Unscented Kalman Filtering?

Question

I want to use an unscented kalman filter (UKF) in ROS to do state estimation. I can't find any ROS packages that implement an UKF so I've been looking around at filtering libraries for C++ that I could use within ROS. Does anyone have any recommendations? I've found the following libraries but many are poorly documented or inactive. Any suggestions?

UKF support:

Bayes++

bfilt

No UKF support:

Mobile Robot Programming Toolkit (The main page lists UKFs but the code doesn't appear to implement them)

Orocos Bayesian Filtering Library

KFilter

---

Originally posted by Andrew Chambers on ROS Answers with karma: 96 on 2011-12-21

Post score: 5

---

Original comments

Comment by Martin Peris on 2011-12-21:
Hi Andrew! OpenCV implements Kalman Filter, Extended Kalman Filter (EKF) and is fully integrated with ROS. But I am afraid it doesn't implement UKF (as far as I know).

Answer 1

Since you already know you want to do an unscented KF, I assume you understand the mathematics. I would suggest just implementing the math yourself using the matrix/vector capabilities of either OpenCV or Eigen libraries which are already part of ROS.

---

Originally posted by Kevin with karma: 2962 on 2011-12-29

This answer was ACCEPTED on the original site

Post score: 4