The Robotics Engineering Task Force
Posted 4 May 2003 at 16:31 UTC by steve
Intel has recently lead the push to form a standards body to establish
software protocols and interfaces for robotics systems. The organization
is called the Robotics Engineering
Task Force and is modeled after the Internet Engineering Task Force
which sets the standards for the Intenet. The next step planned for the
group is forumlating their stance on intellectual property. Let's hope
they follow the lead of the Internet's developers here too and allow
only Free and Open technology to be used as standards. If they follow
that path, we may see an explosion of innovation and development in
robotics not unlike what we've seen with the Internet. A recent LinuxDevices.com
article talks about the RETF and other robotics developments at Intel.
Reusable code, posted 4 May 2003 at 22:36 UTC by motters »
Looks interesting. I'm all in favour of developing resusable software
modules for robots, since I've often seen the same code written by
different people over and over. We need open standards and preferably
open source for the most commonly used aspects of robotics software.
A good example is face recognition technology. This ability is very
useful for robots with vision and a reasonable amount of processing
power. There are a few commercial systems but they're not aimed at
robotics, and as far as I know there is no good face recognition system
which is open source and has an easy to use API.
Despite the high-ranking names I see on the RETF list, the thinking is
rather linear. For example the group does not, for the most part,
consider 8 bit controllers as control systems. There also seems to be
a tendency to consider software completely separate from hardware,
which in my experience is not necessarily a norm in mobile robotics.
At this point I see a kind of strange all-encompassing focus on
midlevel and lower communication protocols which do not include 802.11
and Ethernet protocols but does includes serial and I2C.
I realize the game is early for this group, so I shouldn't be too
critical, but I would have expected some high level conceptualization
first, followed by the generation of some standard patterns. I think
if that occurred you would discover that the linear thinking I was
referring to is derived from robotic projects at a university level,
which I view as a disjointed subset of the young (but growing) field of
But like I said, it's early in the game. I sure would like to see this
group focus on patterns first though.
I think they'll be using primarily 32 bit processors for this.
As they want to develop a Operating System for Robots. Then with
a standardized OS, they'll start developing drivers and objects to
allow the different hardware to all interface to the OS.
It looks to be University Driven mostly, so I doubt it'll have any
commercial success for a few years. In the commercial sector, they
expect results, low cost, and profits, which tends to be opposite
of what the Univerties do.
The big catch is that it looks like Intel is driving this, so we'll
likely see Microsoft jump in later with a commercial version of the
Robotics OS. Thus we'll need CPU's with 64meg of RAM/Flash or more to
run our robotics programs.