An Operation System with Reinforcement Learning for an Embedded Robot Soccer System

• Main Authors: Ching-hui Hsu, 許慶輝
• Other Authors: Kao-Shing Hwang
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/04801830043648974443

碩士 === 國立中正大學 === 機電光工程研究所 === 91 === In this thesis, all the behavior in the robot soccer is designed under the concept of self-learning. The architecture of adaptive heuristic critic is considered as a template and a foundation of behavior. It modifies its inner parameters according to environment feedback signal for the next better movement.Three primitive behaviors are fussed into an autonomous behaving robot. These behaviors are goal-seeking behavior, toward-vacancy behavior, and obstacle-avoidance behavior. Each behavior can control the robot individually with the capability of self-learning. They are embedded in a controller module which use MC68VZ328 (DragonBall VZ) based on the core of Motorola 68K as the CPU. The operation system applied is uClinux which is designed for the CPU without MMU. Combining mentioned above with external circuits, a robot with learning mechanism is constructed.