goldbattle
commented
2 years ago To your first question, the fast propagation is just a copy-pasted code of the normal propagation since the existing code will modify the state estimate which is undesirable since for fast propagate we just want to take whatever the most recent state estimate is and predict what the pose is at the IMU rate quickly.
As for you second question, make sure you note that one is a quaternion and the other is SO(3). These are different matrix exponentials exp() that are in use here. They should be approximately equivalent to each other and interchangeable. These are found through integrating (in respect to time) the derivative of the quaternion or the rotation matrix. For example we walk through how to do this quaternion integration on the docs:
This can also be performed for the SO(3) matrix (Joan Sol`a report has some details too).
Note that we have a negative, which is the same as a rotation transpose
You can try for yourself if there is a different by comment this line in and comparing the two predicted values: https://github.com/rpng/open_vins/blob/d84a51c2b914b5739ed6eaff30eb74e7d6a687a3/ov_msckf/src/state/Propagator.cpp#L463-L471
Hello,
I'm trying to understand the IMU propagator and going through
Propagator.cpp. This comment and this comment are exactly the same but the code that follows each is quite different. Could you clarify which method is doing what?Additionally, this formula is the same as this one, but the one used here is different. Why is that, and which one should be used for extrapolating the pose into the future?
Thank you!