Rollerball

In the Fall ‘23 semester, I collaborated on a project for my Artificial Intelligence course under Prof. Mausam. The project’s goal was to develop an AI agent that could play a variant of Chess called Rollerball. My partner and I implemented a robust solution using C++. Our primary focus was on developing the agent’s core intelligence. I implemented the minimax algorithm with alpha-beta pruning, incorporating early cut-off for efficiency. To enhance performance, I integrated move ordering and experimented with Deep Learning to estimate the evaluation function. Working within the constraint of using only C++ and no external libraries, I built a modular header file for basic MLP models, demonstrating proficiency in object-oriented programming, matrix operations, and derivatives. My partner contributed significantly to the project by implementing performance-improving features like quiescence search, transposition tables, and heatmaps.