I want to implement the navigation in a two wheeled robot with the ROS navigation-package using Kinect , How can I calculate the odometry if I have the radius, the wheeled distance, the poses and the velocity of the wheels???

Originally posted by salime on ROS Answers with karma: 21 on 2013-07-10

Post score: 0

Odometry can be calculated by estimating "forward kinematics" of mobile robot. Here is a link for one such explanation http://www.cs.columbia.edu/~allen/F11/NOTES/icckinematics.pdf

Hope this helps...

Originally posted by ayush_dewan with karma: 1610 on 2013-07-10

This answer was ACCEPTED on the original site

Post score: 0

Comment by salime on 2013-07-12:
Thank you... it helped me a lot, but just one question Given that vr= velocity of the right wheel, vl=velocity of the left wheel, R=the signed distance from the ICC to the midpoint between the wheels...

When vr=vl, R becomes infinite and there is no rotation, so how can I manage this situation?

Comment by ayush_dewan on 2013-07-12:
When vr=vl, the robot should move straight, as you mentioned there will be no rotation.

Comment by salime on 2013-07-19:
So, is it ok for the navigation if I keep receiving nan (not a number) values in the equations, I have tried asigning a really big number when I get nan, but of course in some cases this won't work because of the overflow. Do you know how can I control this part?

Comment by ayush_dewan on 2013-07-19:
I am not sure about this, but one thing you can do is not calculate R when difference between vr and vl is very less.

Comment by salime on 2013-07-22:
Thank you for your reply, I did that, the R thing, and now looks like the nan is gone, but the final pose (x,y, theta) keeps getting bigger and bigger when I start moving the robot, even though when I stop moving it.. it keeps getting bigger and bigger in the x, y and theta, and of course these values are not correct for the pose

Comment by salime on 2013-07-23:
I have done some tests in the particular case when vl !=0 and vr != 0, incrementing the x, y, ant theta just in these cases, but the results do not look correct either