-
Notifications
You must be signed in to change notification settings - Fork 123
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
What are reasonable selections of lag_duration for state estimation? #262
Comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hi,
I'm watching your ROSCon talk and had a question on the
lag_duration
for the fixed lag smoother. It makes sense to keep a rolling buffer of measurements to bound the problem, but it doesn't tell us what a "reasonable" set of values for this are.Your example shows
0.5
seconds, but if someone asked me in a vacuum, I would have thought something closer to 1-2 seconds. There's clearly a trade off for how often you cull out measurements to bound the graph size and how long you keep it for accuracy, so I was hoping you could shed some light from your experience what a reasonable range of values are for a usual setup (IMU + wheel ecoders, each publishing 50-100hz) from tuning your systems. This doesn't have to be very scientific, I'm really looking to query your intuition from building this project, though if you had actual numbers, that's good too 😆(e.g. is 5s just insane because of the number of constraints in the graph for a typical robot CPU, or is 0.5s an actual recommended value or just a toy example for the talk, where's the rough point of diminishing returns on accuracy vs buffer length, etc)
A reasonable follow up question would be if there's a better trade off on lowering the frequency of optimization to allow more samples to come in during a time period and then increasing the
lag_duration
to make use of the extra compute time for improved accuracy. I wonder about the accuracy this can obtain if being run at 20hz with usual robot odometry / IMU at 50-100hz with respect to long term drift relative to an EKF.tl;dr
The text was updated successfully, but these errors were encountered: