Research On Obstacle Avoidence Grabbing Of Robot Based On 3D Vision

Service robots are an important branch of future robot development.The most typical task of service robots is to assist human to grasp and place objects.However,service robots are used in non-fixed scenes,so how to avoid obstacles quickly and stably to achieve grasp will be the difficulty for service robots.Therefore,this topic based on visual acquisition of objects and obstacles to achieve robot arm obstacle avoidance and grasping has great practical value and theoretical significance.In this paper,the artificial potential field method is selected to achieve obstacle avoidance path planning.Firstly,this paper designed and studied the visual system of the experiment,calibrates the color camera and infrared camera of Kinect V2 camera,obtains the distortion parameters of the two cameras,and prepares for the alignment of color image and deep image(the depth image is obtained by the infrared camera).Hand-eye calibration was performed on the EYE?TO?HAND experimental platform.The joint coordinate system of the kinova robot is established,and the forward kinematics is deduced by using D-H parameters.The inverse kinematics is solved by Newton iteration method,and verified by the manipulator.Next,the point cloud is studied and processed,including pass-through filtering,radius filtering,etc.The random sampling consistency algorithm is used to segment the point cloud data in plane,and the Euclidean clustering is used to segment the remaining point clouds.The position of the target point is determined by the recognition of the color image,and the corresponding point in the point cloud is found by the image alignment algorithm to realize the search of the target point cloud.The target attitude was obtained by SAC-IA coarse registration and ICP fine registration on the reference standard point cloud.The acquired objects and obstacles are used to establish the simulation environment.The improved artificial potential field method is used for obstacle avoidance trajectory planning.In the experiment,the repulsion force on the control point is related to the distance from the target to the obstacle,the coefficient of gravity and the ratio of repulsion force to gravity,by adjusting the repulsion force and gravity formula,the problem that the target can't avoid the obstacle in the distance caused by the excessive gravity and the problem that the target can't reach because the target is close to the obstacle are effectively solved.The inflexion points of the obstacle avoidance path were calculated,and the inflexion points were screened.Spline interpolation was performed on the trajectory points at the end of the manipulator.The simulation and experimental results show that the error rate of the proposed method is lower than that of the original potential field method,which proves that the proposed method is practical and reliable,and can realize the obstacle avoidance effectively and quickly.