-
Notifications
You must be signed in to change notification settings - Fork 1.2k
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
Can MPPI follow the path with time stamps? #4155
Comments
At this moment, no, but MPPI is pluginized for a reason, you could easily create a new PathFollower/PathAlign critic that takes into account the timing in your pose messages. The MPPI samples themselves have a time forward, velocity, and pose information, so if you path contains the timing information, you'd have all of the information required in front of you - you just need to develop the algorithm to do so. I think this would also be a great option to provide to MPPI users in the future as well, so if you created/adapted a path following critic for using path timestamps, I'd be more than happy to merge, review, and maintain it in Nav2 itself! |
Sure! I'll work on it these days, and when it passes our case, I will submit a merge request ;) |
@Chortine any updates or problems? 😄 |
Indeed, I'm facing some problems. I have implemented my first version of time_critic, the car now drive faster than without this critic. However, it starts zig-zag around the path, instead of nicely aligning with the path like before, specially when the curve of the path is large. I'm currently working to solve this. Do you have any suggestions? Screencast.2024-03-28.14.17.19.mp4 |
Its hard to say without looking at the code and the video is corrupted so its not viewable on my side. I'd suspect you need to run the controller faster to better converge |
Hi, Here is my implementation of time_critic.cpp:
And the controller parameters are:
And I reupload the video, where the car zig-zag when follow the path. Screencast.2024-03-28.14.17.19.mp4 |
This looks like just the path alignment critic -- what is substantively different about this from the current path align critic? Wouldn't you rather be looking at the velocities of the trajectories instead? if you have some sequence of poses in a path with a timestamp, you can find the velocity between them that enables that easily. Then the trajectories you have their velocities, so it should be an easy subtraction to find the difference and then sum those up to average for each trajectory sample 😄 Also, I'd recommend pulling in the latest changes I made (after 45% increase in performance PR), some of the utils that you're using were changed. I don't recall if I fixed any bugs in them, but I think at least one had a bug that was resolved (the rest are just faster now). |
Hi team,
I have a global path planner that outputs the points with time stamps, and the car is supposed to be a the pose at the suggested exact time. I would like to know that if MPPI is able to do such local planning with strict time critics.
I look into the PathFollowCritcs and PathAlignCritics, however, there seems to be no items encourging 'being at the pose at time'. Maybe I've missed something. Can we extend MPPI to do that? Tons of thanks.
The text was updated successfully, but these errors were encountered: