Skip to content ↓

Sea Grant unveils robotlobster

Thomas Consi hasn't yet donned claws and antennae and begun creeping about the ocean floor, but the MIT Sea Grant research engineer is doing his best to mimic a lobster. In collaboration with Jella Atema, the director of Boston University's Marine Program (at the Woods Hole Oceanographic Institute), Dr. Consi is developing an aquatic, lobster-sized robot that will duplicate the crustacean's keen chemical sensing abilities.

Lobsters are remarkably adept at detecting a chemical and then following it to its source. The animals accomplish this through sophisticated olfactory sensors located on two small antennules located between the bigger antennae. Since turbulence breaks up underwater chemical plumes into irregular patches of varying concentrations, finding the source of a chemical discharge is no small feat. For a lobster, trekking through chemical clouds means following waste plumes toward mussels and procuring dinner. For a robotic imitator, or biomimic, successful tracking could lead to a myriad of scientific, environmental and commercial applications. A robot fitted with chemical sensors could study red tides, find underwater vents, track fish or locate an underwater oil leak and send another robot out to fix it.

However, researchers must first determine how the lobster processes these chemical signals to achieve its ace maneuvering. Because actual lobsters are difficult to study in their natural habitat, Dr. Consi and his colleagues have created Robolobster, a surrogate animal that serves as an investigative model. Supported by MIT Sea Grant, the project will aid biologists and animal behaviorists in testing various hypotheses regarding how a lobster navigates. Then, the most promising hypotheses may be investigated with real lobsters.

The nine-inch robot consists of a plastic cylinder containing a computer, motor drivers and sensor electronics. Although Robolobster doesn't look like a lobster, it does manage to "see" the marine chemical environment much as a lobster would. Powered by 16 AA batteries and mounted atop two wheels and a plastic caster, the robot can approximate a lobster's speed and turning abilities. Dr. Consi describes Robolobster as a test-bed for various sensor designs and algorithms, step-by-step procedures for solving problems with a computer.

In lieu of more complex chemical sensors, Robolobster currently is outfitted with conductivity sensors, which measure salinity. In experiments conducted in a freshwater fish tank spiked with salt water, the robot's sensors gauged changes in salinity, steering the vehicle toward the salt source much as a lobster steers toward a chemical. The robot's conductivity sensors will be replaced with chemical sensors, and ultimately arrays of chemical sensors similar in size and activity to those of a lobster.

"It's like computer modeling, but one step better because you are able to give your model real-world stimuli," Dr. Consi explained. Eventually, the robot will be programmed with lobster algorithms postulated by Dr. Atema's research group. The vehicle then will be tested in the same tanks in which lobsters have been tested for plume tracking, thus providing comparable data.

Aside from the sensors, the vehicle is designed for simplicity. "It's good for students to play with," Dr. Consi said. Jamie Cho, a junior in computer science and engineering, designed the software for the little vehicle. That computer code will later be transferred to the chemical sensor. The vehicle's elegant mechanics were designed by MIT Sea Grant fisheries engineer Clifford Goudey, who left plenty of empty space to add sensors to this easily reconfigurable experimental platform.

Although other researchers have constructed biomimics-including surrogate bees, flies and cockroaches-Dr. Consi explains that Robolobster is one of the first robots to use chemical perception for guidance and navigational tasks. As it gains sophistication, the robot will be used to study increasingly higher levels of behavior in the animal on which it is based.

A version of this article appeared in the November 16, 1994 issue of MIT Tech Talk (Volume 39, Number 12).

Related Topics

More MIT News