home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!EID.ANL.GOV!GABRIEL
- Message-ID: <9212201635.AA17233@athens.eid.anl.gov>
- Newsgroups: bit.listserv.csg-l
- Date: Sun, 20 Dec 1992 10:35:51 CST
- Sender: "Control Systems Group Network (CSGnet)" <CSG-L@UIUCVMD.BITNET>
- From: John Gabriel <gabriel@EID.ANL.GOV>
- Subject: A Useful Insight at last
- Lines: 85
-
- [Gabriel to NET 921220 10:01 CST]
-
- Bill P and I have had an offline discussion that generated lots of
- light for me. So I want to share. We have two rather different views
- of Control Theory and hence of BCP. I think this serves to unify
- the two positions in exactly the technical sense of being the least
- general theory that includes both.
-
- I'm going to avoid the word "Control" because it has so many different
- meanings. This does not signify any quarrel with BCP - just that there's
- a mathematical theory including BCP as a special case that I'm comfortable
- with.
-
- The term "Reference Signal" is replaced by "Desired State (of the observed
- environment)" for similar reasons. The term "Reference" has Rock of
- Gibraltar like connotations, but I think we can all agree it's possible
- to change our desires from time to time, just as the BCP signal topology
- allows and encourages - i.e. the "desire" comes from within, not from
- without.
-
- Thus we begin with two states of the external environment, Desired (D),
- and Perceived (P). For the unified theory it is sufficient for D and P
- to have "representations" in the mathematical sense, as points in a
- "metric space". This is a space where "distance" has a meaning, of
- one or more dimensions, each of which may be discrete or continuous.
- This allows us to talk about the distance between D and P, and
- because the space is a metric space, we are guaranteed not to reach
- any absurd conclusions.
-
- Arm and Man live in 6N dimensional metric spaces - 3 spatial coordinates
- for each degree of freedom, and 3 momenta for each degree of freedom.
- If you are doing rigid body kinematics, the 6 becomes a 12, but no matter.
-
- Most of my concepts live in many dimensional spaces where the component
- of state in one of the dimensions has two or three possible values -
- TRUE/FALSE or TRUE/UNKNOWN/FALSE. If we increase the number of possible
- state values in a single dimension to say 256 or 1024 or 8192 ....
- we can approach a continuous state space as closely as we wish, we
- may exactly represent the digital computational models of Man and Arm,
- and in fact we may represent any physical system where there is noise,
- either thermal or quantum, within the precision of experimental observation.
-
- A metric space has one other imporatnt property. Given two points say
- D and P we have a well defined distance between them. Distances MUST
- obey the triangular inequality, if A B and C are in the space, then
-
- AB + BC >= AC
-
- Now the central thesis of BCP can be stated.
-
- People behave to move P closer to D, where this is possible
- at an acceptable cost.
-
- Notice this is a constrained minimisation process. Also unlike the
- error P-R of BCP, DP is always positive, and may not always attain
- the desired minimum of zero. But the version of control theory used
- in BCP does in fact minimise DP for the cases considered in pratcice.
-
- Notice also that D and P also have representations by data streams,
- i.e. by samples in the metric space at time intervals appropriate to
- the Nyquist criterion for lack of aliasing. P type data comes from
- sensors. D type data comes from desires.
-
- In this case Shannon's Theory allows a metric space to be constructed
- (with a weighted Hamming metric) pretty much regardless of the details
- of the data streams, provided they are sequences of arrivals of symbols
- (including pulses of depolarisation across nerve membranes).
-
- I think for the moment I need say little more. In both the continuous
- and discrete cases orthogonal function expansions are possible
- with arbitrary positive definite metric for the scalar products.
- These metrics have to do with power limits, or value systems.
-
- This is why the detailed machinery of Control Theory does the job.
- Essentially the metric space allows Fourier Analysis and Laplace
- Transforms, which lead to the control formalism of BCP. But
- in some cases minimisation algorithms, taking account of the
- possible may be closer to the real world. All this also includes
- a lot of other things, like population genetics if you replace
- the machinery moving D around with Darwinian Selection. The story
- gets a little different but not very much.
-
- Over and out. I'm exhausted.
-
- John
-