Rolling Ball AI

Here is an example of an implementation of ML-Agents in Unity.

How it works:

ML-Agents use Machine Learning to make the Agents become good at a specific task.

The Agent in this project uses physics to roll towards a specific location. The agent gets the location of himself and the target as an observation. Through Machine Learning the Agent becomes better and better at this task and eventually after some time of learning it can basically do its task without failures.

In the 'trailer' you can see the agent in its learning progress.

Windows can give a warning when you try to run the .exe file. This is because the file isn't downloaded very often. You can still run the program by clicking 'more info' and click run anyway.