DPRG mailing list archives: [DPRG] Processing unit for mobile robot

Wed Apr 23 13:35:29 CDT 2008

Hi! Different Ed here for anyone that might get confused.

On Wed, Apr 23, 2008 at 1:09 PM, Randy M. Dumse <rmd at newmicros.com> wrote:
>  So! Maybe you want to expound on Linux a bit, and tell us 1)
>  what is the real nature of the realtime limits of Linux and 2)
>  Why Linux anyway?

I don't know much about (1) but I can try to answer (2).

The big reason I use Linux on my robots are these:
- carry your development environment around with you
- 'advanced' networking (such as HTTP, SSH, or WiFi)
- ease of programing due a lot of RAM, a lot of disk space, a lot of MIPS etc
- ability to directly use libraries born out of research projects,
which tend to be written for desktops (OpenCV is probably what I'm
thinking of here)

There are a lot of reasons NOT to use Linux (especially nothing but a
PC running Linux), and if you can't think of any reason why you'd want
to use Linux already, good chance is that you don't need to bother.

I really like the "PC running Linux with real time stuff offloaded to
a microcontroller".  In my opinion, it neatly solves the real time
difficulties while leveraging the above benefits of Linux without
introducing a large amount of development overhead.  Typically you'll
have two build targets: say Atmel Atmega and Intel x86.  And in this
case, you are able to use very similar tool chains for both ( GCC and
the Unix stuff).

Hope that helps!  I'm not a zealot, and I like 'the right tool for the
right job'.   But I've done a few robots in this manner over the past
few years and I've not identified any problems with it yet.  I guess
the one thing I never solved very well was high speed video analysis.
It seems that a PC doesn't have the bandwidth, and you'd really need
to have some custom/exotic/something hardware to do 30fps computer
vision.  Lower frame rates worked fine.

Ed P. ( right in line after Ed O. )