A Study of Genetic Programming on Cooperative Model for Traffic Light Game

• Main Authors: Wu Cheng-Yen, 吳政諺
• Other Authors: 孫光天
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/62551814344949196823

碩士 === 臺南師範學院 === 資訊教育研究所 === 92 === The genetic programming(GP) can evolve programs automatically by simulating the evolutionary mechanism on computers. In this research, we use the GP technique and the coevolution mechanism to evolve the cooperative model for the Traffic Light Game. We hope these evolution strategies can derive the similar behaviors of the players in the real children’s Traffic Light Game. Simple functions and terminals are used to coevolve the high elastic strategies for the ghost and the players .Then the ghost can efficiently detour round these obstacles and capture players in the simulated environment. Then, the players evolve the more efficiently cooperative models to protect against the ghost . From the simulation results, the efficient strategies of a ghost and players can be coevolved with simple functions and terminals . This research can not only be used to related researches, but also can be applied to develop more complex cooperative models and strategic combination (eg. the cooperative models for robots、self-learning of virtual agent on line or human learning model).