disappearing hardware - cs.cmu.edujasonh/courses/ubicomp-sp2007/papers/04-disap… · however,...

36 PERVASIVEcomputing 1536-1268/02/$17.00 © 2002 IEEE

R E A C H I N G F O R W E I S E R ’ S V I S I O N

Disappearing Hardware

For many tasks today, the use of comput-ers is not entirely satisfactory. The inter-actions take effort and are often difficult.The traditional, and still prevalent, com-puting experience is sitting in front of a

box, our attention completely absorbed in the dia-log required to complete the details of a greater task.Putting this in perspective, the real objective is thetask’s completion, not the interaction with the toolswe use to perform it.

To illustrate the point further, if somebody asksyou for an electric drill, do they want to use a drill,or do they really want a hole? The answer is prob-ably the latter—but computers are currently very

much a drill, requiring knowl-edge, training, effort, and skill touse correctly. Creating a hole isrelatively simple, and many hid-den computers invisibly accom-plish what seem to be simpletasks, such as regulating our cars’brakes. But unlike other inani-mate objects, a computer systemmight be able to infer the resultautonomously and affect thedesired outcome, increasing both

its potential and end-user complexity. Realizing thispotential while managing the complexity is the fun-damental challenge facing computer systemresearchers.

An important trend over the last decade is theemergence of specialized, task-specific hardware

(such as spell checkers, calculators, electronic trans-lators, electronic books, and Web pads). Thesedevices have a specialized interface and address thedesired goal of ease of use. In contrast, a PC is a gen-eralized machine, which makes it attractive to pur-chase—the one-time investment having the potentialfor many different uses. However, in other ways itadds a level of complexity and formalism that hin-ders the casual user.

Mark Weiser wanted to explore whether we coulddesign radically new kinds of computer systems.These systems would allow the orchestration ofdevices with nontraditional form factors that lendthemselves to more natural, tacit interaction. Theywould take into account the space in which peopleworked, allowing positional and manipulative1—rather than just keyboard and mouse—interactions.Along with specialization and the use of embeddedcomputers, support for mobile computing and wire-less data networks is an important facet of thisvision—in other words, invisible connectivity. A goalof this exploration is that we would learn to buildcomputer systems that do not distract the user; ide-ally, the user might even forget the hardware is pres-ent.2 In essence, Weiser was proposing that well-designed computer systems would become invisibleto the user and that our conscious notion of com-puter hardware would begin to disappear. Someyears later, Don Norman popularized this conceptin his book The Invisible Computer.3

In this article, we survey the progress towardWeiser’s vision from a hardware viewpoint. Where

In Mark Weiser’s vision of ubiquitous computing, computersdisappear from conscious thought. From a hardware perspective, theauthors examine how far we’ve succeeded in implementing this visionand how far we have to go.

Roy Want and Trevor PeringIntel Research, Santa Clara

Gaetano BorrielloUniversity of Washington and Intel Research, Seattle

Keith I. FarkasCompaq Western Research Laboratory

JANUARY–MARCH 2002 PERVASIVEcomputing 37

have we been, where are we, and where arewe headed? What characteristics will makehardware disappear from our conscious-ness, and what will it take to achieve them?

Where we’ve beenThe research community embraced

Weiser’s call to explore ubiquitous com-puting. For example, his vision inspiredthe work at the Xerox Palo Alto ResearchCenter (PARC) in the early 1990s andsuch projects as ParcTab,4 Mpad,5 andLiveboard.6 Olivetti Research’s ActiveBadge7 and Berkeley’s InfoPad8 projectsalso embraced this research direction, asdid other notable centers of excellence,such as at Carnegie Mellon University,IBM, Rutgers University, Georgia Tech,and the University of Washington. Unfor-tunately, many of the early systems werebased on technologies that were barelyadequate for the task, so they fell short ofdesigner expectations.

Figures 1a and 1b illustrate the extent of hardware improvement over the lastdecade. In 1990, no Wireless Local AreaNetwork standards existed; the processorssuitable for mobile devices operated at onlya few megahertz, while PCs were typicallyshipping with up to 50-MHz processors.The early electronic organizers (pen-basedPDAs had not been invented) proudlyclaimed 128 Kbytes of memory, while PCsshipped with 30-Mbyte disks. The displayswere also quite crude: laptops used mono-chrome VGA, and the few handhelddevices available mainly used character-based displays.

Industry soon responded to the challengewith a tighter focus on mobile computing.A flurry of early products hit the market,particularly in the tablet style, that tried tomake using computers feel more like using

a pen and paper. Most of these productshave fallen by the wayside: Momenta andEO, IBM’s early ThinkPad, and later theApple Newton, the Casio Zoomer, and Gen-eral Magic’s pad. For these designs, the ben-efit-to-cost ratio was just not large enough.To be successful, these new devices had toeither be better than the traditional pencil-and-paper technology they were replacingor provide desired new functionality. Thephysical hardware was the dominating fac-tor, and almost every design aspect affectedacceptance: size, weight, power consump-tion, computation speed, richness of inter-face, and simplicity of design.

We started to cross the acceptabilitythreshold only in the latter half of thedecade with the Palm Pilot. It was smallerand lighter, focused on simple applications,and incorporated a novel one-buttonapproach to data synchronization. Finally,an electronic organizer was useful for a sig-nificant number of people and had realadvantages over the more traditional paperproducts such as Day-Timers. The com-puter industry was beginning to move inthe right direction.

Where we areFor hardware to disappear from our

consciousness, we require transparency of

use: if we notice it’s there, it’s distractingus from our real task. For example, if wenotice that we are using a slow wireless net-work connection instead of just editing ourfiles, then the action of accessing the files isgetting in the way of the real task, which iscontained in the files themselves. If the linkis fast and robust, we will not notice it andcan focus on the content. Likewise, if a dis-play can present only a poor representa-tion of a high-quality underlying image, wesee a bad display. A high-quality displaysuspends our belief that the image is onlya representation.

The four most notable improvements inhardware technology during the last decadethat directly affected ubiquitous comput-ing are wireless networking, processingcapability, storage capacity, and high-qual-ity displays. Furthermore, the current pop-ular adoption of emerging technology, suchas cell phones and PDAs, strongly indicatesthat the market is generally ready foradvanced new technology. This adoption,however, requires common standardsacross many products and locales.

Wireless networkingAlthough progress in wireless connec-

tivity was initially slow, it has increased.This area has witnessed two distinct devel-

(a) (b)

Figure 1. Hardware improvement overthe last decade: (a) the Xerox ParcTab,the first context-sensitive computer(1992). The design shows the limited display available at that time—a 128- ×64-pixel, monochrome LCD. (b) a typicalPDA available today with a color 240- ×320-pixel VGA (transreflective) screen.

38 PERVASIVEcomputing http://computer.org/pervasive


opment trends. The first is in short-rangeconnectivity standards, such as Bluetooth(IEEE 802.15) and the IrDA (Infrared DataAssociation) standards, which are primar-ily for simple device-to-device communi-cation. Bluetooth, which will get its firstreal test in the marketplace in 2002, wasdesigned as a short-range cable replace-ment, allowing for proximate interactionand the discovery of resources in the user’slocality. IrDA had a similar aim. Butbecause infrared signaling requires a lineof sight, users had to physically placedevices next to each other, often an incon-venience. This technology, which predatesBluetooth by many years, has been con-sidered a market failure. (The sidebar listsURLs for Bluetooth, IrDA, and other areasof interest in this article.)

The second trend is in wireless LANtechnology, such as the 11-Mbit-per-sec-ond IEEE 802.11b standard and the morerecent 54-Mbps IEEE 802.11a standard.

Wireless technologies provide for twobasic needs: the ability to detect locationand the more basic ability to communicate.In many cases, the ubiquitous computingvision can to some degree be implementedby interpreting simple context information,such as a user’s location.9 Accurately deter-mining in-building locations is difficult.The Global Positioning System (GPS), forthe most part, can locate an object only towithin 10 meters and does not work inbuildings. Short-range wireless standardssuch as Bluetooth let us more easily discernlocation and context within a limited areawithout having to support sophisticatedwide-area location technologies. Further-more, short range implies lower power andthe ability to build a smaller device from amore compact energy source, making itmore attractive for many human-centricapplications.

The emergence of the latter IEEE wire-less standards allows for communicationcells that span many hundreds of feet withsufficient bandwidth to make us feel as ifwe were connected to a wired LAN, butwithout a physical connection’s con-straints. IEEE 802.11b has already beenwidely adopted, and 802.11a is expectedto follow with higher bandwidths. Next-generation digital cellular networks suchas 2.5G (for example, General PacketRadio Service and NTT’s DoCoMo—withgreater than 24 million users) and the com-ing 3G networks will extend these capa-bilities to cover entire metropolitan areas.

The wireless networking of today andthe immediate future thus enables portableubiquitous hardware that remains con-nected to the global infrastructure. How-

ever, to date, wireless networks have laggedbehind the bandwidth capabilities of theequivalent wired networks, leaving boththe opportunity and the user desire forimproved wireless hardware.

Processing capabilityFor 30 years, processing capability has

basically followed Moore’s law, which canbe summarized as “The number of activedevices we can place on a given area of sil-icon doubles every 18 months”10 (This isactually a revision of Gordon Moore’s1965 estimate of doubling per year and thelater 1995 estimate of doubling every twoyears.) This trend’s obvious consequenceis that we can continue to increase thecapability of devices fabricated in a givenarea of silicon. Instead of designing systemsbuilt from separate board-level compo-nents, we can integrate diverse functional-ity onto a single chip, resulting in remark-ably compact consumer electronics.Similarly, the reduced capacitance result-ing from smaller transistor dimensions

means that we can operate these devices athigher speeds, increasing their effective per-formance. Additionally, reduced transistorsizes decrease power consumption, allevi-ating some of the perpetual problems sur-rounding energy storage technologies.

The combination of more transistors ona given area of silicon and a reduced powerbudget has brought us the capabilities ofmid-1980’s desktop computers in today’sbattery-operated, handheld PDAs. Twoexamples are the Motorola Dragonballand Intel StrongARM processors, the mostcommon processors used by today’s PDAs.Besides providing low power consumptionand high performance, these processorsintegrate their DRAM and LCD con-trollers and a host of other interface I/Ocapability on the same die.

These trends directly affect ubiquitouscomputing for mobile devices in two ways.First, we can better match algorithmic com-plexity and execution speed to real-worldproblems. Second, the resulting power con-sumption allows for a reasonable operatingtime before batteries fail. These propertieshave let us build ubiquitous computing hard-ware that we can adapt to a greater range oftask-specific activities. Nevertheless, furtherimprovement is still required to satisfy thecomplete ubiquitous computing vision.

Storage capacityA less visible industry trend is the rate

at which storage capacity is improving forrotating magnetic storage and solid-statedevices. For 25 years, the capacity ofrotating disks has been roughly doublingevery year, a rate of improvement fasterthan Moore’s Law! The current storagedensity found on a disk drive is approxi-mately 10 Gbits per square inch.11 Today,IBM markets the Microdrive, a one-square-inch device with 1 Gbyte of stor-age in a compact flash-card format. If thistrend continues, in another decade we willbe able to carry 1 Tbyte of data in a sim-ilar form factor.

From a ubiquitous computing view-point, storage is becoming so inexpensiveand plentiful that, for many devices, inter-nal storage capacity is not a limiting fac-tor for basic operation. Moreover, we can

A truly ubiquitous computing experience

requires high-quality displays to let us see

through the display process and effortlessly

acquire the underlying information.


begin to use storage in extravagant waysby prefetching, caching, and archiving datathat might be useful later, lessening theneed for continuous network connectivity.Flash memory, DRAM, and Static RAMhave all benefited from progress in inte-gration, making large capacities at lowprices possible. A typical DRAM chip hasapproximately 32 Mbytes of capacity;flash memory has 16 Mbytes. Top-endCompactFlash cards, with 512 Mbytes ofcapacity, closely rival the IBM Microdrive’scapacity.

Storage is fundamental to most ubi-quitous storage applications. Storagelimitations for common tasks hinder theubiquitous computing experience, forc-ing us to leave behind information andto carefully select what must be mobile.Abundant storage negates these issues,letting us focus on the important under-lying tasks.

High-quality displaysBecause vision is one of our most impor-

tant and acute senses, the need to renderinformation at a very high quality cannotbe overestimated. Low-quality displaysdistract us by drawing our attention topixelation, granularity, and poor repre-sentations.

The last decade has seen a remarkableimprovement in display technology: mostcommercial laptop PCs now have a 13-inch color TFT (thin film transistor) LCDdisplay at XGA resolution with a viewingangle of at least 140 degrees. However,these are mainly transmissive displaysrequiring a backlight that accounts forapproximately one-third of the device’stotal power consumption. So, from a sys-tem viewpoint, these displays are far fromideal. Some LCDs have a resolution thatexceeds 300 dpi, making them suitable for x-ray-quality pictures. Quality has also in-creased for very large displays, such as the60-inch diagonal plasma display that Sam-sung showed at Comdex 2001. Such com-mercially available displays provide themeans for shared display workspaces,which have been the subject of researchfor some time (for example, the StanfordI-Room12).

At the smaller end, PDA displays havealso improved. In the past year, PDAssuch as the Compaq iPAQ have used trans-reflective color LCDs. These displays cantake natural light entering their surfaceand reflect it back through the LCD stack,yielding considerable energy savings.Although all these displays look promis-ing, they still have scope for improvement:the contrast ratio is lower than mostprinted magazines, and the resolution alsoneeds to increase to the level of printmedia.

Ultimately, the need for improvement indisplay technology is bounded by thehuman eye’s visual acuity, but we still havesome way to go. A truly ubiquitous com-puting experience requires high-quality dis-plays to let us see through the displayprocess and effortlessly acquire the under-lying information.

Adoption trendsFor the year 2000, the annual sales of

PCs were approximately 150 millionunits,13 a remarkable statistic about therate of adoption of computer technologiesinto all aspects of modern life. However,in that year 8 billion embedded processorsmade their way into the infrastructure ofindustry and electronic consumer devices.So, the fraction of PC sales is a mere 2 per-cent of the total processors sold. From thisstatistic, it is obvious that processors arebeginning to be deployed ubiquitously.

Similarly, another market trend is the

convergence of previously separate tech-nologies. Devices such as cell phones,PDAs, and digital cameras are beginningto merge. Their combined capabilitiesreduce the number of devices a user mustcarry or own. Such convergent devices willlikely succeed because numerous deviceswill no longer burden users.

In contrast, the problems that a unifieddevice’s size and complexity create havegiven rise to single-function informationappliances. These appliances are easilyadopted because of their simplicity andlow cost. When personal mobility is notthe main motivating factor, such divergentdesign approaches become attractive—forexample, customizing a computer to besolely a spell checker. Such a device isphysically better suited to its task, but thenwe have many diverse devices to keeptrack of and maintain. Even so, organizedusers can “accessorize” their devices for aparticular task. There is an ongoing ten-sion in ubiquitous-hardware designbetween convergent and divergent devices.

We can see that our first taste of ubiqui-tous computing is already in the PC’s twostrongholds—the office and home—and aunique third domain, the automobile. Theoffice has benefited from an integration ofapplications and services driven by theneed for coordinated enterprise solutions.For example, the Blackberry, a wirelessdevice that integrates with corporate email,has created a considerable followingamong the corporations that have adopted

Related URLs� Aibo: www.aibo.com

� Blackberry: www.blackberry.net

� Bluetooth: www.bluetooth.com

� DoCoMo: www.nttdocomo.co.jp (in Japanese)

� Electronic ink: www.eink.com

� IEEE 802.11 Wireless Local Area Networks:

http://grouper.ieee.org/groups/802/11/index.html

� Infrared Data Association (IrDA): www.irda.org

� Intel StrongARM processors:

http://developer.intel.com/design/pca/applicationsprocessors/index.htm

� Power MEMs research:

http://web.mit.edu/aeroastro/www/labs/GTL/research/micro/micro.html

� Universal Plug and Play: www.upnp.org

� Versus Technology: www.versustech.com



it. Such examples show how business pres-sure is pushing the development of inte-grated systems, which are based on ubi-quitous computing principles.

The home is also a natural driver for ubi-quitous computing system design. We areat an early evolutionary stage where somehomes have established a high-speed net-work connecting multiple PCs. Consumerproducts are emerging that exploit thisinfrastructure—the home computer systemis slowly subsuming all the home commu-nication, entertainment, information deliv-ery, and control systems.

The automobile is a particularly out-standing area of success for ubiquitouscomputing. Modern cars have computersystems that integrate control of the engine,transmission, climate, navigation, enter-tainment, and communication systems.Success in this domain is largely becauseeach car manufacturer has had completecontrol over all aspects of its subsystems,unlike the home and office domains. Also,power for computation is not a limitationbecause it is a small fraction of what isneeded to accelerate the car.

Where we’re headedUbiquitous computing focuses on getting

computing “beyond the desktop.” Thisimmediately presents a set of research chal-lenges associated with removing the cus-tomary stationary display-and-keyboardmodel, because for most people the PC isstill their focus. Most current ubiquitouscomputing research projects fall into twocategories: personal systems, which includemobile and wearable systems, and infra-structure systems, which are associatedwith a particular physical locale. For bothcategories, novel interaction modalities,such as speech or pen processing, become anecessary component because they don’trequire bulky displays or input devices.

