Learning and generalising object extraction skill for contact-rich disassembly tasks: an introductory study

Abstract Remanufacturing automation must be designed to flexible and robust enough overcome the uncertainties, conditions of products, complexities in planning operation processes. Machine learning methods, particular reinforcement learning, are presented as techniques learn, improve, generalise many robotic manipulation tasks (most them related grasping, picking, or assembly). However, not much has been exploited remanufacturing, disassembly tasks. This work presents state art contact-rich using algorithms a study about generalisation object extraction skills when applied The capabilities two state-of-the-art agents (trained simulation) tested evaluated simulation, real world while perform task. Results show that at least one can skill. Besides, this identifies key concepts gaps for algorithms’ research application on