AI learns Resident Evil 4 via behavioral cloning

This project highlights the efficiency of Behavioral Cloning for basic competency while underscoring the "data wall" in high-level tactical AI.

• –Integrates LSTM layers to provide the model with temporal context, essential for non-reactive game states like reloading or dodging.
• –Successfully handles single-enemy encounters but fails at complex "fight-or-flee" heuristics, indicating the limits of imitation without massive datasets.
• –Built on SDLArch-RL, a Gymnasium-compliant framework that bridges retro emulation with modern deep RL stacks like Stable-Baselines3.
• –Source code and notebooks are openly available, providing a practical look at applying Behavioral Cloning to modern AAA titles.