I generated a trajectory for a point target by using coordianted turn model with a turn rate w = 0.08 rad/s ~ 4.6 degrees/s. When i use extended kalman filter with coordinated turn model, the estimated turn rate converges to 3.8 (~ 0.066 rad/s if estimate is in degrees). If I accept estimated turn rate is in degrees, my filter is OK. I just need to convert deg2rad. Any ideas?
Can you clarify your question a little bit? If you are asking what is better: filtering in the space of radians or degrees, I would say they are equivalent since one is just a scalar multiple of the other. I have sometimes observed the when working with really small angles, using radians can result in floating point errors. However, when working with angles, I would use a Bingham filter which models the uncertainty in the angle using a Bingham distribution as apposed to an ekf that uses gaussian distribution. Such an approach would help reconcile with the fact that theta=they+2π
Dear Aykut,
I suggest you to see links and attached files in subject.
- Interacting Multiple Model Extended Kalman Filter for Tracking Target ...
https://www.researchgate.net/.../44939745_Interacting_Multiple_Model_Extended_K...
- An Improved Interacting Multiple Model Filtering Algorithm ... - MDPI
www.mdpi.com/1424-8220/16/6/805/pdf
- Models and Algorithms for Tracking Target with Coordinated Turn Motion
https://www.hindawi.com/journals/mpe/2014/649276/
- Bearing only Tracking of Maneuvering Targets ... - Semantic Scholar
https://pdfs.semanticscholar.org/.../ab775a24bbf693f8a0ec19d63a...
- MS Thesis - UPCommons
https://upcommons.upc.edu/.../MS%20Thesis%20-%20Furrukh%2...
Best regards
I solved the problem. Use of an inappropriate Q matrix has caused bigger turns.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Computer Graphics and Animation
- 3D