A high-precision electric sandwich plate control system

Abstract

The application relates to the technical field of transport flat plate, and particularly relates to a high-precision electric sandwich plate control system, which comprises a fusion perception system based on a laser radar and an inertial measurement unit, state estimation of a Kalman filtering algorithm, a global path planning model, model predictive control trajectory tracking, and an ultra-wideband positioning and visual sensor fine adjustment module. The application can realize sub-centimeter-level positioning and pixel-level parking precision through multi-sensor fusion, effectively correct cumulative errors, improve the navigation reliability of a complex path area, and optimize the system stability through a feedback mechanism, so that the efficient material transportation and precise positioning requirements in the industrial automation field are met.

Description

Technical Field

[0001] This invention relates to the field of transport flatbed technology, and more specifically to a high-precision electric sandwich pallet control system. Background Technology

[0002] In the field of modern industrial automation, material handling and precise positioning technologies are crucial, directly impacting production efficiency and product quality, and serving as one of the core pillars for the manufacturing industry's transition to intelligent manufacturing. With the increasing demand for automation and precision in industrial settings, electric sandwich pallets, as a flexible transport tool, play an indispensable role on production lines, and their application potential is widely recognized.

[0003] However, many current solutions still have significant shortcomings in practical applications. Some systems rely too heavily on manual operation, resulting in low levels of automation and difficulty in improving efficiency. Furthermore, positioning accuracy and stability often fail to meet the demands of complex working conditions, particularly in multi-station switching and dynamic environments. These limitations lead to frequent errors and delays in the production process, hindering overall efficiency.

[0004] A deeper analysis of the challenges in this field reveals that the core issue lies in the integration of automated navigation and precise positioning. Firstly, the stability of automated navigation technology directly impacts the performance of electric sandwich pallets in complex paths. Inaccurate or inflexible navigation can lead to deviations from the planned route during transport, consequently affecting the efficiency of subsequent workstations. Navigation issues, in turn, raise challenges regarding positioning accuracy. Even with accurate navigation, failure to achieve precise stopping and fixing at the target workstation can cause critical steps such as measurement or processing to fail due to deviations. This progressive technical challenge from navigation to positioning threatens the reliability of the entire transportation and operational process. Summary of the Invention

[0005] The purpose of this invention is to address the aforementioned shortcomings in the prior art by providing a high-precision electric sandwich panel control system.

[0006] The objective of this invention is achieved through the following technical solution: a high-precision electric sandwich panel control system, comprising the following steps:

[0007] S1. A fusion sensing system based on lidar and inertial measurement unit collects real-time environmental data and motion state data of the electric sandwich board from the dynamic environment, and filters the collected point cloud data and acceleration and angular velocity data to generate a first environmental dataset and a first motion dataset.

[0008] S2. Based on the first environmental dataset and the first motion dataset, the Kalman filter algorithm is used to estimate the real-time position and attitude of the electric sandwich board, and the accumulated error is corrected by environmental feature point cloud matching to generate a second position estimate and a second attitude estimate.

[0009] S3. Based on the second position estimate and the second attitude estimate, a global path planning model is constructed. The coordinate information and path constraints of the target workstation are obtained from the pre-established factory map database. It is determined whether the Euclidean distance between the current position and the target workstation coordinates exceeds the preset threshold range. If it does, the first optimized path is generated through dynamic programming algorithm.

[0010] S4. Based on the first optimized path, monitor the deviation of the electric sandwich board from the predetermined route in the complex path area in real time. Compare the local environmental features scanned by the lidar with the path template library. If the deviation value exceeds the preset threshold, trigger the path correction mechanism to generate the adjusted second optimized path.

[0011] S5. Based on the second optimized path, and combined with the real-time speed and angular velocity data of the electric sandwich board, a trajectory tracking system based on model predictive control is constructed. The key path point is extracted from the second optimized path as the control target. It is determined whether the distance error between the current trajectory point and the key path point is greater than a preset threshold. If it is greater, the first trajectory correction command is generated by adjusting the driving parameters.

[0012] S6. Based on the first trajectory correction instruction, drive the electric sandwich board to approach the target workstation along the second optimized path. When approaching the target workstation, obtain the third location data through the ultra-wideband positioning base station deployed near the workstation, and compare the third location data with the coordinates of the target workstation with high precision. If the position error between the two is less than the preset threshold, enter the final docking preparation state.

[0013] S7. Based on the final docking preparation state, activate the vision sensor module, acquire the first image data from the target workstation marker, and perform edge detection and feature extraction on the first image data through an image processing algorithm to generate a second image feature dataset.

[0014] S8. Based on the second image feature dataset and combined with the third position data, construct an image-guided fine-tuning control system, determine whether there is a pixel-level deviation between the center position of the marker point in the second image feature dataset and the preset standard position, and if so, generate a second fine-tuning control command by calculating the deviation vector to complete the final precise docking of the target workstation.

[0015] S9. Based on the final precise docking status, the docking efficiency and processing stability of the workstation are monitored in real time through a multi-sensor system. The docking completion time and processing error data are extracted from the sensor data. If the docking completion time or processing error exceeds the preset performance threshold, the relevant data is fed back to the navigation model and positioning module for parameter optimization, and a third optimized path and a third fine-tuning control command are generated.

[0016] The present invention is further configured such that: the first environmental dataset includes the distribution of environmental feature points after filtering point cloud data collected by lidar; the first motion dataset includes motion state data after filtering acceleration and angular velocity data collected by inertial measurement unit; the second position estimate and the second attitude estimate respectively represent the real-time spatial position and orientation angle of the electric sandwich board in the dynamic environment; the first optimized path is the shortest feasible path from the current position to the target workstation; the second optimized path is a new path generated after path correction; the first trajectory correction instruction is used to adjust the driving trajectory of the electric sandwich board; the third position data is sub-centimeter-level positioning data provided by an ultra-wideband positioning base station; the second image feature dataset includes the edge contour and feature point distribution of the target workstation marker; the second fine-tuning control instruction is used to drive the electric sandwich board to complete pixel-level deviation correction; and the third optimized path and the third fine-tuning control instruction are used to improve the long-term stability and adaptability of the system.

[0017] The present invention is further configured such that, based on a fusion sensing system of lidar and inertial measurement unit, the following steps are taken to collect real-time environmental data and motion state data of the electric sandwich panel in a dynamic environment, and to filter the collected point cloud data and acceleration and angular velocity data to generate a first environmental dataset and a first motion dataset:

[0018] The fusion sensing system based on lidar and inertial measurement unit scans the environment around the electric sandwich board with lidar to generate raw point cloud data containing obstacles, boundary lines and ground features, and collects acceleration and angular velocity data of the electric sandwich board with inertial measurement unit.

[0019] The original point cloud data and acceleration and angular velocity data are subjected to preliminary filtering processing. The median filtering algorithm is used to remove noise points in the point cloud data. At the same time, the acceleration and angular velocity data are smoothed using a low-pass filter to generate denoised point cloud data and motion state data.

[0020] Based on the denoised point cloud data and motion state data, a weighted average algorithm is used to synchronize the time of multiple frames of data, and coordinate transformation is used to unify the data to the same reference system to generate a first environment dataset and a first motion dataset.

[0021] The present invention is further configured such that, based on the first environmental dataset and the first motion dataset, the real-time position and attitude of the electric sandwich panel are estimated using a Kalman filter algorithm, and the accumulated error is corrected by matching environmental feature point clouds to generate a second position estimate and a second attitude estimate.

[0022] Based on the first environmental dataset and the first motion dataset, the extended Kalman filter algorithm is used to estimate the real-time position and attitude of the electric sandwich board. The initial position estimate and the initial attitude estimate are generated by alternating between the prediction stage and the update stage.

[0023] Based on the initial position estimate and initial attitude estimate, the nearest neighbor matching algorithm is used to match the environmental feature point cloud. The best matching point pair is selected by calculating the Euclidean distance and the angle between the normal vectors between the point clouds.

[0024] Based on the optimal matching point pair, the initial position estimate and initial attitude estimate are corrected using the least squares method to generate the second position estimate and second attitude estimate.

[0025] The present invention is further configured to, based on the second position estimate and the second attitude estimate, construct a global path planning model, obtain the coordinate information and path constraints of the target workstation from a pre-established factory map database, and determine whether the Euclidean distance between the current position and the target workstation coordinates exceeds a preset threshold range. If it does, the step of generating a first optimized path through a dynamic programming algorithm is as follows:

[0026] Based on the second position estimate and the second attitude estimate, a global path planning model is constructed. The A* algorithm is used to search for a set of candidate paths from the current position to the target workstation, and the cost of each candidate path is evaluated based on the path length, obstacle density and curvature change.

[0027] Based on the candidate path set, a dynamic programming algorithm is used to select the path with the lowest cost as the first optimized path, while recording the coordinates and turning angles of key nodes on the path.

[0028] The present invention is further configured such that, based on the first optimized path, the deviation of the electric sandwich board from the predetermined route in a complex path area is monitored in real time, and the local environmental features scanned by lidar are compared with the path template library. If the deviation value exceeds a preset threshold, a path correction mechanism is triggered to generate an adjusted second optimized path. The specific steps are as follows:

[0029] Based on the first optimized path, the local environmental features of the electric sandwich board in the complex path area are collected in real time, local point cloud data is generated by LiDAR scanning, and key feature points in the point cloud data are extracted.

[0030] The key feature points of the local point cloud data are compared with the preset feature points in the path template library, and the deviation value between the two is calculated using the Hausdorff distance algorithm.

[0031] If the deviation exceeds the preset threshold, the path correction mechanism is triggered, and the adjusted second optimized path is generated by re-calling the global path planning model.

[0032] The present invention is further configured such that, based on the second optimized path and combined with the real-time velocity and angular velocity data of the electric sandwich plate, a trajectory tracking system based on model predictive control is constructed. The key path points are extracted from the second optimized path as control targets. The step of determining whether the distance error between the current trajectory point and the key path point is greater than a preset threshold, and if so, generating a first trajectory correction command by adjusting the drive parameters, specifically includes:

[0033] Based on the second optimized path, the coordinates and turning angles of the key path points on the path are extracted and used as the control target of the trajectory tracking system.

[0034] By combining the real-time velocity and angular velocity data of the electric sandwich board, a model predictive control algorithm is used to predict the position of trajectory points in multiple future time steps, and the distance error between the current trajectory point and the critical path point is calculated.

[0035] If the distance error is greater than the preset threshold, the first trajectory correction command is generated by adjusting the speed of the drive motor of the electric sandwich plate and the angle of the steering mechanism.

[0036] The present invention is further configured such that, based on the first trajectory correction command, the electric sandwich board is driven to approach the target workstation along the second optimized path. When approaching the target workstation, a third location data is obtained through an ultra-wideband positioning base station deployed near the workstation, and the third location data is compared with the coordinates of the target workstation with high precision. If the position error between the two is less than a preset threshold, the final docking preparation state is entered.

[0037] Based on the first trajectory correction command, the electric sandwich plate is brought closer to the target workstation along the second optimized path through the coordinated control of the drive motor and the steering mechanism.

[0038] When approaching the target workstation, the third position data of the electric sandwich board is obtained by the ultra-wideband positioning base station deployed near the workstation, and the third position data is compared with the coordinates of the target workstation with high precision. The root mean square error algorithm is used to calculate the position error between the two.

[0039] If the position error is less than the preset threshold, the system will enter the final docking preparation state.

[0040] The present invention is further configured such that, based on the final docking preparation state, the visual sensor module is activated to acquire first image data from the target workstation marker, and edge detection and feature extraction are performed on the first image data using an image processing algorithm to generate a second image feature dataset.

[0041] Based on the final docking preparation state, the vision sensor module on the electric sandwich panel is activated, and the first image data of the target workstation marker is acquired through the camera.

[0042] The first image data is preprocessed by using a Gaussian filtering algorithm to remove image noise and using a Canny edge detection algorithm to extract the edge contours of the marker points.

[0043] Based on the edge contour, the Harris corner detection algorithm is used to extract the feature point distribution of the marker points and generate a second image feature dataset.

[0044] The present invention is further configured to, based on the second image feature dataset and combined with the third position data, construct an image-guided fine-tuning control system, determine whether there is a pixel-level deviation between the center position of the marker point in the second image feature dataset and the preset standard position, and if so, generate a second fine-tuning control command by calculating the deviation vector to complete the final precise docking of the target workstation. The specific steps are as follows:

[0045] Based on the second image feature dataset, the pixel coordinates of the center position of the marker point are calculated and compared with the pixel coordinates of the preset standard position.

[0046] If there is a pixel-level deviation between the center position of the marker point and the preset standard position, a second fine-tuning control command is generated by calculating the deviation vector.

[0047] Based on the second fine-tuning control command, the final precise stopping at the target workstation is achieved by adjusting the speed of the drive motor of the electric sandwich plate and the angle of the steering mechanism.

[0048] The beneficial effects of this invention are as follows: This invention achieves precise perception of the surrounding environment and the motion state of the electric sandwich panel through a fusion sensing system of lidar and inertial measurement unit, improving navigation accuracy. The application of the Kalman filter algorithm effectively corrects the accumulated errors in position and attitude estimation, ensuring reliability in dynamic environments. The combination of a global path planning model and a dynamic programming algorithm makes path planning more flexible and efficient, capable of handling the challenges of complex path areas. The introduction of a model predictive control algorithm significantly improves the accuracy and stability of trajectory tracking; furthermore, the collaborative work of the ultra-wideband positioning base station and the vision sensor module achieves sub-centimeter-level positioning accuracy and pixel-level docking accuracy, solving the high-precision requirements in multi-station switching. Through the feedback mechanism of the multi-sensor system, the long-term stability and adaptability of the system are further optimized, meeting the core requirements of efficient material transportation and precise positioning in the field of industrial automation. Attached Figure Description

[0049] The invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0050] Figure 1 This is a system flowchart of the present invention;

[0051] Figure 2 This is a schematic diagram of the structure of the electric sandwich panel of the present invention;

[0052] Among them: 1. Electric sandwich panel; 11. LiDAR; 12. Vision sensor module. Detailed Implementation

[0053] The present invention will be further described in conjunction with the following embodiments.

[0054] Depend on Figures 1 to 2 As can be seen, the high-precision electric sandwich panel 1 control system described in this embodiment includes a lidar 11, an inertial measurement unit, an ultra-wideband positioning base station, a vision sensor module 12, and target workstation markers.

[0055] The lidar 11 is installed on the electric sandwich panel 1. The lidar 11 is connected to the control unit via a data cable and is used to collect raw point cloud data around the electric sandwich panel 1 in real time. The inertial measurement unit is fixed inside the electric sandwich panel 1 near the center of gravity and works synchronously with the lidar 11. The inertial measurement unit communicates with the control unit through a signal interface and provides acceleration and angular velocity data. The two components together constitute a fusion sensing system for acquiring environmental information and motion state information of the electric sandwich panel 1.

[0056] Ultra-wideband positioning base stations are deployed near the target workstation in the factory, typically mounted on fixed supports above or to the side of the target workstation. These base stations are connected to a receiver on the electric sandwich panel 1 via a wireless communication module, providing sub-centimeter-level positioning data as the target workstation approaches. A vision sensor module 12 is installed at the center of the front end of the electric sandwich panel 1, with its lens facing the target workstation marker. The vision sensor module 12 is connected to an image processing unit via a data cable to acquire first image data of the target workstation marker. The target workstation marker is positioned at the center of the surface of the target workstation and features a high-contrast geometric design for easy identification by the vision sensor module 12.

[0057] In actual operation, the lidar 11 first scans the dynamic environment surrounding the electric sandwich panel 1, generating raw point cloud data containing obstacles, boundary lines, and ground features. Simultaneously, the inertial measurement unit acquires the acceleration and angular velocity data of the electric sandwich panel 1. These two sets of data are processed by a median filter algorithm and a low-pass filter, respectively, to generate denoised point cloud data and motion state data. Subsequently, a weighted averaging algorithm is used to synchronize the time of multiple frames of data, and coordinate transformation is used to unify the data to the same reference frame, forming the first environment dataset and the first motion dataset.

[0058] Based on the first environmental dataset and the first motion dataset, the control unit uses the extended Kalman filter algorithm to estimate the real-time position and attitude of the electric sandwich panel 1. In the prediction phase, the algorithm predicts the next position and attitude of the electric sandwich panel 1 based on the acceleration and angular velocity data in the first motion dataset. In the update phase, the algorithm corrects the prediction results using point cloud data from the first environmental dataset. To further reduce accumulated errors, the control unit uses a nearest neighbor matching algorithm to match the environmental feature point clouds. It selects the best matching point pairs by calculating the Euclidean distance and the angle between the normal vectors of the point clouds, and uses the least squares method to correct the initial position and attitude estimates, ultimately generating second position and second attitude estimates.

[0059] During the path planning phase, the control unit constructs a global path planning model based on the second position estimate and the second attitude estimate. The A* algorithm searches for a set of candidate paths from the current position to the target workstation, and evaluates the cost of each candidate path based on path length, obstacle density, and curvature variation. A dynamic programming algorithm selects the path with the lowest cost as the first optimized path, while simultaneously recording the coordinates and turning angles of key nodes along the path. During path execution, the LiDAR 11 collects local environmental features within the complex path area in real time, extracts key feature points from the local point cloud data, and compares them with preset feature points in the path template library. The Hausdorff distance algorithm is used to calculate the deviation between the two. If the deviation exceeds a preset threshold, a path correction mechanism is triggered, and the global path planning model is re-invoked to generate an adjusted second optimized path.

[0060] During the trajectory tracking phase, the control unit extracts the coordinates and steering angle information of critical path points on the path based on the second optimized path, using these as the control targets of the trajectory tracking system. Combining the real-time speed and angular velocity data of the electric sandwich plate 1, a model predictive control algorithm is used to predict the trajectory point positions for multiple future time steps, and the distance error between the current trajectory point and the critical path point is calculated. If the distance error exceeds a preset threshold, a first trajectory correction command is generated by adjusting the speed of the drive motor and the steering mechanism angle of the electric sandwich plate 1. The drive motor and steering mechanism are connected to the wheel axle of the electric sandwich plate 1 via a mechanical transmission device. After receiving the first trajectory correction command, they work together to bring the electric sandwich plate 1 closer to the target workstation along the second optimized path.

[0061] As the system approaches the target workstation, the ultra-wideband positioning base station transmits third position data to the electric sandwich panel 1 via a wireless communication module. The control unit performs a high-precision comparison between the third position data and the target workstation coordinates, calculating the position error using a root mean square error algorithm. If the position error is less than a preset threshold, the system enters the final docking preparation state. At this time, the vision sensor module 12 is activated and acquires the first image data of the target workstation marker. The image processing unit preprocesses the first image data, using a Gaussian filtering algorithm to remove image noise and a Canny edge detection algorithm to extract the edge contours of the markers. Based on the edge contours, a Harris corner detection algorithm is used to extract the feature point distribution of the markers, generating a second image feature dataset.

[0062] The control unit calculates the pixel coordinates of the center position of the marker point based on the second image feature dataset and compares them with the pixel coordinates of the preset standard position. If there is a pixel-level deviation between the center position of the marker point and the preset standard position, a second fine-tuning control command is generated by calculating the deviation vector. After receiving the second fine-tuning control command, the drive motor and steering mechanism adjust the speed of the drive motor of the electric sandwich plate 1 and the angle of the steering mechanism to achieve the final precise docking at the target station. Throughout the process, the multi-sensor system monitors the docking efficiency and processing stability of the station in real time, extracting docking completion time and processing error data from the sensor data. If the docking completion time or processing error exceeds the preset performance threshold, the relevant data is fed back to the navigation model and positioning module for parameter optimization, generating a third optimized path and a third fine-tuning control command.

[0063] In the above embodiments, the connection and positional relationships between the components ensure the efficient operation of the system. The lidar 11 and inertial measurement unit are connected to the control unit via data cables, providing basic sensing data for the system. The ultra-wideband positioning base station is connected to the receiver on the electric sandwich panel 1 via a wireless communication module, providing precise positioning support. The vision sensor module 12 is connected to the image processing unit via a data cable, responsible for the final precise positioning of the target workstation. The target workstation markers work in conjunction with the vision sensor module 12 to ensure pixel-level docking accuracy. The collaborative work of all components realizes a complete control process from environmental perception to precise docking, meeting the core requirements of efficient material transportation and precise positioning in the field of industrial automation.

[0064] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principle of this invention is further explained below in conjunction with a specific application scenario.

[0065] In an industrial automated production line, the electric sandwich panel 1 is responsible for transporting materials from the starting station to the target station. First, a lidar sensor 11 is installed on the electric sandwich panel 1 to scan the surrounding environment and generate raw point cloud data containing obstacles, boundary lines, and ground features. Simultaneously, an inertial measurement unit (IMU) collects the acceleration and angular velocity data of the electric sandwich panel 1. These two sets of data are processed using a median filter algorithm and a low-pass filter to remove noise and smooth the motion data. Subsequently, the control unit uses a weighted average algorithm to synchronize multiple frames of data in time and unifies the data to the same reference frame through coordinate transformation, forming a first environmental dataset and a first motion dataset. This process ensures that the electric sandwich panel 1 can perceive changes in the surrounding environment and its own motion state in real time, providing a reliable data foundation for subsequent navigation and positioning.

[0066] Based on the above data, the control unit uses the extended Kalman filter algorithm to estimate the position and attitude of the electric sandwich panel 1. In the prediction phase, the algorithm calculates the possible position and attitude at the next moment based on acceleration and angular velocity data; in the update phase, it corrects the prediction results using point cloud data. To further reduce accumulated errors, the control unit matches the environmental feature point cloud using a nearest neighbor matching algorithm, selects the best matching point pairs, and optimizes the initial position and attitude estimates using the least squares method, ultimately generating a second position estimate and a second attitude estimate. This process significantly improves the navigation accuracy and stability of the electric sandwich panel 1 in dynamic environments, solving the positioning deviation problem caused by accumulated errors in traditional methods.

