[RL-baseline] Model v2, experiment #1

[ziritrion]

Second of the new experiments with the new action sets (5k episodes) and Learning Rate 1e-3-

The model managed to get a max Running Reward of ~500. Last episode managed 257.

Action set #1 copied below:
[0.0, 0.8, 0.0], # throttle
[0.0, 0.0, 0.6], # break
[-0.8, 0.0, 0.0], # left
[-0.5, 0.0, 0.0], # left
[-0.2, 0.0, 0.0], # left
[0.8, 0.0, 0.0], # right
[0.5, 0.0, 0.0], # right
[0.2, 0.0, 0.0], # right

Results are copied below:

Sample video below:
https://user-images.githubusercontent.com/1465235/111862476-d133f200-8955-11eb-8482-f223631c8e0f.mp4

[ziritrion]

Updated with 20k runs. Running reward is now at 337, with an achieved max of 556.

Sample video below. It does not brake before sharp turns.
https://user-images.githubusercontent.com/1465235/112334548-cd88ce00-8cbb-11eb-9c0f-de586b3d9a89.mp4