sfwa / ukf

Unscented Kalman Filter library for state and parameter estimation
http://au.tono.my/log/20130531-kalman-filter.html
MIT License
461 stars 172 forks source link

Rotation estimation #69

Closed Catsgomeowmeow closed 3 months ago

Catsgomeowmeow commented 4 months ago

Hey Guys, I'm very impressed by your UKF implementation ( and I also read about your accomplishments back in 2014).

I understand you're (from the looks of it) not maintaining this code base anymore. However, it is one of the most complete and easiest-to-use implementations of an SR-UKF that allows for dynamic measurement.

I'm trying to use it for one of my projects: sensor fusion of photogrammetry Position and Orientation and the IMU data coming from a camera. While the position estimation is working perfectly ( most of the time), I struggle with the orientation component. I've tried many things, including using the quaternion implementation and extending the measurement vectors to support quaternions. However, this failed to provide good results. I've also used simple Euler rotation; in this case, at least I can get the correct orientation when I have orientation data, but while the IMU updates, the orientation seems to drift extremely quickly.

I'm wondering if you guys have any insight you could provide me with to get rotations to be properly predicted with their constraints.