[ros-users] AMCL configuration details required
gerkey at willowgarage.com
Mon Jul 12 17:46:27 UTC 2010
On Thu, Jul 8, 2010 at 11:26 PM, Prasad Dixit
<prasad.dixit at fennecfoxtech.com> wrote:
> I have boxturtle package in which AMCL package
> /navigation/amcl/example/amcl_diff.launch has got below configuration:
Let me say first that that launch file has more parameters set than
are really necessary, which is clearly causing some confusion. I've
ticketed myself to clean them up:
> 1. If Laser model type "Beam" is commented then why laser_z_short,
> laser_z_max parameters are active?
They're not used in this case.
> 2. "laser_likelihood_max_dist" parameter depicts likehood model in use.
> Considering entire nav stack is costmap_common's parameter
> "inflation_radius" should be same as likehood_max_dist?
No. While they perform a similar computation, it would be unwise to
couple the likelihood max distance, which affects localization, to the
inflation radius, which affects navigation behavior.
> 3. Which is model can give accurate measurement beam or likehood?
It depends on your environment. We almost always use the likelihood
field model. While it lacks a strict probabilistic interpretation
(see the Probabilistic Robotics book for more on this), it tends to be
more reliable, especially in environments with small obstacles. Also,
the likelihood field model is more computational efficient than the
beam model, which requires ray-tracing at run time (the likelihood
field model pre-computes the field at startup, so that the run time
work is simply a table lookup).
> 4. Translational/ Rotational movements are required before performing a
> filter update by update_min_d and update_min_a. Are values against these
> parameters (0.2 and 0.5 resp.) in meters? are these parameters active
> because we have not given initial position?
These parameters are always obeyed, whether an initial pose is given,
or not. As specified in the documentation
(http://www.ros.org/wiki/amcl#Parameters), update_min_d is in meters
and update_min_a is in radians. Throughout ROS, we use MKS units
> 5. transform_tolerance is specified twice with two different values.
The latter one is used.
More information about the ros-users