Abstract: The generalization model for multicategory peg-in-hole assembly (MPHA) skills is hard to acquire. An adaptive meta policy learning (AMPL) algorithm with virtual model set is proposed to deal ...