Homayoon Kazerooni is about to lift 500 pounds with one arm. The slender 36-year-old mechanical engineering professor is no bodybuilder, yet when he slips into an extender, he becomes a man of steel.
An oversize multijointed robotic arm, the extender is the brawniest of a half-dozen machines built by Kazerooni and a dedicated cadre of students at the University of California at Berkeley. All black metal, chunky, and menacing, it is hinged to a support pedestal like an abandoned prop from an Arnold Schwarzenegger movie. Motors bulge at every joint, and thick cables snake away to a nearby hydraulic engine the size of a desk.
Shoulder to shoulder with the extender, Kazerooni slides his right hand into a glove inside the machine’s hollow metal brace of a forearm, which extends a foot and a half beyond Kazerooni’s hand. At the end of the brace is the machine’s hand, two six-inch-square metal plates that open and shut like giant tweezers. Kazerooni’s left hand squeezes a hand-held safety switch, and the engine roars to life, drowning the lab in white noise. Kazerooni’s eyebrows cock, and a knowing smile breaks under his mustache. He begins to twist and bend his right arm, swiveling shoulder, elbow, and wrist like a tai chi master. Even though the pedestal- supported extender weighs 250 pounds, it moves fluidly and weightlessly along with Kazerooni’s arm. Then, when Kazerooni closes his right hand, the extender grabs a quarter-ton steel ingot and waves it triumphantly.
Welcome to the latest episode in the herky-jerky history of human amplifiers, machines designed to augment muscle power. Half gadgets, half garments, human amplifiers are the ultimate hard-guy hardware. But while they’ve long been a fixture in science fiction, they have yet to pull their weight in the real world. Engineers have been doggedly plugging away at them for more than 30 years, but they are technically complex. The results have been a series of elephantine contraptions and outlandish concepts. Kazerooni has inherited this dubious legacy and is bent on winning the human amplifier some newfound respect.
In theory, human amplifiers make a lot of engineering sense because they muscle aside some of the weaknesses of conventional robotics. They’re based on an intuitive means of controlling intelligent machines. You don’t twiddle a joystick or peck at a keyboard to command these robots- -you just wear them.
The most natural way of communicating is talking, says Kazerooni. Communicating with a computer through a keyboard--that’s not too natural. It’s the same for maneuvering an object. Maneuvering an object is this--and he mimes lifting and shifting a box. I don’t want to do it with a joystick. I want to feel it.
Human amplifiers sidestep another hurdle that has often tripped up roboticists: puzzling out and duplicating the choreography by which the human brain directs the arms and hands. How do we copy human intelligence into a machine? Kazerooni muses. We still have a long way to go. Humans are all over the place, so we might as well use them.
Throughout their evolution, human amplifiers have zigzagged between fiction and science. In Robert Heinlein’s 1959 novel Starship Troopers, for example, a band of interstellar warriors wore powered strength-magnifying outfits. By the early 1960s the Department of Defense began bankrolling research into superman suits. The aim was to beef up servicemen who manhandled cargo and torpedoes in the cramped quarters of ships and submarines. In 1964 Cornell Aeronautical Laboratory, a company in Buffalo with military funding, studied the feasibility of building a 480- pound robotic suit, or exoskeleton. It was supposed to be powered by a backpack-mounted gasoline engine. The company concluded that duplicating all human motions was impractical and that an exoskeleton with only a certain number of motions would suffice for most military tasks.
In 1965 General Electric moved the concept off the drawing board with a military-funded exoskeleton called Hardiman, a 1,500-pound, 30- jointed contraption that looked as if it had stepped out of a Japanese cartoon. Hardiman had two massive load-bearing legs, two arms, and a hip girdle, all attached to a smaller, jointed framework. The user was supposed to strap himself into the framework, and a network of complex hydraulic and electric linkages would power-assist his movements to drive the surrounding exoskeleton. Dean Martin’s sexy sidekick, Janice Rule, had no problem fighting off villains in a Hardiman knockoff in the 1967 film The Ambushers, but the GE engineers barely got a single arm to function properly. Hardiman was simply too large and bulky to maneuver easily.
