As a physiologist and control engineer, I was amused by the ongoing debate
about "who controls whom in the body." The question itself misses the whole
point of what a control system is about. A control engineer never needs to
ask the question of "who" is in control. Rather, there are more
down-to-earth questions of "what", "how", and "why" that every controller
(or so-called) must address. To make a long story short, the key points of a
control system are:
1) First and foremost, WHAT is the control objective? The brain has a clear
set of objectives such as maintaining homeostasis, making goal-directed
movements and long-term decisions etc., whereas the visceral and
skeletomuscular systems merely provide the means for realizing these
objectives.
2) WHY the objective? According to thermodynamics law, nature tends to
become disorderly and self-destructive. A control objective is usually to
bring about order and stability, and to maintain long-range viability.
3) HOW to achieve the objective? More importantly, how to OPTIMIZE the
objective? An intelligent controller knows how to do its job and do it
well. Optimal control is the ultimate goal of every control system there is,
including neural control. For reference, see:
-- Harris, On the optimal control of behaviour: a stochastic
perspective. Neurosci Methods 1998
Aug 31;83(1):73-88
-- Poon, Self-tuning optimal regulation of respiratory motor output by
Hebbian covariance learning.
Neural Networks 9: (8) 1367-1383 NOV 1996
4) WHAT is the control law (algorithm) and HOW would it accomplish the
objective? Peripheral systems do not have any control law; they simply obey
(in a passive fashion) the central command and underlying physical laws. The
brain has a host of intelligent control laws for differing control systems,
many of which defy engineering explanation. For example, homeostatic control
is not as simple as "feedback" or "feedforward" control as taught in most
neuroscience or medical texts. The main outstanding question is how to
optimally control a system in the face of high degree of uncertainty and
variability, subject to the physical and environmental constraints. Learning
and adaptation are increasingly recognized as the cornerstones of such
intelligent control.
5) Now, returning to the WHO controls WHOM question: Indeed, there are
complex control systems in which there could be multiple controllers
interacting with one another. Each controller must address a similar set of
control questions (1-4) and try to "outplay" one another or else will be
slave to (or controlled by) other controllers. This is a problem in
multi-agent games theory. For example, is NATO controlling Milosevic
through its air raids, or is Milosevic controlling them by sitting it out?
Each controller has its own (selfish) objective and each will apply the best
possible control strategy to accomplish its objective and maximize the gain
or minimize the risk, sometimes to the destruction of its opponents.
Using the same example of the TV case, the TV viewer selects the
channel by pushing the remote controller at will and is clearly in control
of his/her entertainment needs (control objective of viewer). On the other
hand, the TV stations develop and transmit the various programs in the hope
of reaching and impacting their viewers, thereby attracting advertisements
(control objective of TV networks). In this sense, each party is trying to
contol the other party and profit from it. The TV set and the cathode ray
tube (as well as the eyes and the fingers) serve merely as the unwitting
mediator (communicator) of this transaction between two selfish yet
intelligent players.
Chi-Sang Poon
*****************************************************************
Chi-Sang Poon, Ph.D.
Harvard-MIT Division of Health Sciences & Technology
Bldg. E25-501
Massachusetts Institute of Technology
Cambridge, MA 02139
Ph: +1 617-258-5405
Fax: +1 617-258-7906
email: cpoon@mit.edu
*****************************************************************