[RL-baseline] Model v3, experiment #3 #36
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Policy network is almost identical to v2, but we removed the final FCL from the main network and added a new one to both the actor and the critic to allow them to learn more.
We also tweaked the actions compared to the experiments with v2. The main difference is that a new no-move action was added to allow the network to use the car's momentum. Here is the action list for this experiment:
[0.0, 0.0, 0.0], # no action
[0.0, 0.8, 0.0], # throttle
[0.0, 0.0, 0.6], # break
[-0.9, 0.0, 0.0], # left
[0.9, 0.0, 0.0], # right
Around the 10k episode mark, the entropy collapsed and never recovered. Running reward managed to stay positive up until past the 15k mark, where it finally collapsed to very low numbers. Results are below:
Sample video below:
https://user-images.githubusercontent.com/1465235/112736341-4b660700-8f52-11eb-9d19-a1b73e30ec89.mp4