Her answer isn't enough for Rajat Mittal. The professor of
mechanical and aerospace engineering at George Washington University
has developed a deep interest in Ms. Coughlin's dolphin kick. A
superstar's winning move, he thinks, deserves to be pored over by a
supercomputer -- and, in Washington, he has one.
For the past year, Prof. Mittal, 37, has been working to load it
with a three-dimensional incarnation of Ms. Coughlin undulating in a
virtual pool. Even at 100 billion calculations a second, the task is
huge -- and a measure of how far the technological backfield of today's
Olympics will go to win a few more medals. Prof. Mittal's goal, simply
put, is to take the guesswork out of perfection.
"Up till now, what is good or bad in all human performance is based
on intuition," he says. "Once science comes into it, some of this
fuzziness about what's best and what's not will be gone."
On Prof. Mittal's office shelf, a bluegill sunfish is pickling in a
jar. He has a grant to help the U.S. Navy study how the fish swims with
its little front fins. The Navy wants someday to swap a machine for the
moody dolphins it sends out on search missions, and that will require
new levels of computer simulation, not for objects like submarines, but
for things that flap and squiggle under water.
With the Navy paying the bill to drop a fish into an $800,000
supercomputer, Prof. Mittal figured he may as well drop a fishlike
human in, too. He hooked up with Russell Mark, a former college swimmer
who, at 24, is "biomechanics coordinator" for USA Swimming, the sport's
ruling body. His job is to explore how flesh and fluid might cooperate
to make swimmers move faster.
"Look across a pool," says Mr. Mark. "Every swimmer will be using
different techniques because every swimmer is taught different
techniques." It has been so at least since Duke Kahanamoku came up with
a new kick and broke the Olympic freestyle record at Stockholm in 1912.
But there are no set answers. Even James "Doc" Counsilman, a pioneer of
scientific coaching who died this year, had to amend his theory that a
hand slicing across a swimmer's body underwater pulls the swimmer the
way propellers pull airplanes; the hand, it now seems, is still mostly
a paddle.
Before computers, coaches shaped strokes the way engineers sculpt
cars -- by watching swimmers in a flume, the watery equivalent of a
wind tunnel. Coaches today watch laser-scanned swimmers in digital
flumes. That's how Speedo tested its new full-body swimsuit. Mr. Mark
almost did as much. He had two of his champs, Gabrielle Rose and Lenny
Krayzelburg, laser-scanned in Hollywood. But then he got a call from
Prof. Mittal, and jumped into the next dimension.
Laser-scans have a failing: They don't move. Scanned dry-land
athletes can be wired up, animated and pasted into videogames. But they
aren't fighting a fluid. Water's disorderly effect on motion makes
picking apart a swimmer's progress hugely more complex.
"Swimmers push the water, and the water pushes back," says Prof.
Mittal. "It all comes down to turbulence. If we could compute every
instant in a stroke, we could understand it."
That's why he needs a supercomputer. Mr. Mark donated the Rose and
Krayzelburg scans, and a set of videos from USA Swimming's flume in
Colorado Springs. One showed Ms. Coughlin dolphin-kicking. When he saw
it, Prof. Mittal knew she was the swimmer he had to use.
"She swam straight, maintaining an even depth," he says. "All fish
do this, passing a wave through their bodies from head to tail. This
was it -- the natural-selection stroke, the best way to swim."
Lacking a scan of Ms. Coughlin, Prof. Mittal assigned a student to
superimpose her videoed body, frame by frame, onto the scan of Ms.
Rose. He then asked James Hahn, director of GWU's Institute for
Computer Graphics, to essentially insert a skeleton, enabling the scan
to move. The output is a goggled, silver phantom, dolphining across a
black screen, trailing a thin red line undulating across a graph --
sort of like the markings on an electrocardiogram.
Three-dimensional, observable from all angles, this creature is
Prof. Mittal's raw material. All he has to add next is water. Pushing
the limits of his field-computational fluid dynamics, he plans to
factor in every swirl and counterswirl produced by an ever-changing
sequence of motions known as a single stroke. To account for every eddy
within every eddy, he will break each stroke into 20,000 units and
perform 200 million calculations on every one.
By reducing Ms. Coughlin to her elements, Prof. Mittal aims to
attain an absolute awareness of what makes her so fast. "Does her body
size naturally put her into the right range of amplitude?" he asks.
"Should small swimmers kick at higher frequencies than big swimmers? If
so, how much higher? That's what we want to know."
Don't expect an answer in Athens. The supercomputer will be mincing
the dolphin kick for three more years, at least. And when it eventually
applies its software to strokes that pierce the water's surface, the
variables will multiply. Yet Prof. Mittal and his partners see a day
coming when swimmers will have their Olympic body mechanics customized
without ever getting their feet wet.
Some variables will never be downloaded: willpower, for one. And
coaches who believe that natural strokes are best left natural may not
want their swimmers diving into virtual pools.
"Some of our people don't care about technique at all," says Mr.
Mark. But Natalie Coughlin, whose stroke is as natural as they come,
isn't one of them. "You can't change physics," Ms. Coughlin says. "You
might as well figure out how it works." She thinks the water in Prof.
Mittal's supercomputer is just fine.
