Hybrid Trajectory and Force Learning of Complex Assembly Tasks: A Combined Learning Framework

Complex assembly tasks involve nonlinear and low-clearance insertion trajectories with varying contact forces at different stages. For a robot to solve these tasks, it requires precise adaptive controller which conventional force control methods cannot provide. Imitation learning is promising method for controllers that can the from human demonstrations without needing explicitly program them into robot. However, profiles obtain demonstration via tele-operation tend be sub-optimal complex thus undesirable imitate such profiles. Reinforcement learns policies through interactions environment but struggles low sample efficiency equipment tear wear in physical world. To address problems, we present combined learning-based framework robotic hybrid trajectory learning. The main contribution of this work development combines imitation learning, learn nominal motion trajectory, reinforcement scheme an optimal policy, satisfy while adapting requirement task. further improve part, develop hierarchical architecture, following idea goal-conditioned generate policy on skill level offline. Through experimental validations, corroborate proposed high-quality find suitable adapt tasks’ requirements more efficiently.