[0067] During the path planning phase, the control unit constructs a global path planning model based on the second position estimate and the second attitude estimate. The A* algorithm searches for a set of candidate paths from the current position to the target workstation, and evaluates the cost of each candidate path based on path length, obstacle density, and curvature variation. A dynamic programming algorithm selects the path with the lowest cost as the first optimized path and records the coordinates and turning angles of key nodes along the path. In actual operation, the LiDAR 11 continuously monitors the local environmental features within the complex path area, extracts key feature points from the local point cloud data, and compares them with preset feature points in the path template library. If the deviation calculated by the Hausdorff distance algorithm exceeds a preset threshold, a path correction mechanism is triggered, and the global path planning model is re-invoked to generate an adjusted second optimized path. This design effectively addresses the path deviation problem under complex working conditions, ensuring the stable operation of the electric sandwich plate 1 on the predetermined route.

[0068] Upon entering the trajectory tracking phase, the control unit extracts the coordinates of critical path points and steering angle information based on the second optimized path, using these as the control targets of the trajectory tracking system. Combining the real-time speed and angular velocity data of the electric sandwich plate 1, the control unit employs a model predictive control algorithm to predict the trajectory point positions for multiple future time steps and calculates the distance error between the current trajectory point and the critical path point. If the distance error exceeds a preset threshold, a first trajectory correction command is generated by adjusting the drive motor speed and steering mechanism angle. Upon receiving the command, the drive motor and steering mechanism work together to guide the electric sandwich plate 1 along the second optimized path towards the target workstation. This process achieves precise control of the driving trajectory, avoiding path deviation caused by external interference or system errors.

[0069] When the electric sandwich pallet 1 approaches the target workstation, the ultra-wideband positioning base station transmits third location data via the wireless communication module. The control unit performs a high-precision comparison between this data and the target workstation coordinates, and calculates the position error between the two using the root mean square error algorithm. If the position error is less than a preset threshold, it enters the final docking preparation state. At this time, the vision sensor module 12 is activated and acquires the first image data of the target workstation marker. The image processing unit preprocesses the first image data, using a Gaussian filtering algorithm to remove image noise and extracting the edge contours of the markers using the Canny edge detection algorithm. Based on the edge contours, the Harris corner detection algorithm is used to extract the feature point distribution of the markers, generating a second image feature dataset. This process ensures that the electric sandwich pallet 1 can accurately identify the target workstation marker, laying the foundation for subsequent precise positioning.

[0070] The control unit calculates the pixel coordinates of the center position of the marker point based on the second image feature dataset and compares them with the pixel coordinates of the preset standard position. If a pixel-level deviation exists, a second fine-tuning control command is generated by calculating the deviation vector. After receiving the command, the drive motor and steering mechanism adjust the speed of the drive motor of the electric sandwich plate 1 and the angle of the steering mechanism to achieve the final precise docking at the target workstation. This design achieves sub-centimeter-level positioning accuracy and pixel-level docking accuracy, meeting the core requirements of efficient material transportation and precise positioning in the field of industrial automation.

