[From Rick Marken (2004.04.06.0930)]
Bill Powers (2004.04.05.1508 MST)--
I sometimes wonder whether the story of economics, once we get all the
smokescreens out of the way, is not going to turn out to be extremely simple.
I agree. I think it's similar to what happens when I do my taxes at the end
of the year. All my detailed financial activities seem so complex while I'm
doing them. But the summary of these activities in the tax form makes
everything very simple: X money came in, Y money went out. If X is greater
than Y then all's well. If Y is greater than X I have to work on keeping
Linda out of CostCo. Simple.
On another topic: Did you or Bruce Abbott ever write a baseball catching
program that worked by controlling vertical acceleration? I have written an
acceleration control program and have been studying it more carefully. It
works but it is not very robust. The velocity control model is _much_ better
in the sense that it works over a much wider range of starting field
positions for the fielder.
I want to write a paper comparing the three existing models of fly ball
catching that have been proposed: optical acceleration cancellation (OAC),
linear optical trajectory (LOT) and optical velocity control (OVC). This
should be simple but I'm finding that as I try to figure out what's going on
with these models I'm having my modeling skills tested severely (which is OK
with me, ego-wise; I think I'm a better researcher than a modeler), so any
help from the more skilled modelers would be greatly appreciated.
This seems like an appropriate topic for the start of baseball season, no?
Best
Rick
···
--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400
Hi, Rick --
On another topic: Did you or Bruce Abbott ever write a baseball catching
program that worked by controlling vertical acceleration?
I haven't, but it could obviously be done. However, if you set a reference
level for acceleration, you don't specify the initial velocity, so the
velocity could be anything. Suppose you have the outfielder start tracking
some short random time after the ball is hit. I'll bet the velocity
variations would prevent catching the ball, while the velocity control
system would work as well as ever.
I have written an
acceleration control program and have been studying it more carefully. It
works but it is not very robust. The velocity control model is _much_ better
in the sense that it works over a much wider range of starting field
positions for the fielder.
I want to write a paper comparing the three existing models of fly ball
catching that have been proposed: optical acceleration cancellation (OAC),
linear optical trajectory (LOT) and optical velocity control (OVC). This
should be simple but I'm finding that as I try to figure out what's going on
with these models I'm having my modeling skills tested severely (which is OK
with me, ego-wise; I think I'm a better researcher than a modeler), so any
help from the more skilled modelers would be greatly appreciated.
I think your best bet would be to make friends with a local physicist or
engineer you can sit down with whenever you need to. Either that or take a
cut at the paper and send it to me for comments.
Bill
[From Rick Marken (2004.04.07.0950)]
Me:
On another topic: Did you or Bruce Abbott ever write a baseball catching
program that worked by controlling vertical acceleration?
Bill Powers
I haven't, but it could obviously be done. However, if you set a reference
level for acceleration, you don't specify the initial velocity, so the
velocity could be anything. Suppose you have the outfielder start tracking
some short random time after the ball is hit. I'll bet the velocity
variations would prevent catching the ball, while the velocity control
system would work as well as ever.
Something like this is, indeed, going on. The initial optical acceleration
when the ball comes off the bat is, of course, positive but it immediately
goes negative as the ball continues through it's trajectory. I've found that
the performance of the acceleration control model is improved significantly
if the outfielder waits through at least 1 model cycle (dt period) before
stating to control acceleration. After one cycle the acceleration (if the
fielder doesn't move) is increasingly negative. Of course, it's still true
that the optical velocity of the ball can be almost anything when
acceleration is kept at zero. This (as I mentioned in my 2001 "Controlled
Variables" paper -- reprinted in _More Mind Readings_) does seem to be why
the optical acceleration cancellation (OAC) model does worse than the
optical velocity control (OVC) model.
I think your best bet would be to make friends with a local physicist or
engineer you can sit down with whenever you need to. Either that or take a
cut at the paper and send it to me for comments.
I think my difficulties are more with the control modeling itself, rather
than with the physics (oddly enough), though, of course, they are related.
I'm kind of busy with other stuff at the moment but I hope to have a nice
spreadsheet version of the models working by the end of the month. I'll post
the programs then and see if I can get your suggestions.
