graph optimization combined with loop closure detection; real-time generation of dense grid map of environment; and calculation of robot's real-time pose. BIM constraint path planning steps: Overlay structural constraints from the BIM structural feature library onto the dense grid map of the environment to construct a cost map containing obstacle costs and BIM structural constraint costs; On the cost map, plan the global path and local trajectory from the real-time pose to the target borehole point marked in the BIM. Parallel mechanism positioning steps: Based on the robot's current base posture and the target drilling point, determine the target end pose of the parallel drilling mechanism; by solving the inverse kinematics of the parallel drilling mechanism, drive the end of the parallel mechanism to move to the target end pose; Vision-force closed-loop drilling steps: During the drilling process, the actual drilling point coordinates are extracted from images captured in real time by a depth camera, the positional error between the actual drilling point and the target drilling point is calculated, and visual servo positioning is performed; at the same time, the pressure sensor data and torque sensor data of the drill bit feed mechanism are monitored in real time to build a vision-force closed-loop control system; when the monitored pressure value exceeds the first preset threshold, the machine is controlled to decelerate or stop; when the monitored torque value exceeds the second preset threshold, it is determined to be a collision with the reinforcing bar and the drill bit is controlled to retract. Global error correction steps: After completing a single drilling, the current third point cloud data is collected, the third point cloud data is re-registered with the BIM structural feature library, the global pose correction value of the robot is calculated, and the current pose of the robot is updated using the global pose correction value to eliminate accumulated errors.

2. The autonomous navigation, positioning, and drilling method for a tracked drilling robot according to claim 1, characterized in that, The global initialization and registration steps specifically include: converting the BIM model into the first point cloud data in IFC format, and performing voxel filtering downsampling on the first and second point cloud data respectively, with the voxel size set to 5mm; extracting the normal vector and curvature features of the filtered point cloud to form the BIM structural feature library; and using the Iterative Closest Point (ICP) algorithm to register the second point cloud data with the first point cloud data to output the initial pose. .

3. The autonomous navigation, positioning, and drilling method for a tracked drilling robot according to claim 1, characterized in that, The multi-sensor fusion SLAM step specifically includes: providing initial pose estimation through IMU pre-integration; performing line / surface feature matching on the LiDAR point cloud to obtain the laser pose increment; and performing ORB feature matching on the depth camera image to obtain the visual pose increment; using the laser pose increment and visual pose increment as observation vector inputs to the EKF for state prediction and update, with the fusion formula being: in: for k The prior state prediction vector at time t; for k The posterior state prediction vector at time 1; for k The posterior state prediction vector at time t; For the observation matrix, This is the state transition matrix; To control the input matrix; The process noise covariance matrix; To observe the noise covariance matrix; The Kalman gain matrix; for k The prior estimate of the covariance matrix at time t; for k The posterior estimated covariance matrix at time 1; for k The posterior estimated covariance matrix at time t; To control the input vector; For the observation vectors of laser and vision; A dual retrieval method using Scan-Context and Bag-of-Words is employed for loop closure detection, triggering global graph optimization to correct the dense grid map of the environment and the real-time pose.

4. The autonomous navigation, positioning, and drilling method for a tracked drilling robot according to claim 1, characterized in that, The total cost function of the cost map constructed in the BIM constrained path planning step is: in: To plan a specific coordinate position for the robot in space; Let the barrier cost function be used. The cost function for BIM structural constraints; The weighting coefficient for the obstacle cost; The constraint cost weighting coefficient; the weighting coefficient is taken as follows: =0.7, =0.

3.

5. The autonomous navigation, positioning, and drilling method for a tracked drilling robot according to claim 4, characterized in that, The BIM-constrained path planning step specifically includes: using the A* algorithm to search for a discrete global path from the robot's current position to the target drilling point on the cost map; in the execution phase, using the Dynamic Window Method (DWA) combined with real-time obstacle point clouds to generate feasible local trajectories; and performing cubic spline interpolation on the local trajectory to obtain continuous speed control commands. The motion controller of the tracked chassis is sent to the motion controller, where: Instantaneous linear velocity, It represents the instantaneous angular velocity.

6. The autonomous navigation, positioning, and drilling method for a tracked drilling robot according to claim 1, characterized in that, The positioning steps of the parallel mechanism specifically include: Based on the target drilling points marked in the BIM and the current base posture of the robot, the target end-effector pose of the parallel mechanism is obtained by combining them; based on the kinematic model of the parallel mechanism, the Levenberg-Marquardt algorithm is used to solve for the iteratively updated joint angle vectors as follows: in: For the first k Joint angle vector at the next iteration; For the first k Jacobian matrix at the next iteration; For the first k Damping factor at the next iteration; It is the identity matrix; It is a positive kinematic function; The desired end-target pose.

7. The autonomous navigation, positioning, and drilling method for a tracked drilling robot according to claim 1, characterized in that, The vision-force closed-loop drilling process specifically includes: The coordinates of the actual borehole points are obtained by performing Hough line transform detection on the borehole markings using a depth camera. ; Calculate the coordinates of the actual borehole point and the coordinates of the target borehole point. Position error: in: actual borehole coordinates The coordinates of the target borehole point are output by inverse kinematics; The error is adjusted by the PID controller to regulate the joint commands of the parallel mechanism, achieving sub-millimeter-level positioning; Establish a pressure sensor model This refers to the drill bit feed rate. , For experimental calibration constants; when The system will automatically slow down or stop. Establish a torque sensor model: For cutting resistance, Where is the drill bit radius; when If the sudden rise exceeds 12 N·m, it is determined to be a steel bar collision, triggering a retraction and repositioning.

8. The autonomous navigation, positioning, and drilling method for a tracked drilling robot according to claim 1, characterized in that, The global error correction step specifically includes: after completing one drilling operation, performing ICP registration again on the currently acquired third point cloud data and the BIM structural feature library, and minimizing the objective function: Based on ICP, a normal distribution transformation NDT is introduced to form the cost function: in: It is a rotation matrix; It is a translation vector; The first point cloud in the source cloud i One point; For the target point cloud i One corresponding point; The first point cloud in the source cloud j One point; The mean vector of the target raster is used; the optimal pose correction value is obtained by solving the Levenberg-Marquardt iterative method. The optimal pose correction value As the observation input EKF, it completes the smooth update of global pose and controls the cumulative error within 5mm.

9. The autonomous navigation, positioning, and drilling method for a tracked drilling robot according to claim 1, characterized in that, It also includes a task completion determination step: the system counts the number of boreholes that have been drilled. Compared with the preset total number of boreholes in BIM ;like When the operation is completed, a safety shutdown procedure is executed, including shutting down the drill motor, stopping the track drive, and disconnecting the power supply to the lidar and depth camera; the point cloud data, pose data, and force / torque logs of this operation are saved to the local server, and the control system is reset to standby mode.

10. A computer program product, characterized in that, The system includes a computer program that, when executed by a processor, implements the autonomous navigation, positioning, and drilling operation method for a tracked drilling robot as described in any one of claims 1 to 9.

Images