[0071] Throughout the process, the multi-sensor system monitors the docking efficiency and processing stability of the workstations in real time, extracting docking completion time and processing error data from the sensor data. If the docking completion time or processing error exceeds a preset performance threshold, the relevant data is fed back to the navigation model and positioning module for parameter optimization, generating a third optimized path and a third fine-tuning control command. This feedback mechanism not only improves the long-term stability and adaptability of the system but also provides an optimization basis for the execution of subsequent tasks.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A high-precision electric sandwich panel control system, characterized in that: Includes the following steps: S1. A fusion sensing system based on lidar and inertial measurement unit collects real-time environmental data and motion state data of the electric sandwich board from the dynamic environment, and filters the collected point cloud data and acceleration and angular velocity data to generate a first environmental dataset and a first motion dataset. S2. Based on the first environmental dataset and the first motion dataset, the Kalman filter algorithm is used to estimate the real-time position and attitude of the electric sandwich board, and the accumulated error is corrected by environmental feature point cloud matching to generate a second position estimate and a second attitude estimate. S3. Based on the second position estimate and the second attitude estimate, a global path planning model is constructed. The coordinate information and path constraints of the target workstation are obtained from the pre-established factory map database. It is determined whether the Euclidean distance between the current position and the target workstation coordinates exceeds the preset threshold range. If it does, the first optimized path is generated through dynamic programming algorithm. S4. Based on the first optimized path, monitor the deviation of the electric sandwich board from the predetermined route in the complex path area in real time. Compare the local environmental features scanned by the lidar with the path template library. If the deviation value exceeds the preset threshold, trigger the path correction mechanism to generate the adjusted second optimized path. S5. Based on the second optimized path, and combined with the real-time speed and angular velocity data of the electric sandwich board, a trajectory tracking system based on model predictive control is constructed. The key path point is extracted from the second optimized path as the control target. It is determined whether the distance error between the current trajectory point and the key path point is greater than a preset threshold. If it is greater, the first trajectory correction command is generated by adjusting the driving parameters. S6. Based on the first trajectory correction instruction, drive the electric sandwich board to approach the target workstation along the second optimized path. When approaching the target workstation, obtain the third location data through the ultra-wideband positioning base station deployed near the workstation, and compare the third location data with the coordinates of the target workstation with high precision. If the position error between the two is less than the preset threshold, enter the final docking preparation state. S7. Based on the final docking preparation state, activate the vision sensor module, acquire the first image data from the target workstation marker, and perform edge detection and feature extraction on the first image data through an image processing algorithm to generate a second image feature dataset. S8. Based on the second image feature dataset and combined with the third position data, construct an image-guided fine-tuning control system, determine whether there is a pixel-level deviation between the center position of the marker point in the second image feature dataset and the preset standard position, and if so, generate a second fine-tuning control command by calculating the deviation vector to complete the final precise docking of the target workstation. S9. Based on the final precise docking state, the docking efficiency and processing stability of the workstation are monitored in real time through a multi-sensor system. Docking completion time and processing error data are extracted from the sensor data. If the docking completion time or processing error exceeds a preset performance threshold, the relevant data is fed back to the navigation model and positioning module for parameter optimization, generating a third optimized path and a third fine-tuning control command. The first environmental dataset includes the distribution of environmental feature points after filtering the point cloud data collected by the lidar. The first motion dataset includes motion state data after filtering the acceleration and angular velocity data collected by the inertial measurement unit. The second position estimate and the second attitude estimate... The values ​​represent the real-time spatial position and orientation angle of the electric sandwich board in a dynamic environment. The first optimized path is the shortest feasible path from the current position to the target workstation. The second optimized path is a new path generated after path correction. The first trajectory correction command is used to adjust the driving trajectory of the electric sandwich board. The third position data is sub-centimeter-level positioning data provided by the ultra-wideband positioning base station. The second image feature dataset includes the edge contour and feature point distribution of the target workstation marker. The second fine-tuning control command is used to drive the electric sandwich board to complete pixel-level deviation correction. The third optimized path and the third fine-tuning control command are used to improve the long-term stability and adaptability of the system. Based on the first environment dataset and the first motion dataset, the Kalman filter algorithm is used to estimate the real-time position and attitude of the electric sandwich board, and the accumulated error is corrected by environmental feature point cloud matching to generate the second position estimate and the second attitude estimate. The specific steps are as follows: Based on the first environmental dataset and the first motion dataset, the extended Kalman filter algorithm is used to estimate the real-time position and attitude of the electric sandwich board. The initial position estimate and the initial attitude estimate are generated by alternating between the prediction stage and the update stage. Based on the initial position estimate and initial attitude estimate, the nearest neighbor matching algorithm is used to match the environmental feature point cloud. The best matching point pair is selected by calculating the Euclidean distance and the angle between the normal vectors between the point clouds. Based on the optimal matching point pair, the initial position estimate and initial attitude estimate are corrected using the least squares method to generate the second position estimate and second attitude estimate. Based on the first optimized path, the deviation of the electric sandwich board from the predetermined route in the complex path area is monitored in real time. The local environmental features scanned by the lidar are compared with the path template library. If the deviation value exceeds a preset threshold, the path correction mechanism is triggered to generate an adjusted second optimized path. The specific steps are as follows: Based on the first optimized path, the local environmental features of the electric sandwich board in the complex path area are collected in real time, local point cloud data is generated by LiDAR scanning, and key feature points in the point cloud data are extracted. The key feature points of the local point cloud data are compared with the preset feature points in the path template library, and the deviation value between the two is calculated using the Hausdorff distance algorithm. If the deviation exceeds the preset threshold, the path correction mechanism is triggered, and the adjusted second optimized path is generated by re-calling the global path planning model.

2. The high-precision electric sandwich panel control system according to claim 1, characterized in that: The fusion sensing system based on lidar and inertial measurement unit (IMU) collects real-time environmental data and motion state data of the electric sandwich panel in a dynamic environment, and filters the collected point cloud data and acceleration and angular velocity data to generate a first environmental dataset and a first motion dataset. The specific steps are as follows: The fusion sensing system based on lidar and inertial measurement unit scans the environment around the electric sandwich board with lidar to generate raw point cloud data containing obstacles, boundary lines and ground features, and collects acceleration and angular velocity data of the electric sandwich board with inertial measurement unit. The original point cloud data and acceleration and angular velocity data are subjected to preliminary filtering processing. The median filtering algorithm is used to remove noise points in the point cloud data. At the same time, the acceleration and angular velocity data are smoothed using a low-pass filter to generate denoised point cloud data and motion state data. Based on the denoised point cloud data and motion state data, a weighted average algorithm is used to synchronize the time of multiple frames of data, and coordinate transformation is used to unify the data to the same reference system to generate a first environment dataset and a first motion dataset.

