I have only short experience testing amcl. What you are describing happens even in simple environments if the numbr of particles is set too low. Once the algorithm converges to a wrong estimated pose it can be hard to converge to a correct one. Unless you have a wrong setting in scale / resolution parameters of .world and .cfg files, this behaviour is normal with this algorithm.

You could try:

 -increase number of particles: e.g. parameters
  pf_min_samples 10000      pf_max_samples 100000

 -define (init_pose, init_pose_var)

Best Regards,


Roberto Guzman
Departamento de Ingenieria
Robotnik Automation, SLL
C/Berni y Catala, 53 bajo
Tel:       +34 963383835
Fax:      +34 963383580
E-mail:  rguzman@robotnik.es
Internet:  http://www.robotnik.es

2010/8/30 Tiago Veiga <tsveiga@isr.ist.utl.pt>
Hi all,

I'm a new user starting to introduce myself in the world of Player and
I'd like to ask to those of you who have more experience about the
localization driver.

When I start player, the particles of amcl are all spread on space as
suppose, and I've also already increased the variance on the initial
distribution. What I found strange is that once the robot starts to move
the particles converge very quickly (let's say in a few seconds) and
very often to a wrong position. Afterwards, even if I do some random
walking, it's very difficult for the localizer to converge to the real

I found this behavior strange, as I think particles should not converge
almost immediately once I start moving. I've tried in different maps, in
real and simulated robots, with different limits on the maximum and
minimum number of particles, and it doesn't seem to change.

Am I missing something here or may it just be the way the algorithm is?

Thanks in advance.

Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users
worldwide. Take advantage of special opportunities to increase revenue and
speed time-to-market. Join now, and jumpstart your future.
Playerstage-users mailing list