Imagine there’s a massive Bay Area earthquake. Buildings sway and then give; injured people are stuck on the top floors of one. Members of a rescue crew prepare to go in, hastily strapping on the tools of their trade: Helmets? Check. Backpack full of medical supplies? Check. Robotic legs? Check.
Sound far-fetched? Not so fast. Engineers have dreamed for decades of wearable human exoskeletons designed to give people extraordinary strength. Researchers around the world have struggled with a deceptively complex problem: how to design a mobile machine strong enough to help people lift heavy loads, yet lightweight and agile enough to walk smoothly in concert with them. The challenge has been to develop a robotic exoskeleton that can increase the strength of its human wearer without getting in his way, slowing him down, or squashing him flat.
General Electric was the first big firm to attempt the task. In the mid-1960s, it designed a behemoth called the HardiMan, which was envisioned as a suit of “mechanical muscles” that would enable its operator to lift up to 1,500 pounds. HardiMan was hardly lithe. It looked like a pair of girders strapped to its operator’s waist, forearms, and feet, topped by two clawlike hands. GE ultimately had to admit that it wasn’t yet possible to make a machine small enough for a person to safely wear. In fact, HardiMan was so gigantic and fearsome-looking that although the company’s engineers did test one of its arms, they never turned on the whole suit for fear that if it unexpectedly convulsed, it might rip apart the person inside.
For the next several decades, exoskeletons rarely made it past the drawing board. Much of the scientific interest came from within the biomedical community, which began to explore their potential to help paralyzed people walk again. Most notably, last year a Japanese company succeeded in creating HAL-3, the Hybrid Assistive Leg, a set of motorized leg braces with a backpack power source designed to help the elderly and disabled walk normally.
But if the world of mechanical engineering has largely failed to produce working exoskeletons, 20th-century pop culture was churning them out at a fantastic rate. Starship Troopers, Robert Heinlein’s 1959 novel about war between humans and aliens, is widely acknowledged as the exoskeleton’s literary debut. In the book, futuristic infantry soldiers wear “powered armor” that allows them to effortlessly jump over buildings and shoot arcs of fire in their wake. Heinlein’s troopers were outfitted in helmeted full-body suits so bulky they gave soldiers the appearance of a “hydrocephalic gorilla.” They served as full life-support systems complete with air and water supply, as well as a helmet-mounted visual display that the soldiers manipulated by pressing their chins on a control plate, and a communications system they activated by biting down on sensors inside their mouths. “That is the beauty of a powered suit: you don’t have to think about it,” the author presciently imagined. “You don’t have to drive it, fly it, conn it, operate it; you just wear it and it takes orders directly from your muscles and does for you what your muscles are trying to do. This leaves you with your whole mind free to handle your weapons and notice what is going on around you … which is supremely important to an infantryman who wants to die in bed.”
Another popular early version of the exoskeleton came from Stan Lee’s 1960s comic Iron Man, in which a war-wounded inventor, held in a prison camp, builds himself a metal suit of armor in order to protect his shrapnel-pierced body, and then goes on to battle Communist villains. Over the next few decades, movie and television writers created a host of characters that were part man and part machine: Robocop, The Terminator, The Six Million Dollar Man, and Dr. Miles Hawkins, the wheelchair-bound exoskeleton-wearing hero from M.A.N.T.I.S., to name just a few. Japanese anime gave viewers another set of hybrid heroes, as well as hugely successful series such as Robotech and Gundam which introduced American viewers to the related genre of mecha, giant human-shaped robots. More recently, video games such as Halo have made exoskeleton-like body armor de rigueur accessories for the first-person shooter set.
Perhaps the most memorable fictional depiction of an exoskeleton occurred during the climactic moment of James Cameron’s sci-fi thriller Aliens. Hero Ellen Ripley strapped herself into a bright-yellow forklift-like exoskeleton and used it to smack around a drooling, hissing insectoid alien. Her use of the exoskeleton essentially leveled the playing field: Where the alien had claws, Ripley had massive robotic pincer arms; where the alien had a protective outer carapace, Ripley now had her own metal shell. Moral of the story: It takes a bug to fight a bug.
All of these fantasies have contributed to the popular notion that exoskeletons are inherently violent. After all, Americans have been fed a steady diet of pop culture that has portrayed exoskeletons as imposing battle machines. It’s a perception that UC Berkeley mechanical engineering professor Homayoon Kazerooni eagerly seeks to dispel. Kazerooni and his team of Berkeley grad students recently unveiled their contribution to the field of exoskeleton research: the Berkeley Lower Extremity Exoskeleton, or BLEEX. From the outside, BLEEX appears far more sci than fi. It resembles a set of metallic leg braces topped by a large hiking backpack, at the base of which is a small plastic box containing the exoskeleton’s computer. The pilot stands neatly inside it, looking for all the world like someone about to go camping someplace very, very steep.
BLEEX cannot leap tall buildings, or even jump at all. It does not wield mechanized pincers or flamethrowers, or as yet even arms. Its exterior cannot repel weaponry of any type, although its interior features some intensely complicated feats of engineering. Yet thanks to the vast quantities of robot-based entertainment that most of us have ingested, Kazerooni and his colleagues keep having to answer questions about whether BLEEX is a violent contraption bent on global domination. “I get a lot of ‘So, you’re building the next Terminator that’s going to take over the world,'” says Cal graduate student Andrew Chu, who has worked on BLEEX for the last four years. “People are worried that we’re working on some super-military killing machine.”
After all, BLEEX is funded solely by the Defense Advanced Research Projects Agency, the offshoot of the Department of Defense that bankrolls cutting-edge military technology research. Just as sci-fi fans have long envisioned the day when smart but squishy humans could wrap themselves in the metallic embrace of a robot, so too has the Department of Defense, with its desire to make soldiers stronger and more technologically enhanced than ever. When BLEEX was formally introduced to the scientific and military world last month at a DARPA symposium in Anaheim, the Department of Defense hailed it as a boon to modern soldiers, who must carry heavy loads of rations, emergency supplies, batteries, and weapons on their backs. Conference attendees bombarded the Berkeley team with questions about how to make the machine faster, stealthier, and more powerful.
Perhaps it’s unsurprising that among the first public responses to BLEEX were some qualms about the advent of mechanically amplified soldiers. When news of its invention broke on the tech Web site Slashdot.org, some writers voiced satiric if misinformed reservations about how the government might one day use the new technology. “Bush wants his soldiers to carry back the oil a barrel at a time,” groused one poster. “If you protect soldiers from small arms, you only add incentive for everyone else to make larger arms,” worried another. “You can see the obvious cycle.” Someone else referred to the “FEAR OF GOD it would put into the soldiers when they see a 40-story-tall metal killing machine running at 100km/hr towards them.” A more enthusiastic poster wrote, “I, for one, welcome our robotically enabled masters!” Someone else crowed: “Imagine a soldier that could roll over a 70-ton tank. It would be like having an army of He-men. We could rule the world.”
Talk of this type easily gets under Kazerooni’s skin. He is insistent that his motivation is not a military one. Where his government funders may see supersoldiers, Kazerooni says he sees firefighters, rescue crews, UPS workers, factory and warehouse employees, and other people who lift heavy burdens every day. The exoskeleton will make their work easier, prevent back injuries, and maybe even help people with degenerative muscle disorders walk again, he says. “I am committed to make machines that are useful for workers, for people who do hard physical jobs,” he says. Kazerooni never tires of pointing out that the “exo” is merely a helpful machine, not something designed to blast through your doorway or kick over your car.
Like his students, the professor is a pretty tough critic of Hollywood’s ideas about exoskeletons. “Some of these machines that are actually made in movies violate the very first laws of physics,” he says with a laugh. To make the machines look exciting, he notes, prop designers have produced exoskeletons so enormous and top-heavy that if they were built in real life, they would tip over or be extraordinarily clumsy. They’re so inefficiently designed that, in a real battle, you’d be better off if the enemy troops were wearing them, he says. And although he says it would have been technologically possible for his team to design a somewhat larger-than-life machine and still make it work, at a certain point it would have become a vehicle, rather than a wearable device, which defies the whole point of the human-augmentation project. Smaller is better, he says, even if the results are less cinematic.
And what about that rumor circulating on Slashdot that Kazerooni was consulted by the Aliens production team regarding the appearance of Ripley’s power-loader exo? “I think there was a phone call,” he says bashfully. When pressed, he’ll admit that he also gave a bit of advice to the crew working on the Spielberg movie A.I., and recently got a phone call from Emeryville animation studio Pixar asking for more aesthetic guidance. But he stresses that he has never taken money for dispensing technical tips to moviemakers, and that his involvement was limited to phone calls and office visits — he has never worked on-set. “My connection with the movie industry has been no more than to solve a few problems and point them towards the right solution, perhaps,” he says. “On the part of writers and moviemakers, creating a machine that impresses the viewer is the goal. What is our goal? It is to bring science and engineering together to make a useful machine.”
In fact, the next phase of Kazerooni’s project is to make his team’s creation less imposing and menacing.
The magic word you hear around the BLEEX lab these days is “transparent.” This is code for “so lightweight and quick that the user can’t feel the machine.” Ideally, the exoskeleton’s machinery would be cool enough that it will avoid infrared detection, quiet enough that its user can’t hear it, and small enough that it could fit beneath the wearer’s pants and go unnoticed by observers.
After all, the exoskeleton’s defining design feature already is what it doesn’t have: no keyboard, no joystick. Its user simply walks, and the machine walks along with him, providing the motorized muscle to carry the payload in the backpack. As Heinlein had envisaged, there are no “driving” skills to learn and nothing to encumber the hands; the user is free to turn his critical thinking skills to whatever task is at hand. The machine does the gruntwork; the human does the thinking. As Kazerooni puts it, “This is the very first machine that integrates human intellect with the strength of a robot.”
But at least for now, the exoskeleton itself is hard to miss. Kazerooni takes an appraising look at one of the prototypes. One leg of the exoskeleton has been encased in transparent plastic to allow viewers to peek at the elegant metal skeletal structure underneath; the other side has a plastic sheathing in Army green. Kazerooni studies its thigh and shakes his head impatiently. With the plastic outer layer attached, the exoskeleton protrudes about six inches from each of the wearer’s legs. “To me, that is huge,” he says. “This is way too much. I want to make the thing about this size,” he adds, making a chopping gesture that would reduce the thigh by half. Will he be happy if he can get it down to three inches? Or perhaps two? “You know what the specs are?” he asks with a grin. “As small as you can get it.”
Kazerooni’s students are of the right age group to have grown up with plenty of pop-culture images of robots. “I think all of us were robot freaks,” says Chu. “I personally loved Legos, Voltron, Transformers, Robotech. I however am a cynic by nature, so I was the kind of kid who eventually got fed up with the sci-fi and toys since they seemed to disobey the laws of physics and had mechanisms that I just didn’t believe could actually work. I guess that’s why I went into engineering.”
There are now thirteen students and four staff members working on the BLEEX project, all aiming for the less-is-more aesthetic. “Transparency is kind of like the holy grail in this thing,” Chu says. “Ideally, once you strap this thing onto the human and power it up, the human doesn’t feel it’s there. They’ll move however they normally move and the robot just follows them.”
“Like a shadow,” nods Ryan Steger, a fellow grad student with several years on the project.
It’s a novel approach to constructing an exoskeleton. Traditional robotics — not to mention the robots often portrayed in movies, comics, cartoons, and video games — have given us images of robowarriors all but completely encased in machinery, moving in distinctly unhuman ways. BLEEX has sought to upend all that, in part because people feel creepy and claustrophobic if they’re made to haul too much machinery. There are only a few places on the body where people can comfortably bear weight for long periods of time without chafing or bruising, so BLEEX is designed to connect to the person at only three points. There are brackets underneath a pair of modified boots where the machine attaches rigidly to the feet, the backpack hangs off the shoulders and straps across the chest, and fabric straps attach slightly beneath the knees. “Even that is questionable,” Kazerooni says, pointing at the knee straps. “Some people will take it off.”
Part of the Berkeley team’s unique approach to building the exoskeleton is that instead of being driven by the user, the machine mimics the user’s movements. Picture a dad with a young child standing on his feet: The dad stands behind the child, but the two walk together, the dad carefully adapting his pace to the movements of the child. That’s how BLEEX works, using an array of networked sensors located at strategic points in the machine’s “bone” structure and in the soles of the boots to anticipate how the person wants to move, and forwarding information to a central computer. “It’s the same way you have a local area network in your office so all the computers get connected and can pass information,” Kazerooni says. And because the sensors are built into the leg structure, there are no buttons to push, no steering wheel to turn. “You simply walk,” he adds.
Future exoskeleton designs may endow users with other forms of super-strength: say, the ability to travel long distances without tiring. For now, BLEEX is primarily a load-carrying machine, a second set of mechanical legs that can walk along with the wearer. But getting a machine to move as fluidly as a person does is no easy task. For safety’s sake, when the Cal team practices walking the machine either in the laboratory on the university campus or at a field station in Richmond, the users generally walk between a set of guide rails or are connected by a wire rope to an overhead crane that can catch them if they start to tip over. People who have worn BLEEX say that it feels heavy while you’re strapping it and a fully loaded backpack on, but once the power is activated, the load lifts dramatically, giving the impression that you’re carrying only ten or fifteen pounds on your shoulders, even though you are carrying many times that. “It’s not powered in all the same places as the human, so you still feel the load in kind of strange ways when you walk,” Chu says. “The human body has hundreds of muscles in all sorts of places controlling all sorts of motions, while we have six hydraulic actuators.”
Watching BLEEX walk without the plastic outer covering over the legs is an exercise in seeing something that is both solid and fluid at the same time. The user is undoubtedly inside a machine; the metal leg braces run down the outside of each leg from hip to beneath the heels. Among the silver parts of the machinery, a series of thick black cables snake down each leg. But the ease of movement also is readily apparent. The braces easily swing backward and forward from the hips and bend at the knees. When the wearer’s knee is lifted up, the metal brace juts slightly in front of the wearer’s limb; otherwise, the exoskeleton’s shape conforms closely to the outline of the human leg.
Some movements are currently beyond the machine’s capabilities — kneeling or jumping, for example — but BLEEX is able to smoothly shadow users walking at a normal pace, including going up or down inclined slopes and steps. People wearing the machine tend to step confidently, with the deliberate gait of someone wearing ski boots. Occasionally, Kazerooni says, the machine gives a slight jerk; it’s one of the kinks he wants to iron out in future prototypes. “It looks like you are dancing with someone and you are dancing well, but at some point the guy you are dancing with has a move that you don’t like,” he says.
Since users are tethered like acrobats on a safety wire, nobody has ever wiped out while wearing the exo, although they have sometimes confounded the machine’s ability to mimic them. For very rapid movements — an involuntary reflexive kick, for example, or the wearer attempting to turn on a dime — the current design has difficulties keeping up. Those kinds of super-quick movements create additional demands on both the system’s engine and its software controls. “It requires more power because reflexes absorb a lot of energy,” Kazerooni says. “So I have to have a burst of power at the time I need it. Also, it’s a control problem because I have to make sure the burst of power happens at the right time. If it comes late it’s not good enough.”
Copying how people walk was a technical challenge, given that the human leg is naturally a complicated machine containing dozens of bones, muscles, and joints capable of a wide array of movements. “You think you’d just put a joint at the knee, you just put a joint at the ankle,” Chu says. “In actuality, when you look down at the biomechanical scale, the movement is much more complicated. It’s not just a simple rotary joint. There’s lots of sliding, there’s a lot of little joints in the human.” The Berkeley team’s solution was not to try to exactly replicate the human anatomy, but to design a machine with enough “sloppiness” built into it to accommodate the many variations in people’s gaits.
On top of that, building a computer model that could handle not only the complexity of leg movement, but also factor in environmental variables such as the slope of the ground, was a mind-mangling math problem. The Cal team’s solution was to incorporate into its design the computational abilities of its human wearer, who already does a good deal of movement-related math without knowing it. When people walk, they unconsciously do a sort of real-time calculus, figuring out how high to step, how to keep balance, how to cut the best path through irregular terrain. “You’re already wired to do that,” Chu says. “When you look at other people who are doing similar research with walking robots and things like that, they have to have these really, really high-end computers and have these crazy mathematical models that are just guessing at things like friction and the squishiness of your shoes and the terrain. Whereas our approach tends to be to rely on the human to take care of that.”
So the person walks; the machine copies how they walk. The result is a unique melding of the user’s cognitive abilities and the machine’s muscle. “The way we consider it, the exoskeleton is the human plus the machine; it’s not just the machine by itself,” Steger says.
