Skill learning for robotic assembly based on visual perspectives and force sensing

• Published in: Robotics and autonomous systems 2021-01, Vol.135, p.103651, Article 103651
• Main Authors: Song, Rui, Li, Fengming, Quan, Wei, Yang, Xuting, Zhao, Jie
• Format: Article
• Language: English
• Subjects: Force sensing; Reinforcement learning; Robot assembly; Skill learning; Visual perspectives
• ISSN: 0921-8890; 1872-793X
• DOI: 10.1016/j.robot.2020.103651

An environment cannot be effectively described with a single perception form in skill learning for robotic assembly. The visual perception may provide the object’s apparent characteristics and the softness or stiffness of the object could be detected using the contact force/torque information during the assembly process. In the process of inserting assembly strategy learning, most of the work takes the contact force information as the current observation state of the assembly process, ignoring the influence of visual information on the assembly state. This paper proposes robotic assembly skill learning with deep Q-learning using visual perspectives and force sensing to learn an assembly policy. The reward system is designed with an image template matching for assembly state, which is used to judge whether the process is completed successfully. The observations of assembly state are described by force/torque information and the pose of the end effector. To evaluate the performance of the proposed skill learning method, experiments with a KUKA iiwa robot are performed for a plastic fasten assembly in a low-voltage apparatus. The results indicate that the robot can complete the plastic fasten assembly using the learned inserting assembly strategy with visual perspectives and force sensing. •A method of assembly skill learning is proposed, which simulates the assembly process of human watching while operating.•Vision and force information is used to learn insertion assembly strategy through continuous interaction with assembly environment.•A framework is also provided for the learning of assembly strategy. The criterion of assembly success comes from visual information, rather than assembly displacement along Z axis.