Skip to content

rdot-lee/DeepRacer

Repository files navigation

AWS DeepRacer Reward Function

I participated in the "Invent 2018" contest of DeepRacer Virtual Circuit.

It is a reward function at that time. The lap time was about 8.464 seconds after repeated training.

Environment simulation

  • re:Invent 2018 Training

Action space

Action space Value
Maximum steering angle 30°
Maximum speed 2.4 m/s
2.4 m/s
14° 1.6 m/s
20° 1.6 m/s
21° 1.7 m/s
25° 1.7 m/s
30° 1 m/s
-10° 1.9 m/s
-20° 1.5 m/s
-30° 1.6 m/s

Hyperparameters

Hyperparameters Value
Gradient Descent Batch Size 32
Entropy 0.01
Discount Factor 0.999
Loss Type Huber
Learning Rate 0.00006
No# Experience Episodes between each policy-updating iteration 20
No# of Epochs 3

Stop conditions

  • 30 or 60 mins

Reference

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages