No abstract

The method of tactile perception can accurately reflect the contact state by collecting force and torque information, but it is not sensitive to the changes in position and posture between assembly objects. The method of visual perception is very sensitive to changes in pose and posture between assembled objects, but they cannot accurately reflect the contact state, especially since the objects are occluded from each other. The robot will perceive the environment more accurately if visual and tactile perception can be combined. Therefore, this paper proposes the alignment method of combined perception for the peg-in-hole assembly with self-supervised deep reinforcement learning. The agent first observes the environment through visual sensors and then predicts the action of the alignment adjustment based on the visual feature of the contact state. Subsequently, the agent judges the contact state based on the force and torque information collected by the force/torque sensor. And the action of the alignment adjustment is selected according to the contact state and used as a visual prediction label. Whereafter, the network of visual perception performs backpropagation to correct the network weights according to the visual prediction label. Finally, the agent will have learned the alignment skill of combined perception with the increase of iterative training. The robot system is built based on CoppeliaSim for simulation training and testing. The simulation results show that the method of combined perception has higher assembly efficiency than single perception.

The adjustment times of the attitude alignment are fluctuated due to the fluctuation of the contact force signal caused by the disturbing moments in the compliant peg-in-hole assembly. However, these fluctuations are difficult to accurately measure or definition as a result of many uncertain factors in the working environment. It is worth noting that gravitational disturbing moments and inertia moments significantly impact these fluctuations, in which the changes of the peg concerning the mass and the length have a crucial influence on them. In this paper, a visual grasping strategy based on deep reinforcement learning is proposed for peg-in-hole assembly. Firstly, the disturbing moments of assembly are analyzed to investigate the factors for the fluctuation of assembly time. Then, this research designs a visual grasping strategy, which establishes a mapping relationship between the grasping position and the assembly time to improve the assembly efficiency. Finally, a robotic system for the assembly was built in V-REP to verify the effectiveness of the proposed method, and the robot can complete the training independently without human intervention and manual labeling in the grasping training process. The simulated results show that this method can improve assembly efficiency by 13.83%. And, when the mass and the length of the peg change, the proposed method is still effective for the improvement of assembly efficiency.