Teaching Robofish How to Swim

If a school of fish instinctively swims in the most efficient manner, Siddhartha Verma, Ph.D., hopes that a school of underwater robots can learn to do the same thing.

Verma joined FAU last fall semester on a joint appointment as an assistant professor of ocean and mechanical engineering in the College of Engineering and Computer Science and Harbor Branch Oceanographic Institute. Under the heading of fluid dynamics, he studies the specific fluid mechanisms of biolocomotion, or how fish swim. However, Verma spends little time on boats or hovering over tanks. “This is all very difficult to measure experimentally,” he said. “So, we approach it from the computational standpoint.” Computer models, he added, tested and improved against simple observational data, help create conclusions about more complex systems.

The data that Verma develops is used by other researchers at SeaTech – The Institute for Ocean and Systems Engineering. “They need people who can write code well, to control the experiments in a natural setting,” he said. One application of the research would be to design a better robot that moves through water like a fish – or even a whole cluster of them, which would be useful for underwater search and rescue missions. The idea is not new, Verma said. But in recent years, the technology has advanced so much that “if you didn’t look closely, it would be hard to tell that it’s not a real fish.” A concurrent step to developing underwater robots is teaching them to move together, drafting off each other’s wakes. Verma’s work combines the traditional field of fluid mechanics with advanced machine learning, developing algorithms in which the robots themselves find the best way to move efficiently.

Verma, who earned his Ph.D. at the California Institute of Technology, comes to FAU from the Computational Science and Engineering Laboratory in Switzerland. Going from cloud-covered mountains to Boca Raton is “a big change,” he said.