Slashdot Mirror

In Take 1 of this two-part series, I described some of the current technology in wearable computing as seen at ISWC 2000, this year's just-finished iteration of the International Symposium on Wearable Computers, from head-mounted displays to intelligent jackets. In this round, a little more speculation about where the future of wearability lies, with a peek in the door at some of the things being done in top university labs. What forces will shape the future of wearable computing? Hint: GeForce 2's are still hard to carry, and there's never enough power.

In square circle, with realtime 3-D overlay Professor Steve Feiner heads a Columbia University / Naval Research Laboratories joint project called Battlefield Augmented Reality System (BARS), one of the more graphically ambitious projects on view at ISWC 2000. He and graduate students Drexel Hallaway and Tobias Hoellerer demonstrated how, using transparent overlays representing maps and other data entered offline, allows the user to interact with others in 3-D environments, and to view information intelligently superimposed on real-world objects as they walk around. On the second day of the conference, the team was demonstrating maps they had entered over the course of the first day using the show-floor itself. One caveat: instead of a tiny metal box tucked discreetly out of sight, these systems approach the limit of wearability while in their design phase. The two systems on display, built separately, resemble backpacking rucksacks stripped of their bags to reveal a Rube Goldberg symphony of batteries, cables, video cards, and nondescript boxes which turn out to be GPS revievers and dead-reckoning devices.

"Because for what we're using, there's nothing you can buy off the shelf, or build really easily, that would be lightweight, low-power consuming and capable of doing the sthings that we want to do. But because we want to be able to go outside do some of the things that we want to do, we put these backpacks on. ... So we're building things that are bigger than they would have to be if they were built by a real oomputer company. It's mostly the fact that we need less the processor power -- because, as you know, the processing power of a high-end end laptop is damned close to the processng power of a high-end desktop -- the important thing to us is that the graphics card is a real desktop [card], basically the fastest PCI card you can get right now."

All that video power is still not enough for the taste of Feiner or his grad students, but it does present a dazzling display of rapidly-updated, colorful information on the monitors scattered around their exhibit. It looks like a video game, until you relaize that it's done in real-time.

In keeping with it's military funding and applications, these semi-wearables use centimeter-accurate GPS receivers ("Which is a waffling way of saying they can be accurate to within a couple of centimeters," says Feiner, "given that you have good satellite tracking performance."), and "in a lot of our current systems, the computer weighs less than the GPS." When GPS coverage is spotty or non-existent, the system uses its included dead-reckoning device to augment the GPS measurements.

"We like to call these bearables, rather than wearables," he grins. "This whole system weighs something like 20-odd pounds. Compared to a camping backpack, that may be pretty good, but for doing real everyday work? It's totally unacceptable."

"What are it's applicatons? At this size, and this level of kludginess? Research. The question is, what happens when it becomes small enough and rugged enough -- truly, they run sort of hand-in-hand -- to be truly wearable? And by truly wearable, I think less of Lizzy and it's predecessors, something that's bigger than my camera for example -- and more in terms of somethng like this --" he indicates the clamshell cell phone on his belt --"and this is even pushing it in terms of size. That's going to happen, in my prediction, sometime by the end of the decade. that prediction is based on technological feasibility and a lot of it is people, companies, whether they get it or not."

See Spot run! See spot compile the Linux kernel! Carnegie Mellon has a long history of wearables research, going back to the Vu-man series of computers in 1991. The current forefront of research at CMU is represented by Spot, a StrongArm-based research wearable in initial fabrication right now. "We'll get metal in November, and that means I'll be able to do the kernel bring-up around New Year's. Essentially, once you have an electronic layout, when you essentially just say what components are connected to what, with wires, there's an issue of trying to cram it into a box this big." That according to Spot's originator and designer, CMU doctoral student John Dorsey. Dorsey is also the ARM Linux kernel hacker for the board that Spot is based on. "This big," in this case, means a tiny card with the footprint of a PCMCIA card, itself enclosed in a 6"x3"x1" oblong pod, designed to snuggle in any of several spots on the human body. That tiny box holds up to 256MB of SDRAM, as well as up to 64MB of Flash RAM, and a CompactFlash slot. "And that's pretty incredible considering the amount of stuff that's actually on the board, so we hired an outside firm to design that for us, who will lay things out so they will fit in that area."

