home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!europa.asd.contel.com!darwin.sura.net!paladin.american.edu!auvm!MCIMAIL.COM!0005096370
- Message-ID: <02920826064420/0005096370ND2EM@mcimail.com>
- Newsgroups: bit.listserv.csg-l
- Date: Wed, 26 Aug 1992 06:44:00 GMT
- Sender: "Control Systems Group Network (CSGnet)" <CSG-L@UIUCVMD.BITNET>
- From: Henry James Bicycles Inc <0005096370@MCIMAIL.COM>
- Subject: STEERING
- X-To: "EMS: INTERNET" <CSG-L@vmd.cso.uiuc.edu>
- Lines: 88
-
- [From Hank Folson (920826)]
-
- (Chris Malcolm 920825)
-
- >Clearly these apparent paradoxes arise because the "folk" theories of
- >how control is learnt and executed are wrong.
-
- Is there really a paradox here? As you described, the only way to get a
- motorcycle/bicycle to turn right is to get it leaning to the right before
- initiating the final steering actions. The only reason we make a big deal about
- counter-steering is because i
- t is not obvious that we must do it and do do it. IsnUt the use of the term
- "counter-steering" a stimulus-response sort of description?
-
- >What is required is a theory of control which can be applied to the >phenomena
- of counter-steering in which the paradoxes become predicted.
-
- How about this, Chris: One thing that I am coming to understand about control
- theory is that purpose is all that counts. Method is secondary, just a means
- to the intended end. The motorcycle/bicycle steering scenario is a good example
- of this, I think.
-
- The first time someone without a physics degree or an experienced tutor tries to
- ride a bike/motorcycle, they will steer right when they want to turn right. The
- desired perception is turning right, and turning the handlebars to the right
- seems like a good
- idea at the time. The result is a fall or a moment of panic and wobbling with
- no turn. The purpose/goal of turning right remains, however. Steering right
- didn't work, so I suspect that our control systems simply try another
- controlling action AT RANDOM.
-
- Handlebars only go right or left, so the next controlling action is to turn the
- bars to the left. NOTE THAT THE SYSTEM HAS NO INKLING OF WHY THIS SHOULD WORK
- ANY BETTER THAN TURNING THE BARS TO THE RIGHT JUST DID. The bike now falls to
- the right side in r
- esponse to God and Newton. Whatever level of our hierarchy controls for not
- falling quickly takes over. The most recent output was to turn the handlebars
- to the left. So now the system corrects, and turns the bars to the right. This
- is the only controllin
- g action option available, fortunately. (In other words there are not so many
- degrees of freedom that all options could not be tried before you fall to the
- ground.)
-
- Now the bike has velocity, and is leaning to the right, and the bars are turned
- to the right. With the centrifugal (or is it centripedal?) force, and the human
- control system hopefully in a stable range, there is no fall, just a
- beautifully controlled rig
- ht turn EVEN THOUGH THE CONTROL SYSTEM DOES NOT UNDERSTAND WHY.
-
- There is not much in the way of planned output in my perception of how a control
- system functions. The initial controlling action chosen may be selected for
- some reason (e.g. If I want to turn right, it makes sense to turn the bars to
- the right, based on
- what is stored in the mind from past experiences, such as steering a tricycle.),
- but if that first choice does not produce the desired input as the process goes
- around the control loop, or if there is no reason to choose a particular
- action, the control s
- ystem will try one at random. And if that doesn't work, it will continue to
- control by trying another action. One thing I like about this suggested
- interpretation is that a living control system can survive a great many complex
- and/or new situations, with
- out requiring a lot of computing power, and can react very fast. Perhaps those
- lucky enough to randomly pick the best controlling actions first survive better
- than others!
-
- The bicycle steering example is, to me, a good example how a control system can,
- by random actions, handle something that its sensors and internal logic would
- never figure out. The way E. coli tumbles and takes off in a random direction
- when controlling f
- or a new food source is a lower level example.
-
- In the going right by steering left scenario, I did not mention another
- controlling action the system has available: putting a foot down. This is
- perfectly valid, and many beginners probably do this. This takes care of the
- error signal of falling down, bu
- t doesn't do anything about turning right. Those who put their foot down will
- probably take longer to learn how to ride, because their control systems will
- not get the opportunity to complete the steering left to go right strategy.
-
- Another thought to ponder: The system has just learned by trial and error how to
- turn right. How will the system respond to the next new goal which would be to
- turn left? Will the rider try steering left to go left, or will he immediately
- steer right to
- go left? Is this a prediction of your paradox, Chris?
-
- Hank Folson, Henry James Bicycles, Inc.
- 704 Elvira Avenue, Redondo Beach, CA 90277
- 310-540-1552 (Day & Evening) MCI MAIL: 509-6370 Internet: 5096370@MCIMAIL.COM
-