Project Final SReport: View the document
Presentation: View the presentation
Our vision is to explore and enhance reinforcement learning techniques within multiplayer environments through the use of the SIMPLE library. By leveraging this innovative approach, we aim to create an AI that can learn and adapt by playing against previous versions of itself, a method that simulates self-play and offers continuous improvement. Our focus is to implement the game "Boop," a strategic two-player board game, and develop an intuitive user interface that allows users to play either against our trained AI model or another player locally. This combination of self-play and graphical representation will not only advance our understanding of reinforcement learning but will also provide an engaging and accessible game for users of all skill levels.
This project represents an exploration of how reinforcement learning can be effectively applied to multiplayer board games, particularly Boop. While traditional AI methods are common in single-player games, the introduction of self-play in multiplayer environments, such as Boop, enables a model to continually improve through iterative interactions. Using the SIMPLE (Selfplay in Multiplayer Environments) library, we have trained an AI model to play Boop. The project also includes the development of a user-friendly web-based graphical user interface (GUI), providing a seamless experience for users to interact with the model. The GUI allows players to track game progress, view inventory, and even play against the AI or a local opponent. The integration of these elements into a cohesive project aims to foster both a deeper understanding of reinforcement learning and a fun gaming experience.
Team:
Alexandra Prasser, CS senior who loves playing board games
Brian Hlathein, CS senior who also loves gaming
Calvin CS Website Link: View the document