Research On Obstacle Avoidance Of Composite Robot

In the flexible production lines of smart factories,intelligent robots need to assume the task of handling operations.This requires robots to have stronger perception and flexible mobility.Therefore,a composite robot with a perception system came into being.Compared with traditional fixed robots,the biggest difference between composite robots is the perception of changing scenes.The current robot's ability to perceive the scene is still weak.In the face of scene changes,the robot may fail and even cause a production accident.Perception and obstacle avoidance are necessary performances of a composite robot.Therefore,this article takes the obstacle avoidance of a composite robot in an unstructured scene as the task background,and realizes the obstacle avoidance movement of the composite robot in an unstructured environment.This paper studies the scene perception based on 3D lidar.In view of the poor performance of traditional ground segmentation algorithms and clustering algorithms,a ground segmentation algorithm based on cross-sectional area and a Chebyshev distance between ordered point clouds are used to determine the neighborhood radius Improved DBSCAN algorithm.By studying the path planning of the robotic arm based on scene information,the Informed RRT * algorithm is used to plan the path of the robotic arm.Aiming at the complicated operation problems of the traditional inverse kinematics solution method of the robotic arm,this paper proposes the inverse kinematics solution of the robotic arm based on the LSTM network.Main tasks as follows:(1)Build a composite robot model and simulation system.The mobile chassis and control system of the composite robot are designed.Compared with the mainstream 3D vision sensors,Li DAR is used as the scene sensing sensor.Compared with commonly used robot simulation environments,Coppelia Sim is used as the simulation environment,and the built composite robot model is imported into the simulation environment.(2)Use Li DAR to perceive scene information.Firstly,a test scenario is set up in the simulation environment and point cloud data is acquired.The obtained point cloud information is segmented on the ground using the cross-sectional area-based method proposed in this paper,and then voxelized network filtering is performed.An improved DBSCAN algorithm based on the Chebyshev distance between ordered point clouds to determine the neighborhood radius is proposed for clustering operation.Finally,a bounding box operation is performed to extract scene information.(3)Path planning for the manipulator arm of a composite robot based on a Rapidly Expanded Random Tree(RRT)algorithm.This paper compares the running effects of the classic RRT* algorithm and the Informed RRT* algorithm in two-dimensional and three-dimensional environments,and uses the Informed RRT*algorithm to plan the motion of the UR5 robot arm.(4)Drive the robot arm to servo motion.First,establish the positive kinematics model of the robotic arm and verify its correctness.Then use Monte Carlo method to collect the posture data of the end of the robotic arm and the corresponding joint angle data,then divide the training set and test set.Of inverse kinematics for robotic arm training network model.Finally,given the specific path information,test the actual motion effect of the robot arm in the simulation environment.(5)Joint simulation experiments to test the process and method of this article.Firstly,set up a cargo platform scenario in the simulation environment with the flexible production line of the smart factory as the background,and use the perception system in this paper to obtain the relevant information of the scenario.Then use the Informed RRT * algorithm to plan the path information,then use the trained LSTM network model to solve the inverse kinematics of the robotic arm,and finally drive the robotic arm of the composite robot to perform obstacle avoidance movement in the simulation environment according to the specific joint angle information Complete obstacle avoidance experiments.The experimental results show that the process and system constructed in this paper are feasible.