Automobile hoisting device and method

Abstract

The invention discloses an automobile hoisting device, and belongs to the technical field of engineering hoisting. The problems that the gravity center of a component is difficult to accurately determine, the position and posture control of a cargo boom is inaccurate, and the load cannot be adaptively adjusted exist in the hoisting process of an existing automobile crane. The device comprises a truck crane body, a cargo boom tail end execution module, a cargo boom body sensing module, an environment sensing module and a control module. The control module is internally provided with a dynamic gravity center correction unit, a multi-sensor fusion unit and a self-adaptive load control unit, processes related sensor data based on Kalman filtering, extended particle filtering and a model prediction control algorithm, and outputs an adjustment instruction. The method is mainly used for improving the accuracy, safety and efficiency of automobile hoisting operation.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering hoisting. More specifically, the present invention relates to an automobile hoisting device and method. Background Art

[0002] In the field of engineering construction, automobile hoisting operation is a common and crucial task, which is widely used in many scenarios such as building construction, bridge erection, equipment installation, etc. However, there are still many problems to be solved in the current automobile hoisting technology.

[0003] Firstly, in terms of determining the center of gravity of the component to be hoisted, traditional methods often rely on experience and simple calculations, and it is difficult to deal with components with complex shapes and irregular mass distributions. Due to the lack of accurate center of gravity data, the component is prone to tilt, shake or even lose balance during the hoisting process, which not only affects the hoisting efficiency but also poses a great safety hazard. For example, for some large special-shaped components with complex internal structures and uneven mass distributions, it is very difficult to accurately determine their center of gravity positions using traditional methods. Moreover, during the hoisting process, the attitude and force conditions of the component will constantly change, further increasing the difficulty of determining the center of gravity.

[0004] Secondly, the pose control of the boom is not precise enough. The existing hoisting equipment mainly relies on the experience of the operator to adjust the angle and position of the boom, and this method is greatly affected by human factors and is prone to errors. In some projects with high requirements for hoisting accuracy, such as the installation of high-precision equipment, a small pose deviation may lead to serious consequences. In addition, due to the lack of real-time and accurate environmental information feedback, it is difficult for the operator to make correct adjustments in time when encountering obstacles, increasing the probability of collision accidents. In actual operations, the construction site environment is complex and changeable, and new obstacles or personnel activities may occur, resulting in the inability to execute the pre-planned path normally. Moreover, due to the lack of effective dynamic obstacle detection and prediction technologies, it is very difficult to take effective avoidance measures in time when encountering moving obstacles.

[0005] In summary, there are many problems in the existing automobile hoisting technology in aspects such as determining the center of gravity of components, pose control of the boom, path planning, etc. These problems affect the safety and efficiency of hoisting operations. Due to the complexity and multi-disciplinary nature of the technology, a comprehensive and effective solution has not been found yet. Summary of the Invention

[0006] The present invention provides an automobile lifting and hoisting device and method, which can obtain the center of gravity information of the component to be hoisted, the position and posture information of the crane arm and the load condition in real time and accurately, and dynamically correct the center of gravity to reduce the risk of component tilting and shaking. Multi-sensor fusion makes the position and posture control of the crane arm more accurate, and adjusts the movement speed of the crane arm in real time to avoid the occurrence of overload, thereby improving the safety and efficiency of the lifting operation.

[0007] In order to achieve these purposes and other advantages according to the present invention, a vehicle lifting and hoisting device is provided, comprising: The main body of the truck crane, which is equipped with a lifting arm and a steel wire rope; The end-of-arm execution module includes a three-dimensional laser scanner, a six-axis force / torque sensor, and a flexible pressure sensor, wherein the scanning axis of the three-dimensional laser scanner coincides with the axis of the wire rope; A boom body sensing module, which includes a fiber grating sensor array and an inertial measurement unit. The fiber grating sensor array is distributed along the length direction of the boom, and a wavelength demodulation module of the fiber grating sensor array is synchronized with the inertial measurement unit signal; An environment perception module, which includes a laser radar and a camera, wherein the scanning plane of the laser radar is parallel to the rotation plane of the crane arm; The control module has built-in dynamic center of gravity correction unit, multi-sensor fusion unit and adaptive load control unit: The dynamic center of gravity correction unit receives the point cloud data of the component to be hoisted output by the three-dimensional laser scanner and the three-axis force data output by the six-axis force / torque sensor, calculates the center of gravity offset of the component to be hoisted based on the Kalman filter algorithm and outputs a hydraulic proportional valve adjustment instruction; The multi-sensor fusion unit receives strain data from the fiber grating sensor array, pitch angle data from the inertial measurement unit, obstacle coordinate data from the laser radar, and visual feature point data from the camera, generates a boom end posture matrix based on an extended particle filter algorithm, and outputs a posture control instruction; The adaptive load control unit receives strain data of the fiber grating sensor array and weight data of the component to be hoisted, and outputs a crane arm motion speed limit instruction based on a model predictive control algorithm.

[0008] Preferably, based on the automobile lifting and hoisting device, the method comprises the following steps: S1: Scan the work area with a laser radar, collect images with a camera, and use a checkerboard calibration plate to spatially align the laser radar point cloud coordinate system with the camera image coordinate system. Output the coordinate transformation matrix to generate a three-dimensional environment map and perform collision-free path planning. S2: Scan the surface of the component to be hoisted by the 3D laser scanner at the end of the crane arm, and perform feature matching between the point cloud data and the 3D coordinates of the preset hoisting points in the BIM model. When the matching error exceeds the preset threshold, the coordinate correction of the hoisting point is triggered; S3: Control the crane hook to dock with the corrected lifting point, collect three-axis force data in real time through the six-axis force / torque sensor, iteratively calculate the center of gravity position of the component to be lifted based on the Kalman filter algorithm and output the hydraulic proportional valve adjustment command; S4: synchronously collects strain data from the fiber Bragg grating sensor array, pitch angle data from the inertial measurement unit, obstacle distance data from the lidar, and visual tracking data from the camera, generates a boom end posture matrix based on the extended particle filter algorithm, and outputs posture control instructions; S5: Monitor the strain data of the fiber grating sensor array in real time. When the strain value reaches the preset safety threshold, the crane arm movement speed limit instruction is output based on the model predictive control algorithm.

[0009] Preferably, S2 specifically includes: Extract the characteristic edge point set of the components to be hoisted in the BIM model, and align it with the point cloud data scanned by the 3D laser scanner through the ICP algorithm; Calculate the root mean square error after registration and remove abnormal points, retain the matching point cloud for surface reconstruction and triangulation processing, and generate a closed 3D solid model; Align the closed 3D solid model with the crane arm base coordinate system through the least squares rigid transformation, and output the initial hoisting point coordinates in the world coordinate system; The surface point cloud data of the component to be hoisted is obtained by a 3D laser scanner, which is aligned with the coordinates of the initial hoisting point to calculate the coordinate deviation; When the coordinate deviation exceeds a preset threshold, a candidate set of alternate lifting points is generated based on the geometric feature points of the component to be lifted, and the candidate set is calculated by a convex hull algorithm; The force on the candidate lifting points is tested by a six-axis force / torque sensor to select the optimal lifting point.

[0010] Preferably, the three-axis force components of the candidate lifting points are tested in sequence by a six-axis force / torque sensor, and the lifting point with the smallest angle between the resultant force direction and the vertical line of the theoretical center of gravity is selected; The standard deviation of pressure distribution on the contact surface is monitored by a flexible pressure sensor, and the secondary correction process is triggered when the standard deviation exceeds a preset threshold.

[0011] Preferably, S4 specifically includes: The hardware clocks of the lidar, inertial measurement unit, and camera are synchronized through GNSS pulse-per-second signals, and the sampling data of the six-axis force / torque sensor and the fiber Bragg grating sensor are aligned by linear interpolation. Based on the extended particle filter algorithm, the laser radar obstacle distance, IMU attitude angle, camera visual feature points and fiber Bragg grating strain data are integrated to generate a six-degree-of-freedom posture matrix; The posture control instructions are output based on the posture matrix to adjust the boom luffing angle.

[0012] Preferably, S5 specifically includes: A linear mapping model between the wavelength offset of the fiber Bragg grating sensor and the bending moment at the root of the crane arm was established; When the real-time strain value reaches the first-level threshold, the model predictive control algorithm reduces the amplitude change speed and adjusts the pitch angle to optimize the bending moment distribution; When the real-time strain value reaches the second-level threshold, the hydraulic system lock command is immediately triggered and the sound and light alarm device is activated; The rolling optimization window of the model predictive control algorithm is multiple control cycles in the future, and each optimization time does not exceed 200ms.

[0013] Preferably, S1 further comprises: Based on the 3D environment map, a collision-free path is planned with the size of the component to be hoisted expanded by 0.8m as the safety envelope constraint; The laser radar scans the working area at a frequency of 10 Hz to detect dynamic obstacles, extract the coordinates of the obstacle's center of mass, and predict its motion trajectory in the next 2 seconds through the extended Kalman filter; The predicted trajectory is collided with the current lifting path. When the intersection of the paths is less than 1.2m from the end of the boom, a fast expansion random tree algorithm is used to generate a local obstacle avoidance path to ensure that the real-time distance to the dynamic obstacle is not less than 0.5m.

[0014] The present invention has at least the following beneficial effects: The present invention uses multi-sensor fusion technology to achieve accurate perception of the working environment and construction of a three-dimensional map, and combines intelligent algorithms to plan collision-free paths to improve hoisting safety. The center of gravity is calculated in real time through hoisting point coordinate correction and Kalman filtering to ensure component balance and reduce the risk of shaking. The extended particle filter fuses multi-source data to generate an accurate posture matrix to achieve refined control of the crane arm movement; the model predictive control algorithm dynamically adjusts the load to avoid overload and ensure equipment safety. This method significantly improves the stability and efficiency of hoisting operations through intelligent and automated control, reduces human errors, and is suitable for complex working conditions.

[0015] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. DETAILED DESCRIPTION

[0016] The present invention will be further described in detail below in conjunction with the details, so that those skilled in the art can implement it with reference to the text of the specification.

[0017] It should be understood that terms such as "having", "including", and "comprising" used herein do not exclude the presence or addition of one or more other elements or combinations thereof.

[0018] It should be noted that the experimental methods described in the following embodiments are all conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial channels unless otherwise specified. Therefore, it should not be construed as a limitation to the present invention.

[0019] The present invention provides a vehicle-mounted lifting and hoisting device, including: A vehicle-mounted crane body, which is equipped with a boom and a steel wire rope. The boom is used to lift and move the component to be hoisted, and the steel wire rope connects the hook and the boom to achieve force transmission; An end effector module of the boom, which includes a three-dimensional laser scanner, a six-axis force / torque sensor, and a flexible pressure sensor. The scanning axis of the three-dimensional laser scanner coincides with the axis of the steel wire rope. The three-dimensional laser scanner is installed directly below the end of the boom to obtain the three-dimensional point cloud data of the component to be hoisted. The six-axis force / torque sensor is installed at the connection between the hook and the hoisting steel wire rope to measure the three-axis force (forces in the X, Y, and Z directions) and three-axis torque (torques around the X, Y, and Z axes) data borne by the hook in real time, reflecting the force condition at the hoisting point. The flexible pressure sensors are distributed on the contact surface between the hoisting fixture and the component to be hoisted to detect the contact pressure distribution between the fixture and the component in real time; A sensing module of the boom body, which includes a fiber Bragg grating sensor array and an inertial measurement unit. The fiber Bragg grating sensor array is distributed along the length of the boom, and a sensor is installed at a certain interval (0.5 - 1 meter) to monitor the strain condition of each part of the boom in real time, convert the strain data into optical signals and transmit them to the wavelength demodulation module. By analyzing the strain data, the force state of the boom can be understood. The inertial measurement unit is installed at the root of the boom to measure the attitude information of the boom in real time, including pitch angle, roll angle, yaw angle, etc. The wavelength demodulation module of the fiber Bragg grating sensor array is synchronized with the signal of the inertial measurement unit to ensure the temporal consistency of the data of each sensor and provide an accurate data basis for multi-sensor fusion; An environment perception module, which includes a lidar and a camera. The lidar is installed in the middle of the boom. The scanning plane of the lidar is parallel to the slewing plane of the boom to scan the surrounding environment and obtain the three-dimensional coordinate data of obstacles, providing accurate obstacle information for path planning by monitoring the environmental changes in real time. The camera is installed at the top of the boom, and the field of view covers the hoisting operation area to collect visual images in real time and extract the visual feature point data of the component and obstacles; The control module is built-in with a dynamic center-of-gravity correction unit, a multi-sensor fusion unit, and an adaptive load control unit: The dynamic center-of-gravity correction unit receives the point cloud data of the component to be lifted output by the 3D laser scanner and the three-axis force data output by the six-axis force / torque sensor, calculates the center-of-gravity offset of the component to be lifted based on the Kalman filter algorithm, and outputs a hydraulic proportional valve adjustment instruction; given the lifting point information of the component to be lifted, the dynamic center-of-gravity correction unit receives the component point cloud data output by the 3D laser scanner and the three-axis force data output by the six-axis force / torque sensor. First, a 3D model of the component is constructed according to the component point cloud data, and combined with the preset component material density information, the theoretical center-of-gravity position of the component is initially calculated. Then, the force condition of the lifting point is monitored in real time through the six-axis force / torque sensor, and the Kalman filter algorithm is used to dynamically correct the theoretical center-of-gravity position to calculate the actual offset of the center of gravity. According to the center-of-gravity offset, the control module adjusts the luffing angle of the boom through the hydraulic proportional valve to keep the component balanced during the lifting process. For example, when the center of gravity shifts to one side, the luffing angle of the boom on that side is appropriately increased to adjust the tension of the sling and make the component regain balance; The multi-sensor fusion unit receives the strain data of the fiber Bragg grating sensor array, the pitch angle data of the inertial measurement unit, the obstacle coordinate data of the lidar, and the visual feature point data of the camera. The strain data can reflect the deformation of the boom, thus affecting its end position. The pitch angle data directly provides the attitude information of the boom. The obstacle coordinate data and the visual feature point data are used to constrain the pose of the boom to avoid collision with obstacles. The time alignment of all sensor data can be ensured by combining hardware synchronization (such as GNSS second pulse signal) and software synchronization (such as linear interpolation). Based on the extended particle filter algorithm, a pose matrix of the boom end is generated and a pose control instruction is output; by continuously sampling, updating weights, and resampling the particles, the optimal estimate of the boom end pose is finally obtained, generating a six-degree-of-freedom pose matrix. According to the generated pose matrix, a pose control instruction is output to control the luffing, telescoping, and slewing actions of the boom to make the boom end reach the desired pose; The adaptive load control unit receives the strain data of the fiber Bragg grating sensor array and the weight data of the component to be lifted. By real-time monitoring the wavelength shift of the fiber Bragg grating sensor, it calculates the real-time bending moment at the root of the boom. Based on the weight data of the component to be lifted, it evaluates the load state of the boom. Based on the model predictive control algorithm, it outputs the speed limit instruction for the boom movement. Taking multiple future control cycles as the rolling optimization window, it predicts the load change of the boom under different control inputs. If it predicts that the load of the boom will exceed the safety threshold, the model predictive control algorithm will calculate the optimal control input, such as reducing the luffing speed of the boom, adjusting the pitch angle, etc., to optimize the bending moment distribution and ensure that the load of the boom is within the safe range.

[0020] In the above technical solution, the truck-mounted crane lifting device can obtain the centroid information of the component to be lifted, the pose information of the boom, and the load condition in real time and accurately through the collaborative work of multiple sensors. The dynamic centroid correction unit improves the accuracy of centroid determination and reduces the risk of component tilt and sway based on the Kalman filter algorithm. The multi-sensor fusion unit generates an accurate pose matrix using the extended particle filter algorithm, making the pose control of the boom more precise. The adaptive load control unit adjusts the boom movement speed in real time according to the actual load condition, avoiding the occurrence of overload phenomena and improving the safety and efficiency of the lifting operation.

[0021] A truck-mounted crane lifting method, based on the above truck-mounted crane lifting device, the method includes the following steps: S1: Scan the operation area with a lidar, collect images with a camera, perform spatial registration on the lidar point cloud coordinate system and the camera image coordinate system through a checkerboard calibration board, output a coordinate transformation matrix to generate a three-dimensional environment map, and perform collision-free path planning.

[0022] Using the checkerboard calibration board as a reference, align the lidar and the camera with the calibration board at the same time, collect multiple groups of images and point cloud data, perform internal parameter calibration on the camera through the Zhang's calibration method to obtain parameters such as focal length and distortion coefficient. At the same time, use the lidar to scan the geometric features of the calibration board, calculate the external parameter matrix (rotation matrix and translation vector) between the lidar coordinate system and the camera image coordinate system, output the coordinate transformation matrix, and map the lidar point cloud data to the camera image coordinate system to achieve spatial alignment of multi-sensor data.

[0023] The lidar continuously scans the operation area at a frequency of 10Hz to generate point cloud data containing obstacles (such as buildings, stacked equipment). The camera synchronously collects images and identifies the obstacle types (such as static obstacles, moving vehicles) through semantic segmentation algorithms. Integrate the lidar point cloud and the camera image to construct a three-dimensional environment map with an accuracy of ±5cm, and mark the position, size, and motion state of the obstacles.

[0024] Based on the 3D environmental map, taking the size of the component to be lifted and expanding it by 0.8 m as the safety envelope constraint, the RRT algorithm is used to plan the initial lifting path. When planning the path, consider the movement range of the boom, the luffing angle limit and the stability of the component's center of gravity to ensure that the path is continuous and collision-free.

[0025] S2: Scan the surface of the component to be lifted through the 3D laser scanner at the end of the boom, perform feature matching on the point cloud data with the 3D coordinates of the preset lifting points in the BIM model, and trigger the correction of the lifting point coordinates when the matching error exceeds the preset threshold.

[0026] The 3D laser scanner at the end of the boom scans the component to be lifted 360°, generating high-density point cloud data. Extract the characteristic edge point set (such as edges, corners) of the component to be lifted from the BIM model, register the scanned point cloud with the BIM point cloud through the Iterative Closest Point (ICP) algorithm, calculate the Root Mean Square (RMS) error after registration. When the RMS exceeds the preset threshold (such as 5 mm), it is determined that there is a deviation in the lifting point, and the correction process is triggered.

[0027] S3: Control the hook of the boom to dock with the corrected lifting point, collect the three-axis force data in real time through the six-axis force / torque sensor, and iteratively calculate the position of the center of gravity of the component to be lifted based on the Kalman filter algorithm and output the adjustment instruction for the hydraulic proportional valve.

[0028] Based on the geometric parameters and material density of the point cloud model, calculate the initial theoretical center of gravity of the component. Using the Kalman filter algorithm, take the theoretical center of gravity as the initial state value, and perform iterative update in combination with the real-time force data. The state equation describes the movement trend of the center of gravity, and the measurement equation fuses the uncertainties of the point cloud model and the force data. The precise center of gravity position is obtained through recursive update, with the error controlled within ±5 cm. According to the offset of the center of gravity, calculate the luffing angle that the boom needs to adjust. For example, if the center of gravity shifts 10 cm to the left, calculate that the boom needs to tilt 2° to the right through geometric relationships to balance the moment. Output the control signal of the hydraulic proportional valve to adjust the telescopic speed of the boom cylinder, and correct the boom angle in real time to make the component regain balance.

[0029] S4: Synchronously collect the strain data of the fiber Bragg grating sensor array, the pitch angle data of the inertial measurement unit, the obstacle distance data of the lidar, and the visual tracking data of the camera, and generate the pose matrix of the end of the boom based on the extended particle filter algorithm and output the pose control instruction.

[0030] Synchronize the hardware clocks of the lidar, inertial measurement unit, and camera through the GNSS second pulse signal to ensure consistent data timestamps. Linearly interpolate and align the asynchronous sampling data of the six-axis force / torque sensor and fiber Bragg grating sensor to eliminate time deviation. The extended particle filter algorithm fuses the strain data of the fiber Bragg grating (reflecting the deformation of the boom), the pitch angle data of the inertial measurement unit (the attitude of the boom), the obstacle distance data of the lidar, and the visual feature point data of the camera to generate a six-degree-of-freedom pose matrix. According to the pose matrix, calculate the deviation between the current position and the target position of the boom tip. For example, if the pose matrix shows that the boom tip needs to move up 5 cm and rotate left 1°, then output an instruction to control the hydraulic system to adjust the telescoping and luffing actions of the boom.

[0031] S5: Real-time monitor the strain data of the fiber Bragg grating sensor array. The fiber Bragg grating sensor monitors the strain of each part of the boom in real time, converts it into an electrical signal through the wavelength demodulation module, and establishes a linear mapping model between the wavelength shift of the fiber Bragg grating and the bending moment at the root of the boom (such as M = kΔλ + b), and calculate the stress state of the boom in real time. The rolling optimization window of the model predictive control algorithm is the next 3 control cycles, and each optimization time does not exceed 200 ms. Predict the strain change trend of the boom at different action speeds. When the strain value reaches the preset safety threshold, based on the model predictive control algorithm, output an instruction to limit the action speed of the boom. For example, reduce the luffing speed of the boom by 30% and fine-tune the pitch angle to optimize the bending moment distribution.

[0032] In the above technical solution, the truck crane lifting method is based on multi-sensor fusion technology, accurately perceives the operating environment, generates a three-dimensional environment map and plans a collision-free path, improving the safety of the lifting operation. By correcting the coordinates of the lifting point and real-time adjusting the center of gravity of the component, the pose of the boom, and the load control, the accuracy and stability of the lifting process are ensured, the influence of human factors is reduced, the lifting efficiency is improved, and the probability of safety accidents is reduced.

[0033] In another technical solution, S2 specifically includes: Use the edge detection algorithm to extract the characteristic edge point set of the component to be lifted in the BIM model, which contains the key structural features of the component, such as corners, edge intersections, etc., and register it with the point cloud data scanned by the three-dimensional lidar through the ICP algorithm. Specifically: ① Initial alignment: Reduce the position difference between the point clouds through rough manual or automatic positioning. ② Nearest point search: Find the nearest neighbor point in the scanned point cloud for each point in the BIM point set. ③ Rigid transformation calculation: Use the least squares method to solve the rotation matrix and translation vector to minimize the root mean square error (RMS) of the two sets of points. ④ Iterative optimization: Repeat steps ② and ③ until the RMS converges or reaches the preset number of iterations.

[0034] Calculate the root mean square error after registration, eliminate the outlier points, and retain the matching point cloud for surface reconstruction and triangular meshing. Specifically: Use the Poisson surface reconstruction algorithm on the retained point cloud to obtain a smooth closed three-dimensional solid model, and use the Delaunay triangulation algorithm to mesh the model surface to generate triangular patches and generate a closed three-dimensional solid model.

[0035] Align the closed three-dimensional solid model with the boom base coordinate system through the least squares rigid transformation. By selecting corresponding feature points (such as the reference point at the root of the boom) in the model and the base coordinate system, use the least squares method to solve the rigid transformation parameters to ensure that the origin and coordinate axis directions of the model coordinate system and the base coordinate system are the same. Extract and output the initial hoisting point coordinates in the world coordinate system according to the preset hoisting point positions in the BIM model.

[0036] Obtain the surface point cloud data of the component to be hoisted through a three-dimensional laser scanner, register it with the initial hoisting point coordinates, and calculate the coordinate deviation by calculating the spatial deviation between the point cloud and the coordinates.

