My reinforcement learning models + reward functions for beginners to understand Amazon DeepRacer a bit more.
My first model tested on actual race track at Istanbul IX.0 Accenture offices and it performed stable enough.
My 11 second virtual track record is not perfectly stable in actual environments due light and color saturation of physical track. (not every track is good as reInvent event tracks bear in mind)
I included both reward functions + model files for you to test out. Any question can hit me with e-mail.
A small clip from actual track at event https://www.youtube.com/watch?v=efZzpvkNoQ8&feature=youtu.be Both model files have their enviroment details ,speed selection , angle selection etc listed in their respective files.