Outside these two categories, the inter-action of computers with the physicalworld, without direct human involvement,is rapidly becoming more important as thetotal amount of deployable computationincreases. A system’s ability to proactivelymonitor and react to the real world is instru-mental for truly ubiquitous computing,

where not only the hardware but also theresults of actions must be removed from theforeground. Similarly, robotics is an emerg-ing field that mobilizes a computer andenables it to effect change at arbitrary loca-tions in the real world.

Personal systemsPersonal systems give users access to

computing independent of their physicallocation at the cost of them having to carrysome equipment. Discrete portable devicessuch as PDAs and cell phones are currentlythe most useful personal systems. How-ever, these systems tend to be limited bytheir computational ability, integrationwith other devices (for example, your cellphone communicating with your PDA),and interface capabilities (such as the dis-play’s size and quality). Today, computa-tional ability and integration are not hin-dered by a fundamental limit—advancesin processors and short-range wireless tech-nology will eventually solve these prob-lems—but interface capabilities are.

Some systems, typically termed wearablecomputers, rely on hardware such as heads-up displays and one-handed keyboards toprovide the interface to the computer. Thismodel is attractive because it provides afully functional computing experiencewherever the user might be. However, theseinterfaces can be overly intrusive, requiringa great deal of the user’s attention; this mit-igates their widespread acceptance. Cur-rently, these devices are typically fairlybulky belt-worn devices, but they willshrink as technology progresses, therebylending themselves to better industrialdesign and integration.

Personal servers14 (see Figure 2a) can en-hance ubiquitous access to data. They formthe computation and storage center of aperson’s digital experience. Devices suchas a cell phone or PDA-like appliance com-municate directly with this central server,providing a common representation of auser’s data. These devices, therefore, merelyrepresent an interface into this centralrepository. With this model, the personalserver could be located out of easy reach—for example, in a user’s shoe, handbag, orbelt clip—without causing inconvenience.

Or, it could even be combined with anexisting device (such as a cell phone).

Extending this model, a user’s personalserver could also support interactionthrough interfaces in the surrounding infra-structure. Users could access their personaldata through a public kiosk or borrowedlaptop display (see Figure 2b). Thisextended model is attractive because itallows a favorable user experience (inter-acting with personal data through a largedisplay) without requiring users to carrythe display. The personal server has recentlybecome tractable owing to advances inshort-range wireless technologies (forexample, Bluetooth) and low-power pro-cessing (for example, StrongARM), whichcan now support an acceptable userexperience.

Infrastructure systemsUnlike mobile systems, infrastructure

systems instrument a particular locale. So,many of the difficulties shift from issues ofsize, weight, and performance to those ofdeployment, management, and processing.For example, imagine thousands of minia-ture temperature sensors deployed arounda room—How did they get there? How isthe data collected? How are faulty com-ponents identified and replaced?15 Tasksthat are tractable when the human/com-puter ratio is close to one suddenly becomedifficult when the number of computingdevices increases.

Several hardware projects, such as theBerkeley motes,16 have started to explorethis space by creating a fairly small wirelesssensor platform. This lets researchersactively explore the networking protocolsnecessary to organize large sets of nodes.Currently about the size of three stackedUS quarter-dollar coins, these devices willkeep shrinking until they reach the size ofsmart dust17 and can no longer be seen ordirectly manipulated. The Berkeley PicoRadio project is pursing a single-chip sys-tem that incorporates both processing andradio frequency subsystems.18

Although the basic hardware of embed-ded devices can be relatively simple, con-siderable hardware challenges remain.Power is a primary concern. Even though

each node might have a battery lifetime ofseveral months, the mean time to batteryfailure for the entire system can be short ifit contains many devices. Furthermore, theenvironmental impact of these systems isnot well understood—dropping a thou-sand sensing devices out of an airplane ona disaster zone is relatively easy, but howdo you reclaim the devices?

The fundamental problem with shrink-ing hardware devices is that they quicklybecome too small and numerous for peo-ple to relate to them: they are literally outof sight, so they will quickly become out ofmind. One direct byproduct of these mul-tidevice systems is that the individual net-work nodes will not be named in anyhuman-understandable way—there willsimply be too many of them to keep trackof. System security rapidly becomes a bigconcern: how do you know if a node shouldbe listening in on a wireless conversation ifyou can’t even keep track of which nodesyou have in the system?

