Teaching a Computer to Fight Itself (Implementing an OGRE AI)
summary
A Python console simulator implements Ogre rules, A pathfinding, a genetic algorithm over genomes that encode unit types and behaviors, and logs outcomes to a MySQL database; source code lives on GitHub. Training runs up to 100,000 simulated games* and shows learning: random play yields ~5 defender wins per 100 games, the GA boosts results (116 wins per 1,000 games, ~56% better) but plateaus after ~1,000 games and never reaches a complete defense victory. Results and parameters are reproducible from the repo.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:31The First Computer
- 3:37Artificial Intelligence
- 8:32GitHub
- 10:15Ogre
- 10:50Game Overview
- 20:38Unit Behavior
- 29:38Mutations
- 33:23Python Code
- 37:30Training Data
- 40:20Results
- 46:29What drove the ogre behavior
- 47:31Mikes term project
- 51:08Mikes experience with Python
related talks
39mBuilding self evolving video game AI in Python
37mHow I trained an AI Model to Beat the 1990's Arcade Game Double Dragon
31mRisk Analysis
06mCreating and Teaching Games in Python to Teach Artificial Intelligence Concepts
35mHow I Built an AI to Beat the 1990s Arcade Game Double Dragon
46mAI Takeover
31mUsing Python to build an AI to play and win SNES StreetFighter II
47mHow we use Reinforcement Learning (...)