A Robot Autonomous Control System for Multi-task Complex Manipulations

This paper proposes a robot autonomous control system suitable for multi-task complex manipulations. Firstly, the system integrates semantic segmentation and pose estimation techniques to solve the problem of recognition and measurement under the unstructured environment. Secondly, it decomposes complex tasks into combinations of some pre-defined primitive skills and makes use of LGP method to generate reference trajectories. Finally, the force-position hybrid control method is used to complete the reference trajectory tracking and compliant contact with the surrounding environment. The autonomous control system endows robots with the ability to operate complex, long-horizontal tasks in an unstructured environment. At last, the effect is verified by the mathematical simulation.