Spot is not destined to hit The Sharper Image anytime soon, though. "I guess our immediate deisre is to make it available to other resarch labs and universities," says Dorsey. "It's a little bit feature filled for it to be [the basis of] a cheap commercial product at this point."

What about the research that went into Spot -- will you be able to raise your own? I asked Dorsey about the possibility of the plans for Spot being put on the Web for public consumption, and his answer was uncertain, but optimistic. "There's an issue of [rights] when an outside funding agency gives the money to design something like that, they have first dibs on the intellectual property. There's a trend that I see at this conference, probably the most exciting thing I've seen, which is that a lot of companies are trying to make their designs available on the web, and that's the kind of thing I would like to see happen. At least for people who are in the research community, there's no reason not to do that. It's cewrtainly saves horrendous duplication of effort, and really helps us to make progress. The important thing is that you just want to get the platform done, and out there."

Thanks to the GPL, though, Dorsey says it's not as important whether Spot's schematics are available, becauses the clue that a home-builder would need will be in public view: "There's a really strange dichotomy when it comes to what people will permit for software, and what they will permit for software. People don't even blink when I say I want to GPL stuff that I write for research, my thesis or whatever -- but for hardware I think there's just a different mindset. What Ithink is funny when it comes to hardware is that I'm allowed to GPL the kernel port for this board. And any reasonably intelligent person could look at the code for the various drivers and support, and sort of back out what's actually in the box. So it's a little bit silly to say you can't say what the schematic looks like."

Like most of the academic reasearchers in Atlanta, CMU uses Linux heavily in its wearables development. Dorsey does have a short wishlist for future developments in the Linux kernel, though: "Power management, that's the big one. I know that the WRL folks over at Compaq are working on that noe for the SA-11 family [the processor which runs Spot] in particular. I would like to see it show up more often in networking research applications. BSD really does own the game as far as that goes right now, and I need to do all my thesis stuff on Linux, so I'd like to see that happen."

Also with CMU and living proof that it takes more than kernel hackers to build computers onto the human body, Francine Gemperle has a background in industrial design, and now is a graduate student with the CMU Design Studio. Gemperle was hired when she finished her undergraduate degree a few years ago, to help make small computers truly comfortable to the users. Now the only designer working with the group, Gemperle has been working since last February on the conference's organization. "They actually thrust the job of exhibits on me, and it's been fun, I've enjoyed coordinating it."

She's coordinated a group of about 25 volunteers who helped put this conference together, and has spent the last year arranging sponsorships, speakers and exhibitors. "Getting people involved in this conference is a tricky thing, because there are a lot of people who are doing resarch in wearables [not in attendance], but having it cleared through their lawyers and PR departments is another story. So there are many big companies that we know of that are actively pursuing wearables, but who aren't here. I think there were some people here from Motorola beofer, and everybody knows that IBM is pursuing wearables, but they don't really show up here." (IBM was demonstrating their Linux-equipped wristwatches, in fact, but did not exhibit any results of their wearable Thinkpad project.)

Charmed, I'm sure Brad Singletary, a PhD candidate at Georgia Tech, was one of the several poster exhibitors, explaining to passers by his current research project, which is one likely to have an impact on other wearable endeavors as well. Singletary is building a database of faces (which now includes a mugshot of me), recorded with a pair of helmet-mounted video cameras (one color, one black and white) onto a pair of DV decks worn in his clothing, to be used later as the basis of a recognition program.

Pointing to small white "X"s on the visor of his Glasstron display, Singletary says "I can line these X's up with your eyes, and then I push a button which records audio along with the video track, to tell me where your face is in the video stream. I push the button and then it just starts firing automatically."

He's trying to create a simple recognition algorithm rather than the unweildy multi-point ones currently in use. "It's differnet from the traditional face recognition; that's what the military wants, they want full 3-D recognition. I also want to help people with prosophagnosia - face blindness ... I'd like to help those people, give them some way to recognize people."

In addition to that disparate pair of users, he says, "Another is police, regular police ... infrared will allow them to function in the dark. But there are also more prosaic uses: " A typical businessman who walks into a conference, he wants to know 'Who do I know?' Who they are, and if he's met them before."

Of the odd-looking yellow piece on the back of his helmet, Thompson sheepishly admits "The dumbell is for balance. I built this thing as quickly as possible."

Singeltary's graduate advisor, Thad Starner, along with fellow graduate student Daniel Ashbrook and long-time wearables enthusiast and hobbiest (and Vassar College employee) Greg Priest-Dorman, are now not only academics, but entrepreneurs. All are part of Tharner's brainchild, Charmed Technology. Charmed's approach to wearable computers is different from that of well-known commerical producers VIA and Xybernaut: rather than invest in ultra-miniaturized cases and custom motherboard designs, they decided to go with commodity PC-104 boards, contained in a case which gains in affordability what it compromises in size. Designed by Priest-Dorman, the case features a slight bend for greter body conformance, and is designed to be worn either in a vest or carried in a hip-hugging bag.

Perhaps most exciting is that in addition to their plans to start shipping in November kit-based systems based on these designs for about $2,000 (a pittance in the wearables market), they're also making the plans available to anyone who cares to download them. "So we have these plans, and they're actually on the Web site, you can download them and build them yourself, you can take them to a metal shop and have them build it," says Ashbrook. "It's totally open source."

One polyfleece vest to bind them all! One of the few places with a strong claim as the birthplace of wearable computing, MIT remains perhaps the most famous breeding ground. While Charmed makes low-cost and easy assembly a priority,resulting in a slightly larger box, a project spearheaded by graduate student DeVaul called Mithril, out of MIT's Media Lab, takes a completely different tack, pushing the components of its wearable system into near invisibility. While nanotechnology hasn't yet advanced to the point where processor, memory and battery can simply be miniaturized and placed in the wearer's navel, the Borg Lab researchers (yes, they really call it the Borg Lab) are convinced that until wearable computers are nearly forgettable, they'll never catch on. To gain the benefits of ubiquitious computing, in other words, the annoyance can't outweigh the rewards.

I got the scoop on Mithril first from Media Lab Research Scientist Steven Schwartz, who started by showing me the "Smart Vest" of an earlier project. "It comes out the Human Design Group, which used to be called Vision and Modelling. It's professor Sandy Pentland's group, and it's the group where wearables first started at the MIT Media Lab.

"The system started off as PC-104 configuration. And that resulted in system that were medium sized boxes with stacked PC-104 architecture, mostly X86 machines, and primarily running Linux. These machines were used with a variety of sensors, including cameras. But as the wearables project has evolved, we've seen a couple of different needs. Power is one -- power is the real battlefield. Batteries are expensive, they're burdensome, they're heavy. In a big pciture sense, batteries are expensive, right? Not just cost, the whole inconvenience. So we're working on a low-power platofrm. And in order to conduct our own research, we wanted it to look a little more "everyday" like. We wanted to be able to dress and act as we normally dress and act, so that the wearable computers we're using to research were actually integrated with our lives and non-interfering with the actions that we take as mobile individuals."

"For instance, we didn't want the computer to interfere by covering our eyes, or requiring too much unser interaction, or simply be buly or odd-lookingm because because you want to go out and see where wearable computing can help you in your conversatons with people, you certainly don't want that kind of distraction. So we started working on a clothing computer, essentially -- but with one variant. We didn't want people to have to buy a specific outfit, or let's say have to throw out the old outfit because they've upgraded their computer. So we dicided to make a lining, essentially a detatchable, soft chassis. And because we didn't want it to be uncomfortable, we got rid of the boxes, the hard packaging. So we went to soft packaging, distributed the electronics in a sort of very wide area over the upper body, went to very small modules, and started to implement some of the very low power RISC architectures. StrongArm, PowerPC."

