You Can't Build a Better Tuna

RoboTuna

Here's the first short essay I wrote for the Science Writing program at MIT. As a warm-up exercise, we were tasked with observing an object at a local museum and writing 500-800 words on it.

Though over twenty years old, RoboTuna looks like a freshly caught fish partway through filleting.

Its fabric skin is pulled back to reveal slender plastic ribs and motor-guts near its head, but the layer of synthetic scales between ribs and skin still clings to its back half. A wire threads through each belt of long, thin scales, creating a series of overlapping scale-belts down its body. Six joints down its spine give it the potential to ripple like a tuna hurtling after prey. RoboTuna is ready to swim as smoothly as a real fish.

But the metal pole attached to its back breaks the illusion. The pole connects the robot to a large support structure for its electronics.

Yet focusing on the fish part again, I begin to picture a whole school of RoboTunas, self-directed, seeking out shipwrecks and ocean trenches.

Created by ocean engineers at Massachusetts Institute of Technology in the 1990s, RoboTuna now lives in a glass case in the M.I.T. Museum’s robot collection. Each robot is a milestone in the history of getting machines to behave more like living things.

Researchers designed fish-mimicking robots to figure out how sea creatures outperformed submarines in efficient movement, hoping to build better robotic vehicles to survey the ocean floor.

Zoologist James Gray found in 1936 that fish do not seem to have the muscle power required to reach the speeds scientists clocked them at in the wild, according to the museum exhibit.

Michael Triantafyllou, a professor of mechanical and ocean engineering at M.I.T., showed—with help from his RoboTuna—how fish use fluid physics to their advantage to reach high speeds with little power.

Anything moving through a fluid or gas creates turbulence behind itself. Think of the vortex of air behind a speeding truck. The drag created by this vortex can slow the object down when the vortex disrupts flow over the object.

But by timing their tail fin’s movements just right, fish can benefit from the vortices and regain some of the energy spent in creating them, according to a 1995 Scientific American article written by Triantafyllou and his brother George, a professor of hydrodynamics at the National University of Athens. A submarine with a rigid stern cannot benefit from this effect, because the vortex created by its fixed propeller spins sideways to its motion.

In the museum, a photo of RoboTuna swimming in a tank illustrates how real fish do it. Blood-red dye outlines the vortices spinning off to either side of the robot’s tail. The vortices are spaced so that the tail catches a boost from them right when they spin in the direction of the robot’s motion.

Not only did Triantafyllou’s team mimic the form and behavior of the tuna, they tried to imitate the process which created it: natural selection.

There are too many variables for RoboTuna’s motion for a simple formula to predict the best way to swim, according to a 1996 paper by the group. By creating many movement programs with different settings for each variable, the researchers simulated a population of tuna with different genes for movement and tested them all. They rewarded the most efficient programs by letting them produce “offspring”—more programs with the same movement genes. A feature that made random mutations in the genes imitated the way helpful adaptations arise in nature.

The best performing program, not surprisingly, had a motion much like a real yellowfin tuna, according to the group’s 1996 paper.

Yet the longer I look at RoboTuna, the less lifelike it seems. For one thing, it has no eyes. I circle the display to look at the creature’s rear. Its tail is frozen in an S, suggesting it might surge through the glass at any moment.

I think not of free-swimming tuna but the summer I handled a number of dead laboratory fish, their tails crooked and stiff with preservatives. I realize that the machine just made me want to see a living fish again.

The robot’s developers felt similarly.

“The more sophisticated our robotic-tuna designs become, the more admiration we have for its flesh-and-blood model,” Michael and George Triantafyllou wrote in the 1995 Scientific American article.

It took millions of years for evolution to perfect the tuna design. Human researchers have only the few decades of their lives to try and reverse-engineer it.