Kobuki Probabilistic Motion Model

[Sterling McLeod]

Hi,

I'm looking for the motion model to use for localization with the Kobuki base.

I teach an undergraduate robotics course on autonomous navigation. I use the Turtlebot 2's as a platform for the class. I would like some students to implement a Kalman filter algorithm with the Turtlebot 2. However, that algorithm requires probabilistic motion and measurement models, which I don't know. I've tried looking for them online, but can't seem to find anything. Can someone here point me in the direction of such an estimation model for the Kobuki's motion?

[Scott C. Livingston]

A decent approximation is a differential drive dynamics. E.g., consult
Section 13.1.2.2 "A differential drive" in the book Planning Algorithms
by Steven M. LaValle (2006).
http://planning.cs.uiuc.edu

The ODEs include trigonometric functions, so the basic Kalman filter is
not applicable. E.g., you could use an extended Kalman filter as
developed in the book Probabilistic Robotics by Thrun, Burgard, and Fox
(2006), Chapter 3.

~Scott

[Yinon Oshrat]

You can look ay the code for ros and see the kalman filter implementation there

--
Kobuki's nest: kobuki.yujinrobot.com
---
You received this message because you are subscribed to the Google Groups "Kobuki Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kobuki-users...@yujinrobot.com.
To post to this group, send email to kobuki...@yujinrobot.com.
Visit this group at https://groups.google.com/a/yujinrobot.com/group/kobuki-users/.

[Daniel Stonier]

Kobuki provides pose, twist and covariance to the robot_pose_ekf node. You can see the relevant code in the ros driver here. As far as I know the guys working on this have done nothing special with regards to this - it was just a rough estimation of the kobuki's parameters to get things started and was sufficient for amcl localisation to work.

Cheers,

Daniel