DPRG List  

[DPRG] Is subsumption old hat?

Subject: [DPRG] Is subsumption old hat?
From: William Crolley w.crolley at mchsi.com
Date: Sat Sep 8 11:16:31 CDT 2007

The new sensor architecture that is competing with subsumption would be the
various types of sensor fusion.  The advantages of sensor fusion is that
different types of sensors can be combined to avoid the weaknesses of a
single sensor.  This is a major problem for subsumption which works best
with binary info, similar to relay logic.

Currently the most popular sensor fusion method is the extended Kalman
Filter.  This assumes that the sensor data  has Gaussian noise in it.  In
practice it works fairly well even if the noise is not Gaussian.

An example of combining sensors would be to use a compass, a gyroscope, and
odometry to calculate the heading.

William Crolley

> >
> > But here's the point, Dave. Can you name the next architecture
> > that replaced subsumption? Or can anyone primarily in the
> > hobbiest side of interest name the next one? I doubt it.
> Can you name it Randy?  Or are you as clueless as the rest of us?
> > So if robotics reseach has moved on, back in 1992, even if
> > subsumption is useful and robost, which I admit it is, does that
> > mean that an additional 15 years of research not provided
> > something even more useful and robust? Or are we as hobbiest
> > just that far behind the times that we don't even discuss the
> > new advances.
> Why don't you finish reading the book, and then enlighten us?
> > New advances, there? Or not there? If they're there, then why
> > don't we even know their name(s)?
> Because you haven't told us yet!  If we are unaware of the new paradigm,
> how can we discuss whether it is better than the one we currently use?  I
> am sure there are many hobbyist robot builders who haven't even worked
> their way up to subsumption yet, after all for most of us it is just a
> hobby.
> Ed Okerson

More information about the DPRG mailing list