Listen to the night music of cockroaches.
Listen to their tiny, spiny feet as they careen across the tiles in your kitchen. What do you hear? What can you learn? These hardy primordial creatures zip through cluttered spaces in utter darkness at human-equivalent speeds of up to 200 miles per hour. Yet you never hear them crashing headlong into things, even though the cockroach brain has only an infinitesimal fraction of the computing power of the average mammal’s. How do they manage this stupendous feat with such meager neural resources?
“A cockroach has two head-mounted antennae,” explains Johns Hopkins roboticist Noah Cowan. “They’re only about as long as its body, but somehow they’re enough to allow it to react in time to obstacles.”
Cowan, an associate professor of mechanical engineering at the Whiting School of Engineering, has been working with biologist Robert Full at UC Berkeley to understand the functions of the cockroach antenna-and apply these findings to robot design. Such antennae could help mini-bots move swiftly through earthquake-collapsed buildings, for example; they could even make Roomba-type cleaning bots better living room navigators.
Cowan has been working on this for the past decade, at first to understand the main problem of how the sensory feedback from the antennae helps the cockroach adjust its course. But a few years ago, one of his graduate students suggested that the tiny hairs on each cockroach antenna might have more than a sensory function. “It turned out that there’s a mechanical aspect to the antenna that we had completely neglected,” says Cowan.
The hairs are hinged in such a way that they tend to lie down against the tapering body of the antenna, and resist being pulled outward. The net effect is to enable the antenna to maintain its functional contact with a surface even as it skims along it at high speeds. “If it’s moving along a wall and there’s any roughness in the wall, which there usually is, the hairs get bound up in the roughness and cause the tip of the antenna, which is very floppy, to quickly curl back into a J-shape,” Cowan says.
Like most engineers, he’s used to approaching design problems with a modular strategy, in which each part has its own separable function. “But here, the sensory apparatus has a mechanical component, and the mechanical component is a sensor, all integrated into one complete package,” he says. “Teasing that apart, with biological and engineering experiments, has been a lot of fun, and obviously the next step is to translate it into something that works on a real robot.”
It’s just one of many projects at the Whiting School that take inspiration from the designs of living things-designs that have been worked out in the Darwinian trials and errors of millions, even billions, of generations. One could say that it’s all part of a massive, modern handoff of design responsibility, from the ponderous “blind watchmaker” of evolution to the faster and better-sighted tinkerers of the engineering world.