"We started building these, and engineering them to work with MicroOptical's displays, so you could have very small clip-on displays that don't cover much of your vision, yet provide lots of rich information. WHen we think of rich content for wearables, we don't thnk of walking down the street oozing through a UXGA screen, because with tat kind of content you'd probably trip right down the street. Cyborgs don't like this. So the idea is that the type of informatino that you want as a wearable user is stuff that's compatibile with high mobility, and high levels of interaction, unexpected physical events taking place."

"So we built the system called Mithril -- it's a new platform that all the research will be done on. It started over a year ago, when we did the smart vest project," Schwartz says as he steps into a vest from that project. "PowerPC, WaveLan, Linux, Apache. Samba as a remote server. We did it to show that we could put a high performance machine intoo clothing that could be worn summer or winter by a man or a woman. ... my advice to people is that the ultimate geek is chic. When you can hide all that big artillery, heavy metal that we carry around -- the ultimate wearable computer is being able to blend in with the people down the street. To walk down the most expensive block in New York, to see the most incredible radio on display, try it out, hold it, feel it, decide I want to buy it -- and on my eyeglasses is flashes where I can get it on the Internet for the best price."

DeVaul and fellow graduate student Josh Weaver explained their computer-in-vest concept. Each wore a custom-made vest outfitted with special pockets, some visible, some hidden. (That by the way, is the entire world's supply of Mithril systems right now."We'll go home on separate planes," DeVaul joked.) Displaying the the 200Mhz StrongArm processor in one pocket (part of the central core, called Brightstar), then the IBM microdrive and reader in another, DeVaul asks "You want to know how much this draws? Around 4 watts, for the whole system." In another pocket, two smallish batteries (like most projects on the floor, the MIT group swears by lithium-ion camcorder batteries, which in the words of one builder "have the energy density of a hand grenade"). But those are not the only components. Wireless networking lives in yet another pocket, and a small board with three IR LEDs (and one green one, so the wearer can know it's functioning) is worn like a campaign button. Those IR LEDs are used to uniquely identify people or objects, to aid the computer system in knowing where it is, or "remembering" the identity of a similarly equipped person you encounter.

The whole system is tied together with a home-brewed cabling system, part of DeVaul calls the "Mithril Body Network," which integrates power and data delivery (USB and I2C) over a flat braided cable designed to wrap smoothly around the wearer. DeVaul invited me to try on his vest, and it really does feel natural.

"Feeling Natural" in fact, is one big element of the design of Mithril. "One way to think about what [wearable computers] can do is to think of them as personal assistants -- helping you remember information or accomplishs tasks in everyday life," says DeVaul. Context-awareness and scaleability are the driving forces behind his designs, and from the looks of it, building a computer that can not only communicate well with the devices beyond its immediate reach, but do so while remaining unobtrusive is a heroic task. "You only want the computer to give a response, say, if it's appropriate for it to do so at the time. If you're asking me a question, you don't want the computer to answer you instead."

And though perhaps a tricker project to assemble the parts and skills to complete, than Charmed's PC-104 kit, DeVaul promises that anyone with the desire will be able to build a Mithril system. "I believe in Open Source software -- continuing the tradition of Thad Starner, everything, including the PCB designs, will end up on a web site."

When will the future arrive? Never soon enough, maybe, but the future does have a way of creeping up on us. The conventional wisdom at ISWC 2000 is that honest-to-goodness wearable computing affordable to all and with many of the current bugs shaken out, is about 10 years away. One person familiar with MicroOptical's displays said that their plans call for a drastic reducton in price over the next two years, to perhaps as low as one or two hundred dollars. Processor power, meanwhile, marches on. When a 600MHz Transmeta chip is old hat, it will still be enough to run the same applications it does today. I'm looking forward to joining the cyborgs when that day arrives.