DPRG mailing list archives: [DPRG] Re: Allen Robots (Will's emotions)

Tue Mar 6 19:14:04 CST 2007

Well my experience with the subsumption architecture was all hard wired
relay logic.  Very fast, very reliable, and easy to debug.  But basically
impossible to have any intelligent or optimized behavior take place.

A few years ago I did a very extensive literature search on this, and did
not find any one who had achieved any sort of optimized behavior.  The
closest anyone had done was to have actions that appeared intelligent to
people who did not know what the code was doing.  A good example would be
the twitching camera motion that David M. Wilson mentioned in his email.

William Crolley

----- Original Message -----
From: "David P. Anderson" <dpa at io.isem.smu.edu>
To: <dprglist at dprg.org>
Sent: Monday, March 05, 2007 7:13 PM
Subject: [DPRG] Re: Allen Robots (Will's emotions)

> Howdy
>
> "Systemic Temporal Representation?"   STRs?.
>
> I think "temporal" in there is a step in the right direction.  But I'm
not
> sure about "representation," which seems like a slippery slope.
>
> Maybe the query is premature.  Anybody else on this list writing
behavior-based
> or subsumption code at all?  That's a good starting place.
>
> I tried a similar thread over on comp.robotics.misc but to no avail.
Those folks
> are all convinced that subsumption can't solve the problems they are
interested
> in solving, although they become quite defensive when asked, "Like what?"
Mostly
> just theorists, I was unable to find anyone with actual experience
writing
> subsumption code.
>
> Is that the case here as well?    Perhaps before we venture into
discussion of a
> hypothetical code layer above the subsumption layers, that would be
useful to
> know.  Any behavior-based robot coders among us?  And if not, for those
who are
> robot builders, how do you organize your robot code?
>
> best,
> dpa
>
> _______________________________________________
> DPRGlist mailing list
> DPRGlist at dprg.org
> http://list.dprg.org/mailman/listinfo/dprglist