GBG (General Board Game Playing & Learning) is an Open Source software framework developed at TH Köln, University of Applied Sciences. GBG aims to ease the entry for the students into game learning and the reinforcement learning research area which is a very interesting sub-field of artifical intelligence. In 2018,
two students finished their Bachelor theses successfully on the basis of the GBG under supervision of Wolfgang Konen.
The GBG framework can be used for research purposes by comparing performances of different implemented intelligent agents on different implemented games. But the most interesting fact about GBG is, that any other board game or agent can easily be embedded and used. If you are interested you can read more about GBG here.
Thanks to the flexibility of GBG, many new themes in game learning and AI can be considered as possible Bachelor or Master thesis topics. The topics can be either implementation of a new game and analysing the performance of the currently implemented agents on such games or implementing a new agent (learning algorithm) and study its effectiveness on the already existing games.
In case you are interested in any of the following topics (or anythign similar) please contact Wolfgang Konen. Some examples of possible future project themes:
- New game:
- New agent:
- Weight-sharing n-tuples (Convolutional Network)
- Comparing TD-Learning agents and SARSA agent on various GBG games
- Improving agents for larger Hex boards in GBG
- Improving agents for Othello in GBG
- Dots-and-boxes in GBG
Competition Framework in GBG (round-robin tournaments, multi-objective visualization)(already under way or done)
- Client-server architecture for GBG
- Intuitive user-interface for human play against AI in all games
New ideas are also appreciated.