Phil and John,
Thanks for your suggestions.
I decided to try Red Hat 6.0 for many reasons: first it allows
multi threads (it seems that even gdb was patched to support multi-
thread debugging mode), it allows swap files (partitions) of more
than 128Kb (the new kernel 2.2.xx does it) without any extra headache,
and it is still supported by RTL and RTAI (or Zentropix, as John
mentioned).
I haven't finished upgrading yet, but I have to say that I like it
a lot so far. If you haven't tried RedHat yet (I hadn't), you'll be
amazed by how easy it is to install and how neat it looks (thanks to
KDE and GNOME, both quite impressive).
Well, anyway... this is not a Linux mailing list, so let's talk
about what is important. (Phil asks about my application...) We have
developed (and are still developing) systems for vision based mobile
robot (indoor) navigation using Linux and RTLinux. I'll be glad to
provide details about our navigation systems - or you can check our
web page at www.ecn.purdue.edu/RVL (though the system discribed there
does not use Linux) - but for now, our robot has a stereo head with
three cameras (a pair for stereo vision and a third for color images),
five sonar sensors for obstacle detection/avoidance, and a laser range
scaner also for obstacle detection/avoidance. The software for controlling
the stereo head and the sonar were both simple unix processes and that
gave us a lot of inconsistent behaviors, specially with the readings
from the sonar sensor.
What I did recently was to write two rt tasks, one for controlling
the stereo head (this one is pretty much the same as in some examples
found in the RTL distribution for serial port, since the head uses
serial communication) and the other for the sonar. The latter uses the
parallel port for triggering one of the five activation signals that
go to each of the sonar trasmiters and "loop" while it waits for the
status bit to go high - echo received by the corresponding sonar
receiver.
Both rt tasks improved significantly the system performance, but
the latter one (sonar) proved to be even more important for our
application. First it allowed us to remove some special hardware
we had before to cope with the nondeterminism of soft interrupts,
and second it gave us a much more accurate measurement.
Well, as I said, I'll be glad to give more details. Just let
me know.
Guilherme
>Date: Tue, 21 Sep 1999 10:54:19 -0700
>From: Phil Wilshire <[EMAIL PROTECTED]>
>Subject: Re: [rtl] NMT vs. RTAI
>
>John Ahrens wrote:
>>
>> I have a copy of pre-release 2.2 Realtime Linux from Zentropix:
>> http://www.zentropix.com based on RTAI. I haven't used the realtime
>> portions or the tools yet, but it is based on RH 6.0. Might be worth
>> checking out..
>>
>
>
>I think there are still some issues under SMP with the 2.2 kernels.
>This is not a RT Linux issue but RT linux seems to trip the problem.
>
>The UP systems look to be more stable.
>For UP both NMT and RTAI are great performers.
>
>In either case get the latest from RTAI or NMT.
>
>Good luck
>
> Any more you can share on your application ??
>
> Phil Wilshire
-------------< G. N. DeSouza >-------------
---------< [EMAIL PROTECTED]>---------
--< http://rvl1.ecn.purdue.edu/~gnelson >--
--- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
----
For more information on Real-Time Linux see:
http://www.rtlinux.org/~rtlinux/
