-
Notifications
You must be signed in to change notification settings - Fork 80
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
add prior constraint #18
Comments
Hey @weisongwen! I assume you mean (unary) priors on individual poses? If so, then you're in luck :-)! It is actually possible to express these priors directly in the model that SE-Sync already employs: you can simply introduce a single distinguished pose x_w that represents the "world" coordinate system that you want to use, and then express priors as relative measurements against x_w. To transform the estimates returned by SE-Sync back into this distinguished world frame when you're done, you'll then just need to multiply everything by x_w^{-1}, where x_w is the estimate for the "world" pose returned by SE-Sync. Note that while this construction may at first glance appear somewhat contrived, it's actually a straightforward mathematical reflection of the fact that there is no such thing as an "absolute" coordinate system. That is, even when we are using measurements that we typically think of as "absolute" (e.g. GPS measurements of position), what we are actually doing is measuring position relative to a commonly-adopted "standard" reference frame (in the case of GPS measurements, the lines of latitude and longitude that are laid out on the surface of the Earth). The construction that I described above simply expresses this fact in a more concrete way than we typically do :-). |
Closing this issue, as I've not heard anything further. |
@david-m-rosen , Thanks for your detailed reply and sorry for my late reply. Yes, I mean the unary constraint, such as the GPS positioning. I am still looking into the algorithm as it is really quite tough to fully understand. I will let you know if I have any updates based on your comments. :-) |
Hey @weisongwen, No worries -- I just want to stay on top of any reported issues to keep the repo in good working order :-). A couple of additional comments that might help to clarify the discussion:
Hope this helps :-)! |
Hi @david-m-rosen , Many thanks for your detailed explanation. Based on your explanation, my understanding is that: a higher-dimensional manifold is employed for searching in SE-Sync to approach the global optimal. I highly believe that your work can have a pretty deep impact on the state estimation field. Currently, I am doing collaborative positioning based on camera, imu, GNSS, and inter-ranging measurements from multiple agents. I will try to compare your SE-Sync to this case. I will let you know if I got any interesting findings, Thanks a lot :-) |
Hey @weisongwen, That's the right intuition. If you'd like an alternative take on this, you might also be interested in checking out the video for our recent paper that shows how to adapt the SE-Sync approach to solve rotation averaging problems using GTSAM -- it has some nice visualizations in it that attempt to convey what the algorithm is doing a bit more transparently :-). And please do keep me posted on the project! It's always super fun to see how folks are using this :-). |
@david-m-rosen , Many thanks for your video and paper link. This is really very very interesting. Sure, I will keep you updated! Best, |
@david-m-rosen Thanks for your great work. is it possible to add prior constraint (in this case, the optimization is formulated as a MAP-maximum a posterior problem)?
Best,
Weisong,
The text was updated successfully, but these errors were encountered: