Got it.
Erling
···
Richard S. Marken
"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
–Antoine de Saint-Exupery
Got it.
Leeanne
···
Richard S. Marken
"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
–Antoine de Saint-Exupery
[From Rick Marken (2003.09.02.2000)]
Just testing my new OSX Mac operating system and mail program. I don't have
anything to reply to so I'm sending this through the listserve interface.
Hope everyone had a nice summer.
Best
Rick
[From Rick Marken (2003.09.03.0905)]
Bill Powers (2003.09.03.0753 MDT)--
Still working on a new tracking-simulation program. Any luck finding the
article by McRuer, Graham, Krendel, and Riesener (1965)? It's reference 19
in the McRuer and Jex (1967) article, and is apparently where the
experimental conditions are discussed.
I don't have that reference. Please sent it to me and I'll order it.
John has said that the basic "crossover model" data have been replicated
"thousands of times," so it ought to be possible to come up with one or two
reports of actual experimental data analyzed by the method of Sheridan and
Ferrel (1974) as described on p. 160 of _Control theory for humans_. Now
that we know the plots in Ch. 14 of the latter were _not_ computed from
experimental data in that way, it's more important than ever to see real
data and computations.
Yes, indeed.
I keep coming back to the fact that control _without_ adaptation can be
achieved over a wide range of plant parameters by a two-level model when
the plant is a simulated damped mass on a spring. Apparently, this is not
possible when the plant ranges from a pure zeroth order system to a pure
second-order system. The model can work without adaptation over a span of
two orders, but not three. Or I haven't found a way so far.
Maybe people can't work without adaptation over a range of three orders.
The simulation I'm working on will wrap up most of what we've been talking
about, with the hand-stick dynamics represented, along with plants
consisting of (1) a pure order 0, 1, or 2 function, or (2) a mass on a
spring with adjustable mass, spring constant, and damping. Disturbances can
be applied to the target, and to the input and output of the plant, in all
combinations. The behavior of the simulation can be shown in the time
domain, or in the form of Bode plots, the frequency domain. I'm thinking of
adding the ability to take experimental data for comparison with the
simulation performance, although with as many parameters as we have in this
simulation I'm not quite sure how to automate the adjustment of parameters
for best fit to the real data. I may have to fall back on good old Vensim.
Or we might have to match the hand-stick dynamics in one experiment, then
use the result for analyzing the overall performance.
I think you should definitely take experimental data. People should be able to see
whether their data at least resembles the model data, which should be quite
different for the different orders of control.
Best regards
Rick
···
--
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org
[From Bill Powers (2003.09.03.0753 MDT)]
Still working on a new tracking-simulation program. Any luck finding the
article by McRuer, Graham, Krendel, and Riesener (1965)? It's reference 19
in the McRuer and Jex (1967) article, and is apparently where the
experimental conditions are discussed.
John has said that the basic "crossover model" data have been replicated
"thousands of times," so it ought to be possible to come up with one or two
reports of actual experimental data analyzed by the method of Sheridan and
Ferrel (1974) as described on p. 160 of _Control theory for humans_. Now
that we know the plots in Ch. 14 of the latter were _not_ computed from
experimental data in that way, it's more important than ever to see real
data and computations.
Also, both I and my simulations are having a huge problem with controlling
a pure double-integral simulated plant at the frequencies implied by the
McRuer and Jex Fig. 2. McRuer and Jex said that an example of a pure double
integral plant is the response of an aircraft's heading to aileron control,
and I know I can do that at least at low frequencies (it would be hard to
vary any aircraft's heading appreciably at 10 radians per second without
ripping the wings off -- that would be equivalent to 1.6 360-degree turns
per second!). However, there is feedback available for both angular
velocity and acceleration in an aircraft, while in a computer simulation
only visual velocity information can be had.
I keep coming back to the fact that control _without_ adaptation can be
achieved over a wide range of plant parameters by a two-level model when
the plant is a simulated damped mass on a spring. Apparently, this is not
possible when the plant ranges from a pure zeroth order system to a pure
second-order system. The model can work without adaptation over a span of
two orders, but not three. Or I haven't found a way so far.
The simulation I'm working on will wrap up most of what we've been talking
about, with the hand-stick dynamics represented, along with plants
consisting of (1) a pure order 0, 1, or 2 function, or (2) a mass on a
spring with adjustable mass, spring constant, and damping. Disturbances can
be applied to the target, and to the input and output of the plant, in all
combinations. The behavior of the simulation can be shown in the time
domain, or in the form of Bode plots, the frequency domain. I'm thinking of
adding the ability to take experimental data for comparison with the
simulation performance, although with as many parameters as we have in this
simulation I'm not quite sure how to automate the adjustment of parameters
for best fit to the real data. I may have to fall back on good old Vensim.
Or we might have to match the hand-stick dynamics in one experiment, then
use the result for analyzing the overall performance.
I have a force-feedback joystick, but the feedback is pretty sluggish and
anyway the only way I can program the forces is in Dark Basic -- don't know
how to do it with Delphi. Bruce Nevin's idea of the plank with rubber bands
stretched between nails looks better and better, though hard to adjust.
When I get sick of working on this program, I play with others. The demo I
showed at the meeting (where 150 control systems control 150
randomly-defined functions of 150 environmental variables, only it's now
500) is now in a form where reorganization of the input function matrix is
working. It brings all the errors very close to zero. However, when all the
reference signals are changed arbitrarily, it seems to have to reorganize
again instead of having learned how to control better in general. As
always, more work is needed.
Best,
Bill P.
[From Rick Marken (2003.09.03.1200)]
Bill Powers (2003.09.03.1218 MDT)--
Rick Marken (2003.09.03.0905)]
>I don't have that reference. Please sent it to me and I'll order it.
D. McRuer, D. Graham, E. Krendel, W. Reisener, Jr.. "Human pilot dynamics
in compensatory systems -- theory, models, and experiments with controlled
element and forcing function variations." AFFDL-TR-55-15, January 1965.
Condensed version in J. Franklin Inst. _238_ January-February 1967
Thanks. It's ordered.
Best
Rick
···
--
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org
[From Bill Powers (2003.09.03.1218 MDT)]
Rick Marken (2003.09.03.0905)]
> Bill Powers (2003.09.03.0753 MDT)--
>
> Still working on a new tracking-simulation program. Any luck finding the
> article by McRuer, Graham, Krendel, and Riesener (1965)? It's reference 19
> in the McRuer and Jex (1967) article, and is apparently where the
> experimental conditions are discussed.I don't have that reference. Please sent it to me and I'll order it.
D. McRuer, D. Graham, E. Krendel, W. Reisener, Jr.. "Human pilot dynamics
in compensatory systems -- theory, models, and experiments with controlled
element and forcing function variations." AFFDL-TR-55-15, January 1965.
Condensed version in J. Franklin Inst. _238_ January-February 1967.
best,
Bill P.
[From Rick Marken (921002.1600)]
The following is a test of the emergency PCT understanding system. This
is only a test:
"PCT makes all current textbooks on psychological methods obsolete."
True or false.
Have a nice weekend.
Rick
···
**************************************************************
Richard S. Marken USMail: 10459 Holman Ave
The Aerospace Corporation Los Angeles, CA 90024
E-mail: marken@aero.org
(310) 336-6214 (day)
(310) 474-0313 (evening)
[From Rick Marken (2004.11.15.1510)]
I'm just testing my e-mail system. Pay no attention to this post (unless
you've been doing those brain wave exercises and can't help it;-))
Best
Rick
···
--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400
--------------------
This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.