Konferensartikel

Our robot perceives environment with a single CCD camera; and it moves that two motors controlled independently. All robots are ``complete autonomous. Therefore; we do not use a host computer. The robot has a Note PC for ``image processing ; ``object recognition and ``select an action. In ``image processing and ``object recognition program; it is easy setting with GUI to correspond to a change of environment that extracts color of objects and establish a threshold. ``Select an action is applying reinforcement learning(see section ``Reinforcement learning). The robot has the circuit board to control the motor. This board was connected to the parallel port in the Note PC. It is composed of a motor control unit and an interface unit by CPLD (Complex Programmable Logic Device). The personal computer can control right and left motors independently by this board in five states each backward and forward. A specific role such as Goal-keeper; Defender and Attacker is given to each robot; because the robot is completely autonomy. The robot works cooperatively by the allotment of the role like this.

Inga nyckelord är tillgängliga

[1] Watkins; C.J.C.H. and Dayan;P. \Technical note: Q-Learning" Machine Learning; Vol.8 No.3 pp.279-292; 1992

[2] Shuichi Enokida; Masato Fukuda; Takeshi Ohashi; Takaichi Yoshida; and Toshiaki Ejima; \Stochastic Field Model for autonomous robot learning"; http://www.mickey.ai.kyutech.ac.jp/KIRC/kirc.html

[3] Digney; B. L.; \Learning Hierarchical Control Structures for Multiple Tasks and Changing Environments"; From animals to animats 5:SAB 98;1998