Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Systematic search and reinforcement learning for the board game ''Turning Points''

Vagenas Sotirios

Full record

---

Summary

Games have always been one of the main forms of entertainment for people. They combine fun, spiritual growth and skill development. For these reasons, games constitute one of the most important areas of research and application of Artificial Intelligence and Machine Learning. This thesis focuses on the board game “Turning Points”. This is a challenging two-player 4 Χ 4 square board game, in which each move corresponds to the placement of a directed arrow to an empty position, possibly causing some rotations of the existing arrows. The goal of the thesis is the design of an agent for the game TurningPoints, who will learn to play and improve against various opponents. The implementation of the agent combines the Minimax with alpha-beta Pruning search technique, an evaluation function and the reinforcement learning algorithm TD-Learning. The combination of these elements results in the extraction of several competitive players, as evidenced by the experimental evaluation procedure. These players faced human players competitivelyand efficiently. Interestingly, they also remain competitive in larger boards compared to the basic one in which they have been trained. Finally, a graphical user interface was implemented to enable game playing with users.

Available Files

Services