Sarah Eisenach, Niall Williams
Faculty Sponsor: Dr. Ramanujan
Tetris is a block packing game in which the player attempts to pack as many falling blocks before running out of space. Success is measured by the number of lines cleared, and maximizing the lines cleared is an NP-complete problem. We have implemented an AI using reinforcement learning to clear as many lines as possible in the game of Tetris. Our AI is modeled as an Markov Decision Process (MDP), using classification-based modified policy iteration (CBMPI), which searches over the policy space of the game.
