AWS DeepRacer Reward Function

I participated in the Kumo Torakku contest of DeepRacer Virtual Circuit.

It is a reward function at that time. The lap time was about 25 seconds after repeated training.

Environment simulation

• Kumo Torakku Training

Action space

Action space	Value
Maximum steering angle	30 degrees
Steering angle granularity	7
Maximum speed	8 m/s
Speed granularity	3

Hyperparameters

Hyperparameters	Value
Gradient Descent Batch Size	64
Entropy	0.01
Discount Factor	0.999
Loss Type	Huber
Learning Rate	0.0003 or 0.001
No# Experience Episodes between each policy-updating iteration	20
No# of Epochs	3 or 10

Stop conditions

• 60 or 120 mins

Reference

• https://forums.aws.amazon.com/forum.jspa?forumID=318
• https://github.com/aws-samples/aws-deepracer-workshops
• https://github.com/sasasavic82/deeracer-reward
• https://qiita.com/kai_kou/items/8a45c687baca8c9465f6
• https://qiita.com/Alt_Shift_N/items/2c37fbb26d739b7f3046