Journal of Electrical Engineering and Automation

This study proposes a robot skill learning method for flexible assembly based on deep reinforcement learning, aiming to improve the robot's assembly adaptability and task completion accuracy in complex and variable industrial environments. By constructing a state-action mapping strategy network, the robot's autonomous learning and strategy optimization of the assembly process are realized. Combining simulation training and real-world transfer, the effectiveness and robustness of the proposed method in various typical assembly tasks are verified. Experimental results show that this method has significant advantages in assembly accuracy, strategy generalization, and dynamic response, which can meet the technical requirements for high flexibility and intelligent assembly in intelligent manufacturing systems.

Flexible assembly; Deep reinforcement learning; Robot skill learning; Strategy network; Intelligent manufacturing