Efforts then languished for well over a decade until Jeffrey Moore, an engineer at Los Alamos National Laboratory, dusted off the concept in 1985. Moore proposed an exoskeleton called Pitman, a powered suit of armor for infantrymen. But to his surprise the Defense Department, not normally gun-shy about funding far-out research, spurned his proposal. (At the time, exoskeletons were faring better in science fiction: in Aliens, James Cameron featured a cargo-moving suit that helped Sigourney Weaver do battle with the queen. Unbeknownst to the viewing public, the suit was made of plastic, and a weight lifter hidden behind it moved the limbs.)
Moore now realizes he stuffed Pitman with technology too pie-in- the-sky for even the Pentagon to swallow. To control the exoskeleton, he envisaged a network of brain-scanning sensors in the helmet. The devices would have measured the shifting magnetic fields generated while the brain dictated limb movement.
Life is a matter of timing, and I was ahead of my time, says Moore matter-of-factly. Those words come as the Army’s Human Engineering Laboratory in Aberdeen, Maryland, begins research into another powered suit of armor for infantrymen. On paper, the suit--part of an Army initiative called Warrior’s Edge--appears to be Pitman without the mind-reading trappings. Even so, Moore notes, for the generally conservative Army to move in this direction now is significant. The story of this whole amplifier effort has been little pieces of work here and there. Until now there’s never been anyone with deep pockets to bring it all together for a specific application.
Outside the defense establishment human amplifier research is also alive and well, if not exactly kicking. Take, for example, the prototype legs-only amplifier that’s already cavorting through the streets of Claremont, California. Dubbed SpringWalker, it’s the brainchild of a tiny outfit called Applied Motion. Cofounder John Dick, who conceived SpringWalker six years ago, doubles as a researcher at NASA’s Jet Propulsion Laboratory. Someday Dick hopes to have a contraption that will let a person run 30 miles an hour and bound to the top of four-foot-high walls. Soldiers could use SpringWalker to move quickly across hostile terrain, and civilians could spring around for exercise and recreation.
Take a pair of spindly stilts that are jointed like the hind legs of a gazelle, throw in some pulleys and a pair of bungee cords, and you’ve got SpringWalker. Unique among human amplifiers, the current SpringWalker involves no motors. You strap your feet to two pedals above a pair of leg frames with backward-bending knees. As you move your feet in a jogging motion, the long leg frames, extending below the pedals, provide leverage that doubles the push-off force of each footstep.
In addition, cables run from each leg frame to a thick loop of bungee cord on an attached back frame. When you put SpringWalker into a walk, the cables stretch and relax the bungee loop. As the loop is stretched, it captures much of the force that the leg frame exerts against the ground, energy that would be wasted during an unassisted jaunt. Raise your foot, and the bungee releases that energy, bouncing you into the air.
SpringWalker was designed to lend the user a kangaroo gait rather than the lumbering stomp of a weight lifter. But watching one of his partners bounce and rattle along a Claremont street during a test run, Dick notes that SpringWalker’s development is about as mature as a nineteenth- century bicycle. Our test pilot goes up and down quite a bit, he admits. We squander the energy in his gait with a lot of vertical motion. As it stands, it’s not any more efficient than just trotting along.
That will change, he vows, when he attaches small motor-driven winches to the cables. Then the motion of the wearer’s feet won’t yank the bungee loop directly. Coordinated by sensors on each leg frame, the motors will reel in and release the cables. The resulting power-assisted gait, Dick promises, will ultimately propel the wearer to run a two-minute mile with all the exertion of a leisurely jog.
At Berkeley, surrounded by his extenders, Homayoon Kazerooni is well aware of the checkered past and present of human amplifiers. And unlike the robotroops conjured up by Pentagon exoskeletons, Kazerooni’s extenders are destined for civilian work: unloading trucks, rearranging warehouses, toting steel girders, or simply bolstering crippled arms or legs.