Best
Rick
···
--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400
[From Bill Powers (2004.04.08.1415 MST)]
Rick Marken (2004.04.07.0950) --
Something like this is, indeed, going on. The initial optical acceleration
when the ball comes off the bat is, of course, positive but it immediately
goes negative as the ball continues through it's trajectory. I've found that
the performance of the acceleration control model is improved significantly
if the outfielder waits through at least 1 model cycle (dt period) before
starting to control acceleration.
If the outfielder manages to keep vertical visual angular acceleration at a
small negative value, the visual velocity will slowly be decreasing from
whatever its initial value was. Chapman's criterion was that the tangent of
the angle should _increase_ at a small constant rate, which probably means
that the apparent vertical velocity should be kept constant, and that the
acceleration should be zero. I think there are probably quite a few
controlled variables of this sort that would get the outfielder within
range to catch the ball -- keeping the velocities small, whether they are
upward or downward, requires keeping the accelerations close to zero, and
over some range these should all have about the same final effect. If you
can distinguish among them by comparing the predicted paths of running,
maybe you can offer a basis for preferring one assumption over another..
If that's the case, then I suggest that your best strategem would be to put
out a call for joining forces with all these writers, because they are all
on the side of perceptual control theory and opposed to those who say that
outfielders compute where the ball is going to come down and run there to
catch it, or that they oscillate their way chaotically to the right spot,
or that they are stimulated to run by the sight of the ball and reinforced
for running the right way and at the right speed by catching the ball. Give
these guys a name for what they're proposing: PCT. Then the debate simply
becomes an attempt to determine from the data what perception the
outfielders are controlling.
It's perfectly possible that one outfielder controls for a slowly rising
ball while another controls for a zero-acceleration ball. This is, after
all, not a genetically-determined behavior, but a learned one. Maybe the
real question is, what method will lead to catching balls the most
reliably? If your baseball-analyst coworkers could determine that, they
might have something to contribute to MLB.
Best,
Bill P.
[From Rick Marken (2004.04.09.1100)]
Bill Powers (2004.04.08.1415 MST)--
I think there are probably quite a few
controlled variables of this sort that would get the outfielder within
range to catch the ball
I agree.
If that's the case, then I suggest that your best strategem would be to put
out a call for joining forces with all these writers
At least one of the baseball researchers is going to work with me on this. I
hope they will all join in eventually.
because they are all on the side of perceptual control theory
Yes, they all certainly understand that ball-catching is a control task.
What they don't understand (I think) is what a controlled variable _is_ or
how to determine what variable is under control in particular behavioral
situations. I am trying to help them towards this understanding, as gently
as possible.
Then the debate simply
becomes an attempt to determine from the data what perception the
outfielders are controlling.
Exactly. That's what I think the debate really is about. I'm just trying to
show how to do this, using some version of The Test. Because these
researchers don't understand the concept of controlled variable, they have
been trying to determine the "informational basis of catching" by simply
looking at the shapes of various optical trajectories from the fielder's
perspective. I'm going to try to explain that this is like trying to
determine which of the three squares in my "Mind reading" demo is being
controlled by looking only at the paths of the three squares that are being
moved (http://www.mindreadings.com/ControlDemo/ThreeTrack.html).
It's good that these researchers are looking at potential controlled
variables; they just haven't got the other aspects of The Test down yet. But
the fellow I'm working with has data that may be an excellent basis for
doing The Test. He's the one who looked at what dogs are seeing when they
catch Frisbees. The irregular trajectory of a Frisbees may provide the kind
of disturbance that can let us test at least one hypothesis about the
optical variable the dog is controlling when it catches a Frisbee.
It's perfectly possible that one outfielder controls for a slowly rising
ball while another controls for a zero-acceleration ball.
This is a great point. Thanks. I'll put it in my write-up.
Maybe the
real question is, what method will lead to catching balls the most
reliably? If your baseball-analyst coworkers could determine that, they
might have something to contribute to MLB.
I agree, I think. What's MLB?
Best
Rick
···
--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400
[From Bill Powers (2004.04.09.1236 MST)]
Rick Marken (2004.04.09.1100)--
I agree, I think. What's MLB?
Major League Baseball.
Best,
Bill P.,