Coordinated Object Locating of Dual-arm Robotic System Using Marker-based Multi-view Pose Estimation for Autonomous Workpiece Loading

Abstract

Industrial robots play crucial roles on machining and manufacturing automation. Recently, more and more highly repetitive and hazardous jobs have been done by industrial robots. However, current automatic machining systems by robots are still not flexible and robust enough in the workpiece-loading process. In this paper, a dual-arm robotic object-locating system equipped with a depth camera is proposed for autonomous workpiece loading to improve the flexibility and robustness of the robotic machining systems. It can automatically locate the workpiece, and then load it by inserting the gripper fingers into the grasping position. Firstly, 6D pose of the in-hand workpiece occluded by the robotic gripper is estimated by the proposed marker-based multi-view pose estimation method based on point pair features (PPFs). According to the estimated pose, a dual-arm pose is generated for locating grasping position of the in-hand workpiece considering the orientational constraint, and the dual-arm motion to reach that pose is planned online by a sampling-based planning algorithm. However, due to some error factors, such as system modeling error and pose estimation error, there may be a locating error which leads to the failed loading. Therefore, in this paper, an object-locating control strategy is developed for compensating the locating error. Experiments are conducted to evaluate the performance of the proposed methods and to verify the feasibility of the proposed system. Analysis related to the experimental results is provided.