Subscribe to:
Post Comments (Atom)
skip to main |
skip to sidebar
Experiments in SLAM, MSRS, and vision
Researchers
Software
Archive
Blogroll
Who?
- Chris Kilner
- Coder in the land of cheese and wine.
Yada yada
3 comments:
Could you explain what the parameters rerr[1.3] and verr[1.3] mean. I know they model the noisy odometry but which values are affected by them exactly?
Thx.
Hi Jason,
they should all be variances for gaussian distributions of the VelocityMotionModel as described by Thrun et al in Probabilistic Robotics p.123
http://robots.stanford.edu/probabilistic-robotics/corrections/pg122-139.pdf
Before a batch of sampling, I use them to calculate the range of errors:
// translation error
_v_range = tErr1 * Math.Abs(_vel.Translation * _time) + tErr2 * Math.Abs(_vel.Rotation* _time);
// rotation error
_w_range = rErr1 * Math.Abs(_vel.Translation * _time) + rErr2 * Math.Abs(_vel.Rotation* _time);
// random error
_g_range = tErr3 * Math.Abs(_vel.Translation * _time) + rErr3 * Math.Abs(_vel.Rotation* _time);
where each of these relates to an item after the comma in Thrun's text.
I can't say I have a deep feel for the parameters. Anything but miniscule values makes the range far too large. I don't think they are particularly physical - just bucket loads of uncertainty to blur the overcertainty of modeling and sensing :)
OK, thank you.
All I wanted to know was where they influence calculations, because the names did not really help while guessing :)
Post a Comment