Smart Input Functions & Hierarchies

[Avery Andrews 950705.1231]
  (Bill Powers (950701.0815 MDT)

Continuing to think about visual flow and rpm w.r.t. velocity
perception, I certainly agree that there would be hierarchical
organization on the perceptual end, something like this:

                 ^

···

                 > Velocity
                 >
              --------
  ---------> | Smarts |
   other --------
   inputs ^ ^
             > >
       ------------- -------
      > Visual Flow | | RPM |
----- | Sensor |-|Sensor|-------
       ------------- ------

But its not so clear to me that there need to be control systems
associated with either visual flow or RPMs. w.r.t. RPMs, I conjecture
that (a) over a wide range, the actual RPMs don't matter (b)
someone who designs a ship's propulsion system ought to know what an
acceptable range of RPMs for given fuel inputs ought to be; if the
actual RPMs fall outside the values expected by the designer, it's
probably time for higher-level intervention, rather than attempted
stabilization by a low level control system (like, something is tangled
up in the propellor, or it's out of the water, etc.) As for visual flow,
I can't see any point in controlling that at all: it's only
interesting when it's indicating velocity. So my suggested system
would just compare actual with reference velocity, and use the
difference to determine fuel supply to the engine.

Or, if you wanted to have an RPM control system after all, the velocity
error signal would determne the RPM reference signal. At the moment
I can't figure out any use at all for a visual flow control system.

So my story is that you can have hierarchical perceptions with or
without hierarchical control systems, and I don't see any challenge to
my original claim that smarting up the perceptual functions is in
general an alternative to smarting up the output functions; the issues
that I think are interesting are (a) are smart input functions different
in their potential effects from smart output functions (b) if so, which
works best under what kinds of circumstances (c) which is actually
employed when by living systems.

Perhaps a misleading feature of my original posting was presenting
visual flow as the primary determinant of velocity and rpm as a
secondary one, an 'ersatz' for visual flow, rather than making them
look like symmetric contributors in a hierarchy; what I think is that
both the hierarchy and the asymmetry exist. The ecologically
significant variable is actual velocity, which is in general going
to be more reliable indicated by visual flow than by propellor rpms,
but the former might not always be available, so then the system does
its best with the latter.

Avery.Andrews@anu.edu.au