It knows its own strength

Though it may not look it, Domo is the first robot built to give a hug.

Typically, robots use small force sensors to tell how hard they’re pressing on something. But this only works if the sensors always remain in contact with the object. For example, if fingertip sensors don’t make contact with a lemon in the palm, you’ll soon end up with a glass of lemonade. Domo, the doctoral work of Aaron Edsinger at the Massachusetts Institute of Technology, was built to replicate a more natural sense of touch. Its “muscles”-motors called elastic actuators embedded in its fingers, wrists, arms and neck-sense how hard it´s gripping an object. The actuators aren´t completely rigid; the joints give a bit, much like ours do. Because of this flexibility, Domo feels an object pushing back against it and fine-tunes its grip like a good handshake-firm, not ferocious.


The talker of the town

Atsuo Takanishi and his Ph.D. student Kotaro Fukui [above] at Waseda University in Tokyo thought they had come up with a simple solution to a complex problem. Want a robot to speak like a human? Why not just build it like a human, vocal cords and all? Unfortunately, though, that´s where the solution´s simplicity ends. In order to simulate the flexibility and resonance of the biological structures, the group had to painstakingly re-create the entire human vocal system in this, the sixth version of the Waseda Talker. They modeled the human tongue, vocal cords, lips, teeth, soft palate and lungs out of plastics and rubbery polymers to mimic the real thing. The result is uncannily clear, natural speech-assuming you speak Japanese.


A new way to balance

Why make a robot that has to balance on a point? Because that also means it can turn on a dime. Ballbot´s omnidirectional ball-foot lets it trace any path along the floor without worrying about turning around or getting stuck in a corner-challenges that make wheel-based ´bots squirm. Its stability starts with fiber-optic gyroscopes that feed balance
data to an onboard control computer. They calculate how the ball has to move in order to keep the platform upright (or move it in the desired direction), and motors in the base make it so. According to Carnegie Mellon University professor Ralph Hollis [below], who built the first version of Ballbot in his suburban basement in his spare time, the robot can even cope with roughhousing. Give it a push, and it will sway and roll to keep its balance before returning to attention.


Gaining a sense of self

The infant robot Nico is slowly developing the smarts and social skills of a nine-month-old. It´ll need them (and more) if robots are ever going to interact gracefully with humans, who have the tendency to create a chaotic environment full of distracting motion. One way to cut the confusion is to pick out what in the environment is the robot itself (an outstretched arm, for instance) and what is another object or person. Yale professor Brian Scassellati and his Ph.D. students Kevin Gold, Ganghua Sun and Marek Doniec [above, from left] programmed Nico to match the video coming in from its eyes with arm-motor movement. It thinks, “I know something is moving, and I´m moving, and so it must be me,´ ” Scassellati explains. The result is a robot that recognizes itself in a mirror, a self-awareness never before achieved in a machine.


Mapmaking on the fly

Many robots are born, bred, and buried in the lab, prototypes eternally protected from unfamiliar environments. That fate won´t befall the University of Southern California´s Segway Robotic Mobile Platform (RMP), a ´bot designed to quickly map unknown and rapidly changing areas such as urban battlefields. Computer scientist Gaurav Sukhatme and his students, including Jonathan Kelly [above], loaded a Segway platform with GPS, video cameras, laser range finders and wireless communications equipment. The range finders constantly measure the distance to nearby objects, forming a rough outline of the landscape. Software fills in the rest, referencing patterns from past experience to turn the outline into a useful map of walls, trees and buildings. A version of the robot could be used to make detailed 3-D maps of hotspots in Baghdad, but for now, it´s still in training around the USC campus.