Research On Robotic Unordered Sorting And Motion Planning For Logistics Industry

In view of the chaotic characteristics of sorting operations in the logistics industry and the safety of the operating environment,this paper uses collaborative industrial robots instead of manual completion of item sorting operations,which improves the flexibility and safety of robots in performing tasks in unstructured environments.This paper is based on RGB-D depth camera to conduct theoretical exploration and experimental analysis of key technologies such as pose estimation in disordered sorting tasks,and at the same time study the collision detection mechanism of robots in complex unstructured environments and the movement of high-dimensional configuration spaces Planning method,build a comprehensive physics experiment platform to complete the comprehensive experiment of robot sorting and motion planning,and realize the task-level programming of the robot.The main research contents are as follows:This paper studies the imaging mechanism of Kinect V2 and RealSense D435 based on the working principle of RGB-D depth camera,and carries out theoretical exploration and experimental research on the calibration of internal and external parameters of the depth camera and hand-eye calibration.The color depth information obtained by the RealSense D435 sensor is preprocessed using PCL technology to perform point cloud filtering and clustering.The performance of the texture-free object pose estimation based on the matched LINEMOD method and the ICP-based PPF method is analyzed.Analyze the robot kinematics and modeling in the robot's working space and configuration space,and study the robot's self-collision detection mechanism and the collision detection principle of three-dimensional vision.After studying the RRT algorithm based on sampling,a RRT*motion planning method based on Gaussian sampling is proposed,which improves the problems of randomness,optimality and completeness of the robot motion planning in the high-dimensional configuration space.Finally,the 3D based on Kinect The robot collision detection,motion planning,trajectory interpolation and other algorithms are verified under the obstacle environmentThe experiment built a ROS-based virtual simulation environment and a UR5 robot physical experiment platform,and then used Gazebo physics simulation engine and MoveIt control platform to verify the completeness and optimality of Gaussian sampling-based RRT*algorithm in high-dimensional configuration space planning.Then the data of RealSense D435 was used to analyze the performance of the pose estimation algorithm.Experimental results show that the ICP-PPF pose estimation algorithm performs better in terms of recognition rate and matching accuracy.Finally,a comprehensive experiment of robot unordered sorting and motion planning in a physical platform was performed to complete the grabbing planning task for untextured objects.