A Reinforcement Learning Based Dynamic Walking Control

Author

Mao, Yong ; Wang, Jiaxin ; Jia, Peifa ; Li, Shi ; Qiu, Zhen ; Zhang, Le ; Han, Zhuo

Author_Institution

Dept. of Comput. Sci. & Technol., Tsinghua Univ., Beijing

fYear

2007

fDate

10-14 April 2007

Firstpage

3609

Lastpage

3614

Abstract

A quasi-passive dynamic walking robot is built to study natural and energy-efficient biped walking. The robot is actuated by MACCEPA actuators. A reinforcement learning based control method is proposed to enhance the robustness and stability of the robot´s walking. The proposed method first learns the desired gait for the robot´s walking on a flat floor. Then a fuzzy advantage learning method is used to control it to walk on uneven floor. The effectiveness of the method is verified by simulation results.

Keywords

fuzzy control; learning (artificial intelligence); learning systems; legged locomotion; robot dynamics; robust control; MACCEPA actuators; biped walking; dynamic walking control; fuzzy advantage learning method; quasipassive dynamic walking robot; reinforcement learning; robustness; stability; Actuators; Energy efficiency; Knee; Learning; Leg; Legged locomotion; Robot control; Robot sensing systems; Robotics and automation; Robust stability; biped robot; fuzzy advantage learning; passive dynamic walking; reinforcement learning;

fLanguage

English

Publisher

ieee

Conference_Titel

Robotics and Automation, 2007 IEEE International Conference on

Conference_Location

Roma

ISSN

1050-4729

Print_ISBN

1-4244-0601-3

Electronic_ISBN

1050-4729

Type

conf

DOI

10.1109/ROBOT.2007.364031

Filename

4209649