On 19/08/10 12:14, DRamachandran@hra.com wrote:
>    So I have a rudimentary version working now. I downloaded the hokuyo
> sample code from
> here:
> 
> http://www.hokuyo-aut.jp/02sensor/07scanner/download/urg_programs_en/index.html
> 
> and adapted md_scan.c so that it outputs the data to a file which I then
> read from within a ros node.  Kind of hackish, but the problem was that
> the c++ version didn't compile on my machine (linker errors within SDL
> (?) ). 
> 
> I'm not sure if what I have is enough however. I guess the ros drivers
> do some postprocessing on the scan data before outputting, which I
> don't. I tried integrating with some odometry for my robot (which I know
> is fairly accurate) to do simple localization in AMCL and the particles
> jump all over the place and then the covariance blows up very fast to
> NaNs. Is this what you would expect with unfiltered laser data?
> 
> What should I do next? I can try integrating this code into the ROS
> hokuyo driver. It should be as simple as replacing the fopen with the
> setSerial code in the hokuyo sample code. I think that be best done by
> someone who knows what he is doing.


Out of curiosity, does the hokuyo_aist library have the same problem on
the Mac that the ROS node does?

http://gearbox.sourceforge.net/namespacehokuyo__aist.html

Install it and try running the hokuyo_aist_example binary. (You should
consider using the SVN version, it's got a vastly updated API with many
new features, including support for the new scanners due to be released
at the end of the year).

If that library works for you, I could be convinced to wrap it up in a
ROS node this week.

Geoff