A Motion Control Method Based on the Knowledge Discovery and Data Mining for Humanoid Robots

This paper proposes a KDD (knowledge discovery and data mining) approach to motion control for humanoid robots. The aim of this approach is to discover the knowledge for generating stable motions in balance. In this paper, we propose a motion control system, which can generate a stable and anti-tumble motion by the concept learning and the searching in the motion space: extracting some generalized motion guideposts by decision tree learner and motion generation with tracking the guidepost by hill-climbing search. The motion control system has three parts: training, learning, and generating part. Former two parts are to acquire the balancing property of itself body and movement by decision tree learning for numbers of executed training motions, and latter one part provides the search technique for the motion control based on the acquired knowledge concerning balance and stability in the motion. In this paper, some performance results by humanoid robot HOAP-1 are reported: stable and anti-tumble motions to stand up from a chair. This paper also reports some performance for the change in the environments: standing up from a chair on slope and different in height.

[1]  Shohei Kato,et al.  A Dynamic Allocation Method of Basis Functions in Reinforcement Learning , 2004, Australian Conference on Artificial Intelligence.

[2]  S. Kato,et al.  Motion control for humanoid robots based on the motion phase decision tree learning , 2004, Micro-Nanomechatronics and Human Science, 2004 and The Fourth Symposium Micro-Nanomechatronics for Information-Based Society, 2004..

[3]  Jun Morimoto,et al.  Acquisition of stand-up behavior by a real robot using hierarchical reinforcement learning , 2000, Robotics Auton. Syst..

[4]  J. Ross Quinlan,et al.  C4.5: Programs for Machine Learning , 1992 .

[5]  Richard S. Sutton,et al.  Reinforcement Learning: An Introduction , 1998, IEEE Trans. Neural Networks.

[6]  Masayuki Inaba,et al.  Online generation of humanoid walking motion based on a fast generation method of motion pattern that follows desired ZMP , 2002, IEEE/RSJ International Conference on Intelligent Robots and Systems.

[7]  Hirochika Inoue,et al.  Real-time humanoid motion generation through ZMP manipulation based on inverted pendulum control , 2002, Proceedings 2002 IEEE International Conference on Robotics and Automation (Cat. No.02CH37292).

[8]  Shohei Kato,et al.  A Concept Learning Based Approach to Motion Control for Humanoid Robots , 2004, ICINCO.

[9]  Masayuki Inaba,et al.  Motion planning for humanoid robots under obstacle and dynamic balance constraints , 2001, Proceedings 2001 ICRA. IEEE International Conference on Robotics and Automation (Cat. No.01CH37164).

[10]  Shohei Kato,et al.  A stochastic optimization method of CPG-based motion control for humanoid locomotion , 2004, IEEE Conference on Robotics, Automation and Mechatronics, 2004..

[11]  Richard S. Sutton,et al.  Dyna, an integrated architecture for learning, planning, and reacting , 1990, SGAR.