Summary: Poincar´e-Map-Based Reinforcement Learning
For Biped Walking
Jun Morimoto1,2, Jun Nakanishi1,2, Gen Endo2,3,
and Gordon Cheng1,2
1
Computational Brain Project, ICORP, JST
2
ATR Computational Neuroscience Laboratories
3
Sony Intelligence Dynamics Laboratories, Inc.
2-2-2 Hikaridai Soraku-gun Seika-cho, Kyoto, 619-0288, JAPAN
xmorimo@atr.jp
Christohper G. Atkeson and Garth Zeglin
The Robotics Institute
Carnegie Mellon University
5000 Forbes Ave, Pittsburgh, PA, 15213, USA
cga@cs.cmu.edu
Abstract-- We propose a model-based reinforcement learn-
ing algorithm for biped walking in which the robot learns
to appropriately modulate an observed walking pattern. Via-

Source: Atkeson, Christopher G. - Robotics Institute, School of Computer Science, Carnegie Mellon University
Zeglin, Garth - Robotics Institute, Carnegie Mellon University

Collections: Computer Technologies and Information Sciences; Engineering