3. The high-precision electric sandwich panel control system according to claim 1, characterized in that: Based on the second position estimate and the second attitude estimate, a global path planning model is constructed. The coordinate information and path constraints of the target workstation are obtained from a pre-established factory map database. It is determined whether the Euclidean distance between the current position and the target workstation exceeds a preset threshold. If it does, the first optimized path is generated using a dynamic programming algorithm. The specific steps are as follows: Based on the second position estimate and the second attitude estimate, a global path planning model is constructed. The A* algorithm is used to search for a set of candidate paths from the current position to the target workstation, and the cost of each candidate path is evaluated based on the path length, obstacle density and curvature change. Based on the candidate path set, a dynamic programming algorithm is used to select the path with the lowest cost as the first optimized path, while recording the coordinates and turning angles of key nodes on the path.

4. The high-precision electric sandwich panel control system according to claim 1, characterized in that: Based on the second optimized path, and combined with the real-time velocity and angular velocity data of the electric sandwich plate, a trajectory tracking system based on model predictive control is constructed. The key path points are extracted from the second optimized path as control targets. The process of determining whether the distance error between the current trajectory point and the key path point is greater than a preset threshold, and if so, generating a first trajectory correction command by adjusting the drive parameters, is as follows: Based on the second optimized path, the coordinates and turning angles of the key path points on the path are extracted and used as the control target of the trajectory tracking system. By combining the real-time velocity and angular velocity data of the electric sandwich board, a model predictive control algorithm is used to predict the position of trajectory points in multiple future time steps, and the distance error between the current trajectory point and the critical path point is calculated. If the distance error is greater than the preset threshold, the first trajectory correction command is generated by adjusting the speed of the drive motor of the electric sandwich plate and the angle of the steering mechanism.

5. A high-precision electric sandwich panel control system according to claim 1, characterized in that: Based on the first trajectory correction command, the electric sandwich board is driven to approach the target workstation along the second optimized path. When approaching the target workstation, the third location data is obtained through the ultra-wideband positioning base station deployed near the workstation, and the third location data is compared with the coordinates of the target workstation with high precision. If the position error between the two is less than a preset threshold, the final docking preparation state is entered. Based on the first trajectory correction command, the electric sandwich plate is brought closer to the target workstation along the second optimized path through the coordinated control of the drive motor and the steering mechanism. When approaching the target workstation, the third position data of the electric sandwich board is obtained by the ultra-wideband positioning base station deployed near the workstation, and the third position data is compared with the coordinates of the target workstation with high precision. The root mean square error algorithm is used to calculate the position error between the two. If the position error is less than the preset threshold, the system will enter the final docking preparation state.

6. The high-precision electric sandwich panel control system according to claim 1, characterized in that: Based on the final docking preparation state, the steps of activating the vision sensor module, acquiring first image data from the target workstation marker, and generating a second image feature dataset by performing edge detection and feature extraction on the first image data using an image processing algorithm are as follows: Based on the final docking preparation state, the vision sensor module on the electric sandwich panel is activated, and the first image data of the target workstation marker is acquired through the camera. The first image data is preprocessed by using a Gaussian filtering algorithm to remove image noise and using a Canny edge detection algorithm to extract the edge contours of the marker points. Based on the edge contour, the Harris corner detection algorithm is used to extract the feature point distribution of the marker points and generate a second image feature dataset.

7. The high-precision electric sandwich panel control system according to claim 1, characterized in that: Based on the second image feature dataset and combined with the third position data, an image-guided fine-tuning control system is constructed. The system determines whether there is a pixel-level deviation between the center position of the marker point in the second image feature dataset and the preset standard position. If so, a second fine-tuning control command is generated by calculating the deviation vector to achieve the final precise docking of the target workstation. The specific steps are as follows: Based on the second image feature dataset, the pixel coordinates of the center position of the marker point are calculated and compared with the pixel coordinates of the preset standard position. If there is a pixel-level deviation between the center position of the marker point and the preset standard position, a second fine-tuning control command is generated by calculating the deviation vector. Based on the second fine-tuning control command, the final precise stopping of the target workstation is achieved by adjusting the speed of the drive motor of the electric sandwich plate and the angle of the steering mechanism.

Images