Automatic Generation of Macro-Actions using Genetic Algorithm for Reinforcement Learning

Takeshi Tateyama; Seiichi Kawata; Toshiki Oguchi

doi:10.11499/sicep.2002.0.62.0

SICE Annual Conference 2002

DOI https://doi.org/10.11499/sicep.2002.0.62.0

Conference information

Host: The Society of Instrument and Control Engineers

Co-host: IEEE/Industiral Electronic Society, IEEE/Robotics and Automation Society, IEEE/Control System Society

Automatic Generation of Macro-Actions using Genetic Algorithm for Reinforcement Learning

Takeshi Tateyama, Seiichi Kawata, Toshiki Oguchi

Author information

Keywords: reinforcement learning, macro-action, Semi-Markov Decision Processes, classifier system

CONFERENCE PROCEEDINGS FREE ACCESS

Pages 62

Details

Abstract

The main problem of reinforcement learning is that learning converges slowly. As one of the solution, McGovern proposed “macro-action”. However, a human expert needs to design macro-actions which adapt to an environment. In this paper, we propose a new method that enables to generate the macro-actions which adapt to the enviroment automatically using genetic algorithm.

Corresponding author

Register with J-STAGE for free!