Improving Trackmania Reinforcement Learning Performance : A Comparison of Sophy and Trackmania AI

Author(s): Neinders, L.J. (2023)

Abstract:
The thesis compares the performance of two AI agents, Sophy and Trackmania AI (TMAI), which were trained using reinforcement learning to play racing games. The study examines how Sophy achieved its outstanding performance in "Gran Turismo Sport" (GTS) and identifies ways to improve TMAI's performance, developed for "Trackmania", by transferring modifications from Sophy. The research evaluates the effects of these changes on TMAI's performance.

Document(s):

Neinders_BA_EEMCS.pdf