Human-Robot Interaction Control Using Reinforcement Learning

This book gives brief overview of human-robot interaction control schemes, and presents novel model-free and reinforcement learning controllers. It begins with a brief introduction and state of art of human-robot interaction control and reinforcement learning. It then moves on to describe the typical environment model and some of the most famous identification techniques for parameters estimation. Later chapters address the robot-interaction schemes using impedance and admittance controllers, model-free controllers, and input forces/torques of the human operator. The authors also describe using the reinforcement learning approach for the position/force control task in discrete time, to achieve an optimal robot-environment interaction using a position/force control. They also explore how to design robust controllers based on the modified reinforcement learning under the worst-case uncertainty. Closing topics include inverse and velocity kinematics solution, H2 neural control, and future developments in the field.

本書簡要介紹了人機互動控制方案，並提出了新穎的無模型和強化學習控制器。首先，它簡要介紹了人機互動控制和強化學習的現狀。然後，描述了典型的環境模型和一些最著名的參數估計技術。後面的章節介紹了使用阻抗和輸入力矩控制器、無模型控制器以及人類操作者的輸入力矩的機器人互動方案。作者還描述了在離散時間中使用強化學習方法進行位置/力控制任務，以實現最佳的機器人-環境互動控制。他們還探討了如何基於最壞情況下的不確定性修改強化學習來設計魯棒控制器。最後，本書還包括逆運動學和速度運動學解法、H2神經控制以及該領域的未來發展。