[Playerstage-users] Fwd: player-stage bug?

[Geoff B. <gb...@ki...>]

Hi Lantao Liu,

I have forwarded your message to the playerstage-users mailing list. I 
recommend you learn how to use the mailing list, as it is the best 
source of help when you are having trouble with Player. You can sign up 
for the mailing list here:

https://lists.sourceforge.net/lists/listinfo/playerstage-users

As for your question: the output will be different if there is an object 
where you propose. If you turn off the "filled" option in playerv, you 
will be able to see this. A laser scanner only returns point values, not 
filled regions. If an object is in the position you propose, you will 
get scan returns along the object, not free space.

Geoff

-------- Original Message --------
Subject: player-stage bug?
Date: Sat, 19 Dec 2009 17:55:58 -0600
From: Lantao Liu <la...@cs...>
To: Geoffrey Biggs <gb...@ki...>

Hi Geoff,

I am now a user of playerstage and currently I am encountering a
trouble. I don't know how to use the BBS/platform for problems shooting,
but I got your email from
http://article.gmane.org/gmane.science.robotics.playerstage.user/6173

and I hope I can get some help from you.

My problem is in the AMCL using Hokuyo laser. The hokuyo_aist driver can
only provide interface 'ranger', while the amcl driver requires 'laser'
only, therefore, I used the 'rangertolaser' to obtain a 'laser'
interface.

Now there exists the problem, although I am not 100% sure if it is
because of laser/ranger. More formally, the amcl can work well in a
small environment (4m X 2m dimention, where the laser can always
touch/sense obstacles for every beam), but can not even localize in a
long corridor (2m X 20m dimention, where some beam can not touch
obstacles/walls) in our building, and the localization particles *only*
exist in obstacles!.

The reason I am doubting is: in rangertolaser driver, it may lose some
information when do the data conversion, please see the 1st attached
pictures as example (I converted the interface as follows in my cfg
file: ranger->rangertolaser->lasertoranger):
	left: original hokuyo ranger output in playerv (ranger:0)
	middle: rangertolaser output (laser:0)
	right: lasertoranger output (ranger:1)

The same data, but different between left and right pictures, which are
using the 'ranger' interface, but just after driver conversions...

Back to my amcl problem, and the reason so far I can explain is that:
The rangertolaser driver can only return the ranges which touching
obstacles, and disregard all the others, such as the free space. While
the problem is(Please see my second attachment first):  suppose there is
an obstacle(right figure) exactly 1cm in front of the laser which is
placed in a corridor, it could also block the sensing and output the
same result as the situation with nothing in front(left fig). If so, it
is reasonable that the robot can only exist in obstacles and
particularly around the walls.

I attempted to modify some possible places in rangertolaser.cc but did
not work, and I am not familiar with the driver stuff.

That is so far what I think. And of course maybe other reasons...Could
you please help me on this issue since I have been stuck here for
several days...

I appreciate a lot for any of your suggestions.

Regards,

Lantao Liu

PhD student
Texas A&M University