Purpose The purpose of this paper is using a model-free reinforcement learning (RL) algorithm to optimize manipulability which can overcome difficulties of dilemmas of matrix inversion, complicated formula transformation and expensive calculation time. Design/methodology/approach Manipulability...