curvature calulation equation difference with paper

[hoppss]

auto_dock.cpp
while (!controller_.backup(0.0, -dock_yaw)

controller.cpp
double k = -1.0 / r * (k2_ * (delta - a) + (1 + (k1_ / 1 + ((k1_ * theta) * (k1_ * theta)))) * sin(delta));

hi,
I find the curvature calculation function is different with paper, is it a error ?
Another error, preorient a yaw in base_link, it should be dock_yaw , rather than -dock_yaw.
`

[mikeferguson]

I haven't looked at the yaw aspect, but I can confirm - there are missing parenthesis in the computation of the curvature (adding them almost certainly requires a retuning - but probably fixes some issues).

[erelson]

Thanks for the ticket! I verified that at least the controller.cpp equation is something that we've fixed in internally (that codebase has diverged quite a bit from the open source version and the original paper).

Not sure about the auto dock portion. Offhand, I'd probably use RViz to visualize the dock pose arrow while a research robot goes to dock. I'll keep that in the back of my mind to eventually verify. Feel free to let us know if you have more direct evidence handy!

Empirically, docking currently works fine as is, as far as I know. My intention is to try and get a PR open with the controller.cpp fix, referencing this issue, but no eta on merging. As noted above, retuning and testing would presumably need to be done, but is not a priority. Someday it would get tested and merged...

[CumaOzavci]

@hoppss, @mikeferguson, @erelson can you share the paper you are talking about?

[mikeferguson]

There is a comment in controller.h: Implements something loosely based on "A Smooth Control Law for Graceful Motion of Differential Wheeled Mobile Robots in 2D Environments" by Park and Kuipers, ICRA 2011". You can find a copy here.