The robotic fish kept alone in a tank at the University of Maryland, College Park doesn't dazzle with its agility or speed, but it does promise bigger things to come.
At the flick of a switch, water flows through the tank. The faceless gray plastic creature less than a foot long knows this, and you know that it knows this because it slides languidly from side to side to shelter behind a white plastic pipe, minimizing its energy use. The robot cannot see the pipe, but it can feel it.
It may not look like much, but this is progress for associate professor Derek A. Paley, his College Park research team and partners at Bowling Green State and Michigan State universities, who have been working on the hardware, mathematical calculations and computer program for a couple of years. The fish showed its stuff for the first time in a demonstration in February, a step toward a free-swimming robotic fish that responds to its surroundings much as a natural fish does. "How we actually achieve that remains to be seen," said Paley, of the University of Maryland's department of aerospace engineering and Institute for Systems Research.
The team has nine months left on a three-year, $625,000 grant from the Office of Naval Research before it will have to find new sources of support, and there's much work to be done. The next goals are a robot that uses the sense of water flow and pressure to avoid underwater obstacles and navigate tunnels.
Tom McKenna, who manages the Naval Research program, said the idea is to create an autonomous underwater vehicle that can find stationary objects by changes in water flow. It also wants the robot fish to be able to work in a group, sensing their spacing from one another, he said.
He said the robotic fish could be used in oceanographic studies, to find and disarm mines and to gather information on potential adversaries. He said that the research program is not linked to a weapons program and that the Pentagon "forbids weapons on autonomous systems."
The current model won't pass for a real fish. Made of a single piece of plastic knocked out by a 3-D printer, it has no face, no fins, nothing to resemble a fish except the rough contours of the body, which looks rather like a cross-section of an airplane wing.
The fish is equipped to detect water flow and changes in pressure with an array of electronic sensors based on the system of fish sense organs called the lateral line. Rather than hundreds of bundles of sensory cells along the body of a natural fish, this robotic version uses eight metallic "whiskers," each nearly an inch long, to detect the movement of the water. Complementing the flow sensors are four pressure sensors exposed to the water by tiny holes in the fish body, each about the size of the tip of a ball-point pen.
Wires run from the sensors through a post that connects the fish to a motorized device suspended over the 50-gallon clear acrylic tank. The system's electronic array conveys signals between sensors and motor, moving the fish in response to water flow and pressure.
During a recent demonstration conducted by doctoral student Levi DeVries, the fish was holding its position behind an object, in this case a length of PVC pipe, much as real fish do, saving energy they might otherwise use resisting the current. Shifting slowly from side to side, the robotic fish detects changes in the speed of the flowing water, using "flow sensor data to automatically determine its motion in the water tank without human intervention," Paley said in an email.
It's what he calls a "closed-loop control," or "feedback loop," meaning that there's no external control — from, say, a human operator. This is one element that distinguishes this robotic fish project from others.
Paley's group is one of several around the world looking at how the sensing and swimming capabilities of fish can be adapted to underwater robots. James Tangorra, an associate professor at Drexel University in Philadelphia who is also working on underwater robots, said Paley's group has made a few significant contributions.
It has made refinements to the system that tells the sensors what aspects of water flow they should be monitoring, he said, and figured out the best places to put sensors along the robot body, a more complicated matter than it might seem, as it's meant to mimic the function of the natural system of hundreds of sensors. He said Paley's group has also contributed significant work on the metal composite material used for the "whiskers," built by the team at Michigan State.
"We're not going to solve it all that soon," Tangorra said. "But you can have a lot of successes along the way." The gray plastic fish, for instance, swaying from side to side in a foot of water, "knowing" which way to move. That represents about two years of work, DeVries said. "Hopefully," he said, "it's got a bright future."