Walking algorithm of Apollo3D—The world champion in the RoboCup2013 3D soccer simulation competition

Author

Zhao Hecheng ; Liang Zhiwei

Author_Institution

Coll. of Autom., Nanjing Univ. of Posts & Telecommun., Nanjing, China

fYear

2014

fDate

28-30 July 2014

Firstpage

8331

Lastpage

8334

Abstract

Quick and agile walking is indispensable for humanoid robots in the RoboCup soccer competition. So this paper mainly proposed a method to develop a flexible walking based on reinforcement learning for soccer robots. First, a linear inverted pendulum model with a predictive control is applied to generate a motion trajectory of the robot´s trunk. Then, the Cerebellar Model Articulation Controller(CMAC) method is employed to optimize the walking parameters to keep dynamic balance of robots. Our team Apollo3D employed this walking skill, and won the first place in RoboCup 2013 3D simulation competition by winning 53 goals.

Keywords

cerebellar model arithmetic computers; humanoid robots; learning (artificial intelligence); mobile robots; motion control; multi-robot systems; trajectory control; Apollo3D the world champion; CMAC method; RoboCup2013 3D soccer simulation competition; cerebellar model articulation controller; flexible walking; humanoid robots; linear inverted pendulum model; motion trajectory; reinforcement learning; robot trunk; soccer robots; walking algorithm; Foot; Humanoid robots; Joints; Legged locomotion; Solid modeling; Trajectory; CMAC; linear inverted pendulum; preview control; trunk trajectory;

fLanguage

English

Publisher

ieee

Conference_Titel

Control Conference (CCC), 2014 33rd Chinese

Conference_Location

Nanjing

Type

conf

DOI

10.1109/ChiCC.2014.6896397

Filename

6896397