But of course, mechanical muscle is heavy. According to Kazerooni, BLEEX currently weighs about 120 pounds and allows the user to carry 70 to 80 pounds in addition to the machine itself. The team would like to flip-flop those figures within the next two years. In order to minimize the weight-to-power ratio, the team has two tricky problems to solve: finding the right engine and the right power source. Right now BLEEX can run for a couple of hours at most, but DARPA envisions it being used on eight-hour missions. Batteries are heavy to carry and the exoskeleton is meant to work in field conditions, rather than in the lab near an electric socket, so the Cal team chose to use gasoline. Sure, it’s liquid and flammable, but as Kazerooni points out, gasoline packs a lot of power into a small amount of fuel, and is available anywhere in the world. It was harder to find the right small high-performance engine for the job, so the team built its own, a hybrid motor that provides hydraulic power for locomotion, and electric power for the computer and the sensors. Most of the BLEEX prototype engines have been built using parts from remote-controlled airplane engines or what Chu refers to as “cannibalized lawnmower-weedwhacker-type stuff.” This may sound a little silly, but as the students like to point out, it works. “A lot of people joke around, ‘Why the hell are you putting a weedwhacker engine on the back of this thing?’ but it’s like, you crunch the numbers and it’s hard to beat the weedwhacker,” Chu says with a shrug.
Engineers and the military have long believed that exoskeletons could be useful, but the failure of the HardiMan project convinced them that they’d have to wait a little while until they became more feasible. But by the turn of the millennium, things had changed. Kazerooni says the success of BLEEX is largely the result of advances in computer processing, which have made sensors and computer networks faster, more powerful, and more capable of governing nuanced movements. “This machine has an incredible set of computer controls which is very elaborate,” he says. “It’s years of programming to get it into that phase. This never existed during the ’50s with the HardiMan.” Scientific understanding of how humans walk had improved, and technological advances had made component parts — from engines to actuators — both smaller and more capable. “I learned a lot by studying work the military and some other people had done back in the ’60s and ’70s, and a lot of the technological barriers back then had been solved by now,” Chu says. “A lot of those things that brought these thirty-year-old projects to a grinding halt, if you look at the technology nowadays, the same barriers aren’t there anymore.”
That same idea had occurred to the people at DARPA as well. “Military folks have always had to carry large packs and a lot of weight, so it’s not a new need in the military,” says spokeswoman Jan Walker. “But up until now we haven’t been at the point technologically to consider a solution. We felt that there had been technology developments in actuators and power sources and computer processors and software that might allow us to do what we hadn’t been able to do in the past.”
Over the last few years, much attention has been given to the idea that future wars will be fought in urban environments, where soldiers may fight on foot and without the protection of vehicles. No longer encased in tanks or trucks, soldiers are being armed with ever-increasing amounts of weaponry, protective devices, and communications gear they must carry on their bodies. Although DARPA does not currently imagine that every future soldier will go into battle tricked out with an exoskeleton, the agency believes that wearing them would help soldiers lug all of that gear to the battle. “The idea of the exoskeleton would be to allow them to do the approach march and then get to their mission and not be tired from carrying everything,” Walker says. “It would help them be fresher and more capable during the firefight.” She also points out that helping soldiers bear more weight would allow troops to wear heavier body armor and carry more supplies and ammunition.
DARPA also envisions other uses for the technology, such as exoskeleton-equipped soldiers single-handedly loading missiles onto airplanes, thereby making it possible to reload a plane anywhere, instead of having to send it back to a base. Exoskeletons also could help medics evacuate injured personnel from the battlefield, Walker says. But she is careful to stress that DARPA doesn’t try to control design details or specify a new technology’s possible future applications — once a working prototype exists, the different branches of the military services can tweak it to suit to their own needs. “What we want to do is demonstrate the technology,” she says. “We have to develop it, integrate it, and prove that it works. Then, once that technology is available to the military services, they’ll figure out what they want it to look like and how they would use it, if they use it at all.”
In 2001 DARPA launched its Exoskeletons for Human Performance Augmentation program, which budgeted $50 million and five years to develop a working lower-extremity exoskeleton. Originally, six teams were chosen to develop prototypes, and only two teams made the cut last fall to continue receiving funding: the Cal group and the Sarcos Research Corporation in Utah. DARPA reports that Sarcos’ exoskeleton, dubbed ALEX for Autonomous Lower EXtremity system, first walked unaided last December. The Cal team beat them to it by several months.
