Newsgroups: comp.robotics
Path: brunix!uunet!usc!howland.reston.ans.net!torn!newshost.uwo.ca!cogsci.uwo.ca!mheim
From: mheim@cogsci.uwo.ca (Michael Hiemstra)
Subject: Re: Micro-coding Robots?
Organization: University of Western Ontario, London
Date: Fri, 23 Jul 1993 22:02:43 GMT
Message-ID: <1993Jul23.220243.2126@julian.uwo.ca>
References: <22ooc0INNe10@uwm.edu>
Sender: news@julian.uwo.ca (USENET News System)
Nntp-Posting-Host: cogsci.uwo.ca
Lines: 34

In article <22ooc0INNe10@uwm.edu> rick@ee.uwm.edu (Rick Miller) writes:

>I originally concocted this idea to implement neural-net behavior.  Since
>digital neural-net computations are quite intensive, I figured I'd train
>the network simulation, then feed it all possible inputs while storing the
>outputs as data on the memory chip.  Thus, the chip would *behave* exactly
>as the neural-net... but instantaneously, without run-time computation.

I've been pondering ideas of a neural-net controlled robot too.  The
problem with micro-coding a neural net is that it defeats the purpose
of the net (i.e. an ability to adapt to situations as they occur).

There's also a problem with pre-computing the control algorithm for the
robot.  Simulated input and output won't be anywhere near the
real-world conditions that the robot will encounter when it's running.

In college a group I was in made a micro-coded robot.  It didn't work
so hot.  We had latches that would record the status of four
micro-switches for the last three ticks of it's timer.  These were
used as input to an EPROM that we <hand-coded!>, and the output
directly controlled the motors.  It was slow, and didn't react well.
But, with modifications, it's not a bad idea.  A faster clock speed,
outputs feeding back in.  You're right, that it wouldn't have to do
any calculations, so it could be fast if done right.

>RICK MILLER            <rick@ee.uwm.edu> Voice: +1 414 221 3403 FAX: -4744
>16203 WOODS            Send me a postcard, and I'll return another to you!
>53150-8615 USA         Sendu al mi bildkarton, kaj mi redonos alian al vi!


-- 

Opinions are my own -- mheim@cogsci.uwo.ca (Michael Hiemstra)

