This repository is a two weeks playfield who has seen different branches of ml-agents.
There's a lot of legacy code and scripts that need to be rewritten.
Consider coming back in a week or so :)
- (-1.0f/maxSteps) at each step
- +1.0f if Blue reaches Yellow
Action Space (Continuous - 2 actions):
- Horizontal direction
- Vertical direction
The resulting Vector is normalized then scaled by speed. State Space (Continuous - 6 states):
- Relative position of Yellow
- Yellow velocity
- Own velocity