In addition, Kazerooni’s approach to controlling his extenders breaks with the traditional technique known as master-slave control. GE’s Hardiman personified that strategy. Hardiman sported two sets of motors: a slave set that moved the joints of the exoskeleton, and a master set on the inner framework that pushed a fraction of this work back on the wearer, to keep the human aware of the load. That setup works fine in telerobotics, in which a human runs a robot from some distance. But when master-slave control is put into a system in which human and robot intertwine as one, the dual set of motors nearly doubles the weight and complexity of the exoskeleton.
Kazerooni’s extenders require only one set of motors. The lighter, streamlined arrangement is made possible by today’s fast, powerful computers and sensitive electronic sensors called load cells. These dime- size sensors are positioned where the human arm and hand touch the extender’s inner surfaces, as well as where the extender makes contact with the load. The load cells measure the force exerted by the human on the extender (such as the upward thrust to lift a load), along with the weight of the load itself.
Readings from these sensors feed into the computer controller, which commands the extender’s motors to turn just enough to relieve the human arm of most--but not all--of the load. For a 100-pound load, the extender might support 95 pounds while the human arm feels only 5 pounds.
This tactic strips away much of the bulk that ultimately crushed the Hardiman project. But Kazerooni now faces technical stumbling blocks that Hardiman’s creators never confronted.
Perhaps the knottiest problem is the inconsistent behavior of the human arm. Swing a weight from your knees to eye level, and you’ll notice that your muscles don’t work uniformly throughout the maneuver. When your arm approaches certain awkward angles, its muscles stiffen to compensate for the decrease in leverage. How much they stiffen depends on how strong you are.
If you’re wearing an extender while you lift the weight, the load cells will read this extra muscle stiffness as added force, as if the load has suddenly put on weight. The controller will direct the extender motors to pull up harder. The extender will jerk away from your arm, and the load cells will suddenly register a drop in force. The motors accordingly exert less torque, dropping the extender back on your arm. That, again, suddenly boosts the load on the sensors, quickly trapping the extender in an oscillation. Crank up the force and the wavering becomes even more likely.
To counter this tendency, Kazerooni is writing software that takes into account the arm’s idiosyncrasies. The software will also tailor the movements of off-the-rack extenders to humans of any strength.
Kazerooni’s extenders are designed in such a way as to solve these nuts-and-bolts issues. We want to go step by step, he says, and make sure all the fundamentals are taken care of. The aim is to build an extender with the full range of mobility of the human arm. To understand that mobility, Kazerooni has constructed a series of simpler extenders that mimic specific movements. For example, a shoulder-elbow extender, fitted with fast electric motors, probes the centrifugal forces touched off by high-speed arm swings. And a graceful tabletop extender copies the subtle movements of the human wrist.
Although his approach is methodical, Kazerooni wants to get his machines out of the lab and into production as soon as possible. He has no intention of letting Warrior’s Edge walk off with the honor of producing the first functional exoskeleton. Slumped in a corner behind the hydraulic extender is a hint of where Kazerooni’s efforts are leading. It’s an unpowered wooden mock-up of a full-body exoskeleton that makes the suit in Aliens look like an outfit from Romper Room. I have never been late or missed my schedule, Kazerooni insists. If you come back in two years this machine will be walking.
The mock-up is the fifth assembled so far. Each has served as a test-bed for alternative exoskeleton anatomies, anatomies that don’t simply ape the human framework. For example, in the current mock-up, the forearm extenders pivot forward from above like spider limbs, rather than up from below like human forearms. As counterintuitive as it might seem, the arrangement actually lends more comfort and maneuverability to lifting.
Kazerooni is also experimenting with unconventional leg designs. When the exoskeleton is transporting a heavy weight, it must channel the load through its legs into the ground without toppling over. One construction that may lend stability is a double-jointed leg that in effect has two knees, the upper one moving forward and the lower one backward.
If prompted, one of Kazerooni’s students will strap on the mock- up and stride around the lab like an animated scarecrow. When he can dance with the mock-up on, says Kazerooni, watching the student’s antics with interest, we’ll know it’s time to build this machine.