• Project Report
• Project Presentation

The goal of this project was to enable the computer to play a computer game in a way comparable to a human player. This means the game runs completely unmodified and in real-time and the computer just uses the visual information to plan its moves.

The game was approached using 3 algorithms. A traditional computer vision implementation, a supervised learning system, and a reinforcement learning system.

Details on this in the Report and Presentation (Top of the Page).

• Traditional Computer Vision Implementation
• Superviced Learning Run
• Typical SARSA run
• The best Eval Run (SARSA)