Reinforced learning by the chaotic dynamics matching for robot control

点击翻译按钮获取中文摘要

Chaotic dynamics matching which can be applied for a dynamics of the control systems with high-speed and flexibly according to the condition and the task based on the chaos theory is proposed. It achieves the reinforcement learning system that acquires the action autonomous. The following results were achieved. The attraction domain for the identification learning control parameter was theoretically analyzed and the availability of the design approach was shown. The control and the learning characteristic in a real robot were clarified. The possibility to use of human brain information to learning was verified.