Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fix convergeness preliminarily #47

Merged
merged 1 commit into from
Sep 30, 2021
Merged

Fix convergeness preliminarily #47

merged 1 commit into from
Sep 30, 2021

Conversation

KelvinYang0320
Copy link
Member

I have fixed the following issues:

  • The robot will stop moving after the first episode.
  • The robot will kick the target (the yellow ball) away after the task is about to be solved.
    • changed the target from a yellow ball to our logo.
  • The agent is hard to converge.
    • rearranged the reward functions, reset conditions, action mapping(gas & wheel)

I have tried many different reward settings, but the best result I can get is this PR. Take a look at the showcase.
Therefore, I keep the ":warning: NOTE: This agent do not converge well" on readme and comment out the reward settings (1)~(5) instead of deleting the code.
Hope that someone can fix it further.

@KelvinYang0320 KelvinYang0320 linked an issue Aug 12, 2021 that may be closed by this pull request
[find_and_avoid] Fix convergeness issues #41
Open
@eakirtas
Copy link
Member

Indeed the agent converges better than before so we can merge the PR. But I recommend to now close the issue yet.

From problem's prospective I think that avoiding the obstacles without collisions should be an objective of this task. So it should be included in is_done method. In a way, if a robot crash to a wall, then the task could not be completed.

If you agree I propose to merge the PR and keep in mind the above restriction for future work.

@KelvinYang0320
Copy link
Member Author

@ManosMagnus Thank you for the advice.
Since it is find_and_ "avoid", so avoid should be included in is_done, i.e. one of the objective of this task.
I'm with you on that. 👌

Also, you said "now close the issue" or "not close the issue"?
I think we should not close the issue until the "avoid" part is done.

@eakirtas eakirtas self-requested a review September 30, 2021 08:14
@eakirtas eakirtas merged commit cc692ad into aidudezzz:dev Sep 30, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@eakirtas eakirtas eakirtas approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

[find_and_avoid] Fix convergeness issues
2 participants
@KelvinYang0320 @eakirtas