Lego Fruit Collector

This is a small experiment I made in 2020 with Unity's Machine Learning tool (ML Agents), using assets from the Lego Microgame.

You control a Lego minifigure and the goal is to collect fruit and drop them in the cage at the centre of the scene. You will compete against a trained AI agent, controlled by the CPU. Whoever stores more fruit in the cage wins. But remember to avoid touching the crabs, or you will lose points!

The game has no sound. It is the result of my simple introduction to ML Agents, so don't expect it to be perfect. The AI agent is not 100% accurate, but sometimes he sort of knows what he's doing. This was the best result I could get of around 12 different machine learning training sessions.

Hope you enjoy it!