In many ways, Kazerooni seems the logical choice to lead research into building a machine that would integrate human movement and robot strength. The professor has spent the last fifteen years researching the technology related to exoskeletons and human augmentation. Kazerooni attended the Massachusetts Institute of Technology, where he studied mechanical engineering and did research on manufacturing and human-machine systems. He then spent four years as an associate professor at the University of Minnesota before coming to Berkeley in 1991, where he is currently director of the university’s Robotics and Human Engineering Laboratory.
Even as a child, Kazerooni says, he was always at home around machines. He recalls a youth spent building go-karts and tinkering in his high school’s laboratory. “I was always interested in technology and making things with my hands,” he says. “I was fascinated with mathematics. I was always in the shop, tooling around, basically.”
He was drawn to the logic and precision of machines. “I feel comfortable with machines, and if you want to ask me why, it’s that when machines don’t work, don’t perform well, there is a good reason for it,” he says. Kazerooni also paints and sculpts, a process he says is not inherently different from building something mechanical. “There is a beauty in machines,” he says. “I actually look at machines as sculptures, as a piece of art. I think a lot of us do that but we don’t confess.” He points out that many people will admire the design of a car on the street, or even the intricacies of the machinery underneath the hood.
Kazerooni says his interest in exoskeletons was sparked by his disappointment with the limitations of autonomous robots, which are simply programmed to perform repetitive tasks in structured environments — a robot spot-welding in an auto plant, for example. But in a chaotic situation, where he pictured an exoskeleton being the most useful, Kazerooni doubted the abilities of an autonomous machine. A successful rescue worker must be able to make split-second choices, to prioritize decisions, and to rely on sensory input that most robots don’t have. “Even with the state of technology in artificial intelligence, it’s just not going to make the robot as good as a person,” he says. Humans, on the other hand, are comparatively wimpy, but are good at dealing with free-form situations. “They have experience, they have memories of what they have done in the past, they have the ability to think very quickly,” Kazerooni says. “Therefore, I might as well augment a person.”
Kazerooni had a second reason to want to improve human abilities, rather than replace them. Some of his early research had put him into close contact with the manufacturing world, such as the assembly lines at General Motors. Although he could see the need for autonomous robots in hazardous work environments such as nuclear plants, he says he could not condone the replacement of human workers by robots simply because they’re a cheaper labor source. “I could not accept this culturally,” he says. “I could not see how a beautiful person with a beautiful way of thinking about problems could be replaced by a strong but dumb robot.”
So while the Department of Defense may be motivated by making exoskeletons useful to soldiers and field medics, Kazerooni prefers to talk about his invention helping FedEx workers load trucks or rescue workers carry the wounded from earthquake sites. The hallmark of good robotic design, he says, is to make the machine’s production so efficient that everybody who needs it can afford to use it. “It’s always easy to build a very safe car with a price of $60,000,” he says. “That to me is not a challenge. But can you provide the same level of safety for less than $20,000 so everybody can use it? That is beautiful engineering.”
Some might question why a project headed by a professor so interested in its potential civilian and humanitarian applications is being funded by the Department of Defense. Kazerooni points out that DARPA is unique in the tech world for its willingness to pay for long-term research projects that may never lead to a profitable product. “Most institutions fund projects where there is for sure a return there,” he says. “I always like to do these risky projects. DARPA traditionally has been funding projects that are high-return/high-risk things, and I love to do that, so it was a great match.” Kazerooni says he is also interested in collaborating with other groups, including local fire departments and advocates for the disabled or others who have a possible use in sight for exoskeleton technology, as well as manufacturers developing materials that might make the exoskeleton more lightweight.
Likewise, Chu says he isn’t losing any sleep over the exo’s funding source. “The exoskeleton is being funded by DARPA as new technology with potential military applications, not specifically as a weapon of war,” he says, adding that his team has never been asked to design anything vaguely weapon-related. He points out that venture capitalists are unlikely to fund such a new technology, much less the educations of a dozen mechanical engineering grad students. “Without military funding, most of our lab would be more worried about getting enough money to eke out a degree than pushing the edge of technology,” he says.
A box-loading device may not be as sexy as a fighting machine, nor the premise of preventing back injuries quite as exciting as creating übertroops. But Kazerooni is cautious about overhyping his own design, or spinning elaborate theories about how it might be used in the future. All he’ll say is that it’s useful for carrying heavy loads and making them feel light. “I’m pragmatic,” he says. “Everybody around the department tells me that I’m just quite practical about my work, and sometimes being in Berkeley it may not be such a good idea. But practical things are important these days, considering the limited resources we are going to have in the future. Making things practical under the cold light of reality is hard.”
Ironically, Berkeley’s exoskeleton is funded by perhaps the one branch of the federal government that concerns itself with the fantastic more than the practical. As an agency, DARPA’s directive is to dream big and to dream weird. As Walker puts it, “Our mission is to avoid technological surprise, so the US is not surprised by its adversaries, and to create technological surprise for our adversaries.”
The agency was founded in 1957, only a few months after what was arguably the Cold War’s biggest technological surprise: the Soviet Union’s launch of Sputnik. DARPA is dedicated to funding the “far side” of development: that means high-risk projects that may yield a high payoff, fundamental research for new technologies, and engineering problems that are deemed “DARPA-hard.” The agency prizes innovation above all else. Program managers are replaced every four years so that they can’t mother-hen favorite projects; the agency has no laboratories of its own in order to discourage it from developing institutional interests; and it doesn’t take requests from the military about what to develop next. Instead, the idea is to develop technologies that may be useful to the military far in the future. “The military doesn’t know to ask for something if they don’t know it’s possible,” Walker says. “We like to be able to show the art of the possible.”
Over the years, DARPA provided the funding for once-cutting-edge but now-standard technologies. Its Tacit Blue stealth fighter program led to the development of the B-2 stealth bomber; DARPA also funded research into the cruise missile engine that led to the Tomahawk. Several DARPA-funded inventions also have filtered down to quotidian use. The Internet owes its genesis to DARPA’s development of TCP/IP network protocol architecture, and the agency facilitated Stanford Research Institute’s invention of the computer mouse.
But over the years, DARPA has had its share of ideas fall famously flat. Perhaps most notorious are the agency’s research into the possibility of telepathically spying on the Soviet Union, or its attempts to build a mechanical elephant that would enable soldiers to navigate the jungles of Vietnam. Last year, the DARPA-funded Futures Markets Applied to Prediction program proposed setting up a sort of stock exchange designed to predict world events by allowing investors to bet on the likelihood of palace coups and terrorist strikes. News of the project resulted in such moral outrage from Congress and the public that the plan was hastily scrapped and the program’s head, retired Admiral John Poindexter, promptly resigned. And last month, the DARPA Grand Challenge, in which $1 million was to be awarded to the team that could pilot an unmanned robot racer from California to Nevada, ended in infamy after most of the fifteen vehicles flipped over, wandered off course, or otherwise disqualified themselves within sight of the starting line. None of the entrants completed more than eight miles of the 142-mile course.
Some of the agency’s ongoing projects sound no less eyebrow-raising. It is currently funding experiments to help soldiers deal with long periods of sleep deprivation, the development of a Matrix-like mind-machine interface that would use brain activity to control machines directly, a project investigating whether bees can be trained to sniff out explosives, and one that is remote-controlling rats by planting electrodes into their brains and electronically stimulating their pleasure centers to reward them for good behavior.
The folks at DARPA claim to not know what the agency’s failure rate is for getting these projects off the ground, partly because so many of the proposed projects never get to their stated goal but do end up spinning off something else of value. “Maybe you started out trying to develop X,” Walker says. “You didn’t get X, but you got Y, and Y is really great. So is that a failure?”
Where do exoskeletons fit into DARPA’s experimental range? “When I first started this project, I thought we would be on the far end of the weird curve,” Chu observes wryly. “The more I learn about DARPA, the more it seems we’re on the pretty-safe-bet side of the curve.”
Last month, the BLEEX project made its formal debut at the DARPA Technical Symposium in Anaheim, where grantees gathered to show off their progress. “It’s like a science fair with $5 million production value,” Steger says. Except at this science fair, most of the exhibits are top secret. The Cal researchers say most teams’ booths feature videotapes of their projects working perfectly, but never the math or drawings explaining how they work. “They have a big flashy poster and a lot of acronyms,” Steger says. “But when you start asking technical questions, everybody kind of freezes up because it’s their proprietary idea and they don’t want another scientist or researcher to get wind of it,” Chu adds.
The BLEEX project was very well received by the military brass in attendance, the researchers say. Buoyed by the anything-is-possible vibe of the event, symposium attendees had plenty of requests. “Didn’t they come back and say ‘Well, can it swim?'” Steger recalls, turning to Chu for confirmation. “‘Can you make an exo in space?'”
“They had all sorts of crazy stuff like that,” Chu agrees. “Different branches of the military, they wanted to jump buildings, to be ultra-stealth, to run many, many miles without stopping.”
Kazerooni’s team expects that once its basic research is complete, its results will be turned over to defense contractors who will modify their design according to military specs. And what about making exoskeleton technology available to private entities such as FedEx? According to DARPA’s Walker, universities own the intellectual property rights to the new technologies they develop, and barring any constraints that might govern classified research, after their DARPA grant is complete the schools usually have the option of licensing their technology to private manufacturers or spinning off their own companies to market their product. But Kazerooni’s students caution that some of the wilder hopes about what an exoskeleton can do may simply be impossible. “You’re not necessarily limited by the technology,” Chu says. “Some of it is you’ve got hard constraints on what a human can do.”
Even with the help of a machine, people can run for only so long before they get tired or hurt. If you can’t do the splits by yourself, an exoskeleton can’t do them for you. Or take jumping over a building: “If you build a suit roughly the size of a human that a human straps on and you try to make him leap a building, any physics professor in five minutes can prove you’re going to tear the crap out of the human,” Chu says. “It’s almost like we have to train the people who give us the money that there are certain things that are just far, far more difficult to do, and there’s a lot of scientific reasons why we can’t do these things that they want us to do.”
BLEEX, however, will soon be able to do a good deal more than it currently does. For example, while it is very good at mimicking a person who is walking normally, it has a harder time shadowing sudden movements. Since a soldier on the battlefield may suddenly duck, weave, or trip, it’s important for BLEEX to be able to do these things smoothly.
And of course, there is the issue of transparency. The current engine design is still too noisy and too warm, Kazerooni says, although Chu notes that the team soon hopes to have a muffler that will make it “quieter than a cat.” The team wants to somehow alleviate the fact that now, before and after the power switch is thrown, the exoskeleton is still uncomfortably heavy to the wearer. “We may not be able to make a part of the machine smaller in terms of its size, but we might be able to make it for example lighter, or to make its location somehow not as intrusive as it was before,” Kazerooni says.
During its DARPA grant’s two remaining years, the team expects to build another set of improved prototype legs and then go on to add a set of arms by 2005. Research into wearable arms and hand-like manipulation is already much more advanced than research into legs and lower-body locomotion, so the Cal team is tackling that part last. “You can almost get existing robotic systems and tack them onto the top to deal with arms,” Chu says.
In the meantime, they expect that other research teams will soon be turning out rival designs. For example, according to DARPA, the Sarcos team also expects to complete an upper- and lower-body prototype by 2005. “There are going to be a lot of exoskeletons pretty soon, like in a year or two, coming up from bright places: universities, institutions, and laboratories,” Kazerooni says. “They will probably do what we sort of prescribed.” Then he thinks about it for a moment and adds, “Hopefully, better things.”
Even though the exoskeleton has long captivated the imaginations of sci-fi fans, it’s an idea that may become more palatable to the general public as people become accustomed to wearing more and more technology on their bodies. “Progress towards that era is unavoidable,” Kazerooni says. “We’re on that path already. We have GPS with us, we have helmet-mounted displays with us.” And for those who think that being draped in machinery and communications gear is only for soldiers, he points out that ordinary people have turned themselves into mini-cyborgs thanks to the portable technology they carry with them every day. Take the cell phone; it’s small, noninvasive, and it gives people a power once unimaginable — the ability to instantaneously talk from anywhere to someone on the other side of the world. Or how about the technologies many of us wear every day without even noticing: watches, contact lenses, pacemakers, hearing aids. Could there be a day when people strap on exoskeletons as blithely as we put on our glasses?
For now, the Cal team’s foremost accomplishment is to have produced the first working exoskeleton prototype, proving that they are a viable field for future research. Someday soon, exoskeletons will work better. They’ll be smaller. They’ll be able to go further on less fuel. Who knows? Maybe we’ll end up using them to throw alien queens out of the airlock. Or maybe they’ll just make a bunch of postal workers very, very happy. Right now, the far side is the limit. “This is an infant project,” Kazerooni says. “It’s a beginning. In fact, we never really finished the problem. We simply introduced the problem to the community, the problem of the exoskeleton. It has a long way to go.”