[0037] When the coordinate deviation exceeds the preset threshold, based on the geometric feature points (such as the centroid and endpoints) of the component to be hoisted, use the convex hull algorithm to generate a candidate set of alternative hoisting points. The candidate set is calculated by the convex hull algorithm, which calculates the minimum convex polygon of the point set to ensure that the candidate points are distributed in the hoistable area of the component and avoid stress concentration.

[0038] Test the forces on the candidate hoisting points through a six-axis force / torque sensor, the three-axis force (in the X, Y, and Z directions) and torque data, and select the optimal hoisting point.

[0039] In the above technical solution, the hoisting point correction is achieved by registering and processing the BIM model and the actual scanned point cloud data, which can accurately determine the initial hoisting point coordinates and make corrections according to the actual situation. Generating a candidate set of alternative hoisting points and selecting the optimal hoisting point improves the accuracy and reliability of the hoisting point, ensures the stability of the component during hoisting, and reduces the safety hazards caused by inaccurate hoisting points.

[0040] In another technical solution, sequentially test the three-axis force components of the candidate hoisting points through a six-axis force / torque sensor, collect the three-axis force data, calculate the magnitude and direction of the resultant force of each candidate point, and evaluate the angle between it and the perpendicular line of the theoretical center of gravity of the component. Select the hoisting point with the smallest angle between the resultant force direction and the perpendicular line of the theoretical center of gravity as the preliminary correction point; Combine a flexible pressure sensor to monitor the standard deviation of the pressure distribution on the contact surface. When the standard deviation is within the preset threshold (such as 10 N / cm²), it is determined as the hoisting point. When the standard deviation exceeds the preset threshold, trigger a secondary correction process to adjust the hook attitude or replace the candidate point until the pressure distribution is uniform.

[0041] In the above technical solution, the hoisting point selection and correction method comprehensively considers the force direction of the candidate hoisting points and the pressure distribution on the contact surface. The hoisting point with the smallest included angle between the resultant force direction and the perpendicular line of the theoretical centroid is selected, ensuring the rationality of the component force. At the same time, the standard deviation of the pressure distribution is monitored and the secondary correction process is triggered when it exceeds the threshold, ensuring uniform force on the component during hoisting, avoiding local damage, and improving the service life of the component and the safety of hoisting.

[0042] In another technical solution, S4 specifically includes: Synchronize the hardware clocks of the lidar, inertial measurement unit, and camera through the GNSS second pulse signal, and linearly interpolate and align the sampling data of the six-axis force / torque sensor and the fiber Bragg grating sensor; Based on the extended particle filter algorithm, fuse the lidar obstacle distance, IMU attitude angle, camera visual feature points, and fiber Bragg grating strain data to generate a six-degree-of-freedom pose matrix; Based on the pose matrix, output pose control instructions to adjust the boom amplitude angle.

[0043] Specifically, define the six-degree-of-freedom pose state vector , which includes the three-dimensional coordinates and Euler angles of the boom tip.

[0044] Lidar: Match the point cloud features through the ICP algorithm, extract the obstacle distance and azimuth angle, and construct the observation equation .

[0045] Inertial measurement unit: Use the integration of acceleration and angular velocity to predict the pose change and construct the state transition equation .

[0046] Camera: Obtain visual feature points through SIFT feature matching and establish the projection equation .

[0047] Fiber Bragg grating: Based on the strain data, establish a boom deformation model to compensate for the pose estimation error.

[0048] Generate an initial particle set based on the initial attitude of the inertial measurement unit and the lidar point cloud rough positioning. Update the particle pose according to the inertial measurement unit data and add process noise. Calculate the particle weights using the lidar, camera, and fiber Bragg grating data, and resample to retain the high-weight particles. Control the pose error after fusion.

[0049] Convert the six-degree-of-freedom pose matrix into the telescopic amount and slewing angle of the boom cylinder, and calculate the target angle and the pitch angle .

[0050] Adjust the opening of the hydraulic proportional valve in real time to make the end of the boom track the target pose.

[0051] In the above technical solution, the method for generating the pose matrix at the end of the boom solves the problem of inconsistent data by synchronizing and fusing multi-sensor data, improving the accuracy of the pose matrix. Based on the accurate pose matrix, the luffing angle of the boom is adjusted, making the pose control of the boom more precise, better adapting to complex lifting environments, and improving the accuracy and efficiency of lifting operations.

[0052] In another technical solution, S5 specifically includes: Establish a linear mapping model M = kΔλ + b between the wavelength offset Δλ of the fiber Bragg grating sensor and the bending moment M at the root of the boom, where k is the slope and b is the intercept. By fitting the preprocessed data collected, calculate the values of k and b to determine the linear mapping model; When the real-time strain value reaches the first-level threshold, the model predictive control algorithm reduces the luffing speed and adjusts the pitch angle to optimize the bending moment distribution. For example, reduce the luffing speed to 0.5 m / s and reduce the standard deviation of the bending moment distribution by 20% by adjusting the pitch angle; When the real-time strain value reaches the second-level threshold, immediately trigger the hydraulic system locking instruction and activate the audible and visual alarm device; The rolling optimization window of the model predictive control algorithm is multiple future control cycles, and each optimization time does not exceed 200 ms.

