Bio & Life Sciences

Bio-Robots Swim, Swarm, Change, and Shed Light on Evolution

A TadRos predator hits prey

Talk of robots and robotics research tends to conjure worries about manufacturing and futuristic fantasies about the “singularity.” But new bio-robots are designed instead to help us understand our evolutionary origins, and are providing insights into biology.

In a lab on the campus of Vassar College in upstate New York, biology professor John Long and his team are studying robots in a water tank as they fight for evolutionary supremacy. The researchers are using biomimetic autonomous robots to understand how fish-like vertebrates that lived 500 million years ago evolved into the fish of today.

These bio-robots are built to mimic biological systems to test biological hypotheses. “Most people are familiar with building models on a computer, but our robots are models that are physically embodied,” Long says. “And in my lab, they’re also physically autonomous, meaning that they are not remotely controlled by us.”

The University of Edinburgh’s Barbara Webb first conceived of building robotic organisms for study in the late 1990s. She built models of female crickets in order to understand the mechanisms behind their mate-finding behavior.

Vassar’s Long, who describes his work in Darwin’s Devices: What Evolving Robots Can Teach Us About the History of Life and the Future of Technology, uses bio-robots to try to understand how fish developed specific biological characteristics like vertebrae, caudal fins, and a sensory system called the lateral line that detects movement or vibration in water. Each of his Tadpole Robots, or TadRos, has decision-making machinery on an onboard computer chip. They decide how to move based on their goals, and how the movement of other robots in their tank could affect those goals. TadRos’ decisions inform researchers’ ideas about which selection pressures were most important 500 million years ago.

John Long teaches Introduction to Cognitive Science at Vassar College

The team started by programming the TadRos’ brains to approximate fish brains, using what’s already known about fish neuroscience. “We know that a fish can be cruising along, but if it detects a predator, its brain says, ‘Hey, whatever else you’re doing, no matter how much fun you’re having, get the hell out of there,'” Long says. “That’s an escape response, and we can see the neural signature of that in the nervous system of fishes.” Long and his team have built that same priority system into their robots: they swim along looking for food, but override that if they detect a predator.

Modeling the robots’ bodies isn’t easy. Long says that modeling extinct species is difficult because the fossil record is “a beautiful bunch of patterns and pieces of anatomy” that reveals nothing about physiology or behavior. “Dead fossils tell no tales,” he adds. So he and his colleagues fill in the holes in the fossil record with what they believe to be the right traits or mechanisms.

To test whether physical characteristics evolved in response to certain selection pressures, the team builds each robot slightly differently. To test theories on the evolution of the spine, for example, Long’s lab built several robots with different numbers of vertebrae. In a sort of “evolutionary Olympics,” as Long calls it, the scientists developed a judging scheme to assign points to different robots depending on how well they evade predators while feeding and procreating.

The researchers review videos of the TadRos playing the “game of life” with a predator robot to see which are most successful. That information is entered into a companion computer program, which is used to track the actual genes behind each design variation, and a “genetic” algorithm creates instructions for the next generation of TadRos.

Long calls it “CSI: Evolution”—using physical clues from the past to reconstruct what might have happened. “We would never claim we know exactly what happened,” Long adds. “All we can do is circumscribe the possible. We eliminate things that are less likely.”

One theory to come out of the work is that the pressure of escaping predators combined with the pressure to feed may have driven the evolution of vertebrae. Using the predator robot, Long and his colleagues found that the “feed and flee” response made extra vertebrae useful. “Now is that what actually happened? We don’t know,” Long says. “But it’s more likely than [a previous] idea of just selection for feeding alone.”

By building TadRos with varying tail shapes and lateral line systems the team also tested how selection pressures might have affected those traits. Even this simplified model uncovered potentially complex evolutionary patterns: as evolution of vertebrae slowed down, for example, tails got bigger and the oscillations of the lateral lines changed. Long says the bio-robots and the companion computer program enable him to connect evolution to behavior, biomechanics, and genetics.

While the TadRos’ bodies evolved, however, their “brains” and their “instincts”—such as whether to feed or to flee danger—remained constant, raising questions about the nature of intelligent behavior. “If you think that doing a better job at eating and not getting eaten means you’re smarter, then our robots got smarter,” Long says. “But they did so without evolving their brain.”

Long says one of the work’s most interesting discoveries is that the body may be partly responsible for what is recognized as intelligent behavior. “Intelligence isn’t only located in the brain … and the body itself is involved in giving us the kinds of behaviors we think of as intelligent,” he adds.

Long points to other ways this technology could be used to study evolution, such as studying how various organisms may react to environmental changes like global warming and higher water temperatures.

His team is also conducting research with the TadRos to try to explain the evolutionary reasons behind social behavior and the collective intelligence of large groups of animals, as well as how different species branch off from a common ancestor.

In one recent experiment, Long put several physically identical TadRos in a tank, all with only one sensory capacity—to find food, in this case represented by a light source. Unable to sense each other, the TadRos at first collided in the tank. But then Long explains, “after they bump around for a little bit, they start doing something amazing. They form a conga line around the light source.” The study showed how a coordinated group behavior emerged only from a shared goal and physical interactions.

The team contends that their results refute many models of swarm intelligence that assume that individual organisms in a group—fish in a school, for example—need to communicate with each other to act together.

In recent weeks, the researchers took the experiment even further, creating TadRos with different tail shapes. They found that those with similar tails schooled together, creating smaller subgroups by virtue of their biomechanics alone. “This, then, is a precondition for creating new species, which requires reproductive isolation,” Long says. “This is looking at evolution as it’s occurring now.”

Christie Rizk is a reporter and editor based in New York. She currently writes for the New York Genome Center, and has worked as a reporter, editor, and producer at Genome Technology magazine, Thomson-Reuters, and The Brooklyn Paper.

Tags: , , , , , ,

Leave a Reply

Your email address will not be published. Required fields are marked *