Proactive interactionBoth personal and infrastructure-based

systems ultimately require some kind ofuser interface to let humans interact withthem. Nondesktop interface modalities,such as pen, speech, vision, and touch, areattractive in ubiquitous computing systemsbecause they require less of a user’s atten-tion than a traditional desktop interface.However, the mainstream use of these tech-niques depends largely on their hardwarerequirements and interface capabilities.

For example, pen computing has largelybeen successful on PDA-class devices witha well-defined and accessible display. But it

has no place in systems with either verysmall displays or no display at all. Speechand vision interfaces have done well ininfrastructure-based systems that have bothsignificant computation resources and a static environment.19 But they have trou-ble in mobile systems that are computa-tionally impoverished or need to operate indynamic environments that might be, forexample, very loud or dark. Complextouch-based interfaces20 that deal with asignificant amount of data input or outputare problematic. They tend to be task spe-cific, with interface hardware crafted for aspecific application, so they have not gainedwide acceptance.

The most exciting advances in ubiqui-tous interfaces will likely be new displaytechnologies that enable rich visual outputwithout a bulky flat-screen display. E-Ink(electronic ink) Corp., for example, uses asystem of microcapsules to create flexibledisplay surfaces, which would be consider-ably more convenient than a traditionalrigid LCD panel. Such technology, in con-junction with an abstracted computationmodel such as the personal server, brings usone step closer to a world where we canaccess personally relevant informationquickly and conveniently, without relyingon bulky, fragile display systems. Interfacehardware technology poses different diffi-culties for mobile systems than for desktopcomputers. For example, speech interfacesfor mobile devices are inherently problem-atic because they often operate in noisyenvironments, requiring noise cancellationor directional-microphone techniques.

Directly integrating computing with thereal world, without human intervention,

will greatly increase computers’ impact onour lives by removing people as a major lim-iting factor from the processing stream. Afew automatic systems have already signif-icantly affected our lives: thermostats, air-plane autopilots, automated factories, andantilock brakes, for example. Such systemsstill require human intervention: thermostatmaintenance, airplane landings, factory con-struction, and deciding when to apply thebrakes. The challenge is to make these sys-tems proactive, where they can anticipateand react to physical world conditions (forexample, deciding to apply the brakes),instead of just reacting to them (for exam-ple, deploying an air bag when you crash).21

The salient distinction between these twomodels is that one is human-centric, whichrequires close involvement to effect correctoperation, while the other is human-super-vised, requiring minimum involvement butstill achieving an intelligent, useful result.

A proactive system must closely and reli-ably integrate sensors and actuators withthe physical world. This task is closelyrelated to building the infrastructure-basedsystems we described earlier. However,proactive systems will require greatersophistication in the components deployedin the environment, both to enable thecapability to affect the physical world andto quickly, robustly, and accurately pro-cess real-world data. For example, for abuilding-scale temperature-monitoringapplication, slowly reporting distributedtemperatures to a central server would besufficient. However, an earthquake re-sponse system would need to activelydampen vibrations at many nodes through-out a building.


Figure 2. Personal servers: (a) Intel Research’s prototype; (b) using resident display devices such as a wall-mounted display, or acommunity tablet computer, to view personal data stored on a user’s personal server.

Personal server(a) (b)



From a personal perspective, proactivecomputing pushes us toward systems thatcan monitor and affect our bodies directly.For example, automatically handling dia-betes requires the ability to both monitorblood glucose levels and administerinsulin—tasks requiring specialized hard-ware. A system such as this would stillrequire significant advances in softwarereliability to be feasible.

RoboticsNow more popular than ever, robotics

presents an interesting confluence betweenmobile and proactive systems: allowing acomputer system to affect the real worldwithout a priori instrumentation of theenvironment. Most robotic systems eitherperform a very specific task, such as factoryautomation, or are remotely controlled bya person. In general, autonomous robotshave difficulties dealing with dynamic phys-ical situations—primarily with determin-ing where they are, as well as with identi-fying objects in the environment. Similar tohandheld devices, the hardware for con-structing robots is shrinking rapidly, mak-ing them cheaper and more capable.Recently, the first real consumer robots—

namely, Sony’s Aibo (see Figure 3)—havereached the market. Abstractly, robots area novel type of disappearing hardware: theyallow computation to directly affect the realworld without heavily instrumenting theenvironment.

Software issