A generic manipulation task instruction framework using interactive digital twin environment

Small-batch customized assembly lines require assembly robots with the ability to deploy tasks rapidly and learn new skills quickly, while the acquired skills should exhibit hierarchical and sequential characteristics that conform to human habits. This paper presents a novel framework for robotic task learning and deployment within a digital twin environment. The primary aim is to enhance the efficiency of robotic coding through human demonstrations, addressing challenges in automation and robot-human collaboration. Our approach integrates high-level action semantic coding and behavior trees to bridge the gap between high-level task representations and low-level skill execution. Key contributions include: (1) the construction of a digital twin simulation for micro mobile phone parts assembly, (2) the realization of structured human demonstration encoding, and (3) the development of a robust skill learning system utilizing dynamic movement primitives (DMPs). Our findings demonstrate significant improvements in task learning efficiency and adaptability, marking a substantial advancement in the field of robotic automation.