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

MPPI documentation inconsistency #521

Open
ARK3r opened this issue Feb 7, 2024 · 1 comment
Open

MPPI documentation inconsistency #521

ARK3r opened this issue Feb 7, 2024 · 1 comment

Comments

@ARK3r
Copy link
Contributor

ARK3r commented Feb 7, 2024

first line of https://github.com/ros-planning/navigation.ros.org/blob/master/configuration/packages/configuring-mppic.rst#general-words-of-wisdom strongly directs to never set model_dt > 1 / controller_frequency but in the example right above it, the controller frequency is 30.0 which suggests model_dt should be less than 0.33.

I won't make a PR for this since I don't know exactly if I should bring the controller frequency or model_dt lower.

@SteveMacenski
Copy link
Member

SteveMacenski commented Feb 7, 2024

That is general advice which is true. The discrepancy larger will impact performance the larger these values diverge. This advice is especially pointed to people trying to make it work at 10-15-20hz where the model_dts there are so huge the controller isn't super stable.

It still applies to the situation you rightly point out. But that was a tactical trade off between making the discrepancy low, yet having a high enough prediction time (model_dt * number of points), and a small enough number of points to work on a broad range of compute platforms out of the box.

You'd be right to say that if you're tuning and you have CPU you can spend, reducing that to 1/freq is a good potential move. The value add of much smaller after you've hit ~0.04 & at or lower than 1/freq is marginal (if any) and its better to optimize for other things like longer time horizons or more samples in your batches -- unless in exceptional situations

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

2 participants