I wanted to explore Machine Learning this jam so I taught Mr. Pinky to walk by himself. You control the level obstacles to help him home.

More specifically I'm using ML Agents for Unity, with imitation learning so it learns based on recordings of human playing. I've been working with ML agents before but not a lot with imitation learning so this was a good chance to learn it better! Really happy that I managed to get the AI working. After solving quite some bugs with it, I still managed to pull off the rest of the game, just not with as many levels as I hoped.

Click to Play (HTML5)

Training:

Final Result: