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.

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

faster robot reset #59

Open
Armandpl opened this issue Jan 19, 2024 · 2 comments
Open

faster robot reset #59

Armandpl opened this issue Jan 19, 2024 · 2 comments

Comments

@Armandpl
Copy link
Owner

When an episode terminâtes if the motor speed was high the motor continues freely spinning for a second or so. This slows down the whole training as we have to wait for it to stop spinning then wait for the pendulum to come to rest.

  • We could use a pid controller on the speed to slow down the motor faster
  • then maybe help the pendulum oscillations to stop?
  • unsure about the quality of our velocity filter so maybe having its output be a target is bad?
@Armandpl
Copy link
Owner Author

Armandpl commented Jan 31, 2024

Ok so I quickly set up a pid to slow down the motor but it may be too violent and I think it damaged the motor. I should look into this before the next real training, make sure we gently slow the motor down!
We can probably use the speed limits so that the gains 'make more sense'. E.g divide the gains by the max speed, so that Kp = 1 equals to an action of 1 if the motor is at max speed.
One thing I don't like is that the action starts at the maximum, i'd like it to slowly ramp up to gently decelerate? how do

@Armandpl
Copy link
Owner Author

Armandpl commented Feb 2, 2024

I just need to compute a command based on the current speed an how long I want to take to slow down the motor. Instead of having the command be zero which makes the first action 1 and instantly reverses the polarity.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant