Grasp Planning and Control for Robotic Mobile Manipulation Based on Semantic Segmentation

In this paper we propose a docking and object grasping design for an autonomous mobile robot (AMR) based on semantic information. The AMR first navigates and docks to the station, then it uses the eye-in-hand camera to estimate object pose and grasp cluttered objects on the workstation. A coordinating controller is designed to position the mobile base and the 6-DOF robot arm simultaneously to a location that is suitable for object grasping. A grasp planning algorithm is proposed in this work based on the observed 3D point cloud of the target object. The grasp index(GI) is proposed to determine the most suitable grasping pose of the robot arm to ensure a successful object picking. The proposed methods have been implemented on the laboratory developed mobile manipulator. The experimental results show that the average error of AMR docking alignment is 0.027m in x-axis, 0.0I2m in y-axis, and 3.1 degrees in orientation. The average successive rate of random bin picking is 84.96% for three types of objects.