Multi-Game Machine Learning
This project aimed to explore the creation of a
single machine learning agent that can play two distinct
video games, a hummingbird game, and a kart racing
game, in Unity. In the artifact, two agent implementations are presented for each game,
one individual agent, and one with logic mapped closely to the alterative game.
This project highlighted the effectiveness of individual reinforcement learning ML agents in replicating ideal player behaviour (collecting nectar from flowers, completing laps of a racetrack), and the challenges of creating a singular trained ML agent that could play two distinct game types.
See the linked YouTube video showcasing the project, ML agent behaviour in the different simulations, and a project development discussion.
Features:
- Three Game Demo Environments (Hummingbirds, Karting, Both)
- Multiple Individual-game Trained Machine Learning Agents
- Shared-brain Trained Machine Learning Agents