Skill acquisition framework in multi-robot precision assembly based on cooperative compliant control

Abstract

Robotic assemblies are widely used in manufacturing processes. However, high-precision assembly remains challenging because of numerous uncertain disturbances. Current research mainly focuses on a single robot or weakly coupled multi-robot assembly. Nevertheless, more complex and uncertainty-filled tightly coupled multi-robot assemblies have been overlooked. This study proposes an efficient skill-acquisition framework to address this challenging task by improving learning efficiency. The framework integrates a dual-loop coupled force-position control (DLCFPC) algorithm, a parallel skill-learning algorithm, and collision detection. The DLCFPC was presented to address simultaneous motion and force control challenges. In addition, a parallel skill-learning algorithm was proposed to accelerate assembly skill acquisition. Simulations and experiments on a multi-robot cooperative peg-in-hole assembly confirm that the framework enables a multi-robot system to accomplish high-precision assembly tasks even without prior knowledge, demonstrating robustness against disturbances.