[0053] In the above technical solution, the load control method can monitor the strain condition of the boom in real time by establishing a linear mapping model between the wavelength offset of the fiber Bragg grating sensor and the bending moment at the root of the boom. According to different strain thresholds, corresponding measures are taken, such as reducing the luffing speed, adjusting the pitch angle, or locking the hydraulic system and alarming, effectively avoiding boom overload, extending the service life of the equipment, and improving the safety of lifting operations.

[0054] In another technical solution, S1 also includes: Based on the three-dimensional environmental map, with a safety envelope constraint of expanding 0.8 m outside the dimensions of the component to be lifted, plan a collision-free path. The path nodes include parameters such as the boom angle and telescopic length, meeting mechanical constraints such as the maximum elevation angle of the boom being 75° and the minimum working radius being 5 m; Scan the operation area at a frequency of 10 Hz with a lidar for dynamic obstacle detection, establish a motion model of the dynamic obstacle, and the state vector includes position (x, y, z), velocity (v x , v y , v z ), and acceleration (a x , a y , a z), filter the centroid coordinates of the obstacle, extract the centroid coordinates of the obstacle and predict its future 2-second movement trajectory through extended Kalman filtering. The process noise covariance matrix is dynamically adjusted according to the obstacle type (pedestrian / vehicle). The prediction confidence of the vehicle is set to 0.95, and that of the pedestrian is set to 0.85; Discretize the predicted obstacle trajectory into time series points at 100ms intervals. Perform collision detection between the predicted trajectory and the current hoisting path. When the distance from the path intersection point to the end of the boom is less than 1.2m, use the current position as the root node, randomly sample new nodes under the constraint of the safety envelope surface, and use the rapidly-exploring random tree algorithm to generate a local obstacle avoidance path. The path smoothing uses quintic polynomial interpolation to ensure the continuity of the boom movement acceleration, and the maximum angular velocity is limited to 5° / s to ensure that the real-time distance from the dynamic obstacle is not less than 0.5m. The lidar monitors the path area in real time. If an unpredicted obstacle (such as a suddenly crossing pedestrian) is detected, immediately trigger emergency braking. The laser displacement sensor installed at the end of the boom feeds back the distance from the obstacle in real time, and automatically fine-tunes the path when the error exceeds ±3cm.

[0055] In the above technical solution, the hoisting path planning method takes into account the influence of dynamic obstacles on the basis of generating a three-dimensional environment map. By detecting the movement trajectory of dynamic obstacles in real time and generating a local obstacle avoidance path according to the collision detection results, the safety and reliability of the hoisting path are ensured. It can respond to the complex changes on the construction site in a timely manner, avoid the occurrence of collision accidents, and improve the efficiency and safety of hoisting operations.

[0056] The number of devices and the processing scale described here are used to simplify the description of the present invention. The applications, modifications, and variations of the present invention will be obvious to those skilled in the art.

[0057] Although the embodiments of the present invention have been disclosed as above, it is not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the field, additional modifications can be easily achieved. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and the examples shown and described here.

Claims

1. Automobile lifting and hoisting device, characterized in that: include: The main body of the truck crane, which is equipped with a lifting arm and a steel wire rope; The end-of-arm execution module includes a three-dimensional laser scanner, a six-axis force / torque sensor, and a flexible pressure sensor, wherein the scanning axis of the three-dimensional laser scanner coincides with the axis of the wire rope; A boom body sensing module, which includes a fiber grating sensor array and an inertial measurement unit. The fiber grating sensor array is distributed along the length direction of the boom, and a wavelength demodulation module of the fiber grating sensor array is synchronized with the inertial measurement unit signal; An environment perception module, which includes a laser radar and a camera, wherein the scanning plane of the laser radar is parallel to the rotation plane of the crane arm; The control module has built-in dynamic center of gravity correction unit, multi-sensor fusion unit and adaptive load control unit: The dynamic center of gravity correction unit receives the point cloud data of the component to be hoisted output by the three-dimensional laser scanner and the three-axis force data output by the six-axis force / torque sensor, calculates the center of gravity offset of the component to be hoisted based on the Kalman filter algorithm and outputs a hydraulic proportional valve adjustment instruction; The multi-sensor fusion unit receives strain data from the fiber grating sensor array, pitch angle data from the inertial measurement unit, obstacle coordinate data from the laser radar, and visual feature point data from the camera, generates a boom end posture matrix based on an extended particle filter algorithm, and outputs a posture control instruction; The adaptive load control unit receives strain data of the fiber grating sensor array and weight data of the component to be hoisted, and outputs a crane arm motion speed limit instruction based on a model predictive control algorithm.

