Learning synergies based in-hand manipulation with reward shaping

作     者：Zhen Deng Jian Wei Zhang 

作者机构：TAMS GroupInformaticsUniversity of HamburgHamburg D-22527Germany 

出 版 物：《CAAI Transactions on Intelligence Technology》 (智能技术学报（英文）)

年 卷 期：2020年第5卷第3期

页      面：141-149页

核心收录：

学科分类：08[工学] 081101[工学-控制理论与控制工程] 0811[工学-控制科学与工程] 081102[工学-检测技术与自动化装置] 

基　　金：This work was funded by the German Science Foundation(DFG)and the National Science Foundation of China(NSFC)in project Crossmodal Learning under contract Sonderforschungsbereich Transregio 169 

主　　题：manipulation policy jointly 

摘      要：In-hand manipulation is a fundamental ability for multi-fingered robotic hands that interact with their *** to the high dimensionality of robotic hands and intermittent contact dynamics,effectively programming a robotic hand for in-hand manipulations is still a challenging *** address this challenge,this work employs deep reinforcement learning(DRL)algorithm to learn in-hand manipulations for multi-fingered robotic hands.A reward-shaping method is proposed to assist the learning of in-hand *** synergy of robotic hand postures is analysed to build a low-dimensional hand posture *** additional rewards are designed based on both the analysis of hand synergies and its learning *** two additional rewards cooperating with an extrinsic reward are used to assist the in-hand manipulation *** value functions are trained jointly with respect to their reward *** they cooperate to optimise a control policy for in-hand *** reward shaping not only improves the exploration efficiency of the DRL algorithm but also provides a way to incorporate domain *** performance of the proposed learning method is evaluated with object rotation *** results demonstrated that the proposed learning method enables multi-fingered robotic hands to learn in-hand manipulation effectively.