2. A method for lifting and hoisting a vehicle, characterized in that: Based on the automobile lifting and hoisting device according to claim 1, the method comprises the following steps: S1: Scan the work area with a laser radar, collect images with a camera, and use a checkerboard calibration plate to spatially align the laser radar point cloud coordinate system with the camera image coordinate system. Output the coordinate transformation matrix to generate a three-dimensional environment map and perform collision-free path planning. S2: Scan the surface of the component to be hoisted by the 3D laser scanner at the end of the crane arm, perform feature matching between the point cloud data and the 3D coordinates of the preset hoisting points in the BIM model, and trigger the correction of the hoisting point coordinates when the matching error exceeds the preset threshold; S3: Control the hook of the crane arm to dock with the corrected lifting point, collect three-axis force data in real time through the six-axis force / torque sensor, iteratively calculate the center of gravity position of the component to be lifted based on the Kalman filter algorithm and output the adjustment command of the hydraulic proportional valve; S4: synchronously collect strain data from the fiber Bragg grating sensor array, pitch angle data from the inertial measurement unit, obstacle distance data from the laser radar, and visual tracking data from the camera, generate a boom end posture matrix based on the extended particle filter algorithm, and output posture control instructions; S5: Monitor the strain data of the fiber grating sensor array in real time. When the strain value reaches the preset safety threshold, the crane arm movement speed limit instruction is output based on the model predictive control algorithm.

3. The method for lifting and hoisting a vehicle as claimed in claim 2, characterized in that: S2 specifically includes: Extract the characteristic edge point set of the components to be hoisted in the BIM model, and align it with the point cloud data scanned by the 3D laser scanner through the ICP algorithm; Calculate the root mean square error after registration and remove abnormal points, retain the matching point cloud for surface reconstruction and triangulation processing, and generate a closed 3D solid model; Align the closed 3D solid model with the crane arm base coordinate system through the least squares rigid transformation, and output the initial hoisting point coordinates in the world coordinate system; The surface point cloud data of the component to be hoisted is obtained by a 3D laser scanner, which is aligned with the coordinates of the initial hoisting point to calculate the coordinate deviation; When the coordinate deviation exceeds a preset threshold, a candidate set of alternate lifting points is generated based on the geometric feature points of the component to be lifted, and the candidate set is calculated by a convex hull algorithm; The force on the candidate lifting points is tested by a six-axis force / torque sensor to select the optimal lifting point.

4. The method for lifting and hoisting a vehicle as claimed in claim 3, characterized in that: The three-axis force components of the candidate lifting points are tested in sequence by using a six-axis force / torque sensor, and the lifting point with the smallest angle between the resultant force direction and the vertical line of the theoretical center of gravity is selected; The standard deviation of pressure distribution on the contact surface is monitored in combination with a flexible pressure sensor, and the secondary correction process is triggered when the standard deviation exceeds a preset threshold.

5. The method for lifting and hoisting a vehicle as claimed in claim 2, characterized in that: S4 specifically includes: The hardware clocks of the lidar, inertial measurement unit, and camera are synchronized through GNSS pulse-per-second signals, and the sampling data of the six-axis force / torque sensor and the fiber Bragg grating sensor are aligned by linear interpolation. Based on the extended particle filter algorithm, the laser radar obstacle distance, IMU attitude angle, camera visual feature points and fiber Bragg grating strain data are integrated to generate a six-degree-of-freedom posture matrix; The posture control instructions are output based on the posture matrix to adjust the boom luffing angle.

6. The method for lifting and hoisting a vehicle as claimed in claim 2, characterized in that: S5 specifically includes: A linear mapping model between the wavelength offset of the fiber Bragg grating sensor and the bending moment at the root of the crane arm was established; When the real-time strain value reaches the first-level threshold, the model predictive control algorithm reduces the amplitude change speed and adjusts the pitch angle to optimize the bending moment distribution; When the real-time strain value reaches the second-level threshold, the hydraulic system locking command is immediately triggered and the sound and light alarm device is activated; The rolling optimization window of the model predictive control algorithm is multiple control cycles in the future, and each optimization time does not exceed 200ms.

7. The method for lifting and hoisting a vehicle as claimed in claim 2, characterized in that: S1 also includes: Based on the 3D environment map, a collision-free path is planned with the size of the component to be hoisted expanded by 0.8m as the safety envelope constraint; The laser radar scans the working area at a frequency of 10 Hz to detect dynamic obstacles, extract the coordinates of the obstacle's center of mass, and predict its motion trajectory in the next 2 seconds through the extended Kalman filter; The predicted trajectory is collided with the current lifting path. When the intersection of the paths is less than 1.2m from the end of the boom, a fast expansion random tree algorithm is used to generate a local obstacle avoidance path to ensure that the real-time distance to the dynamic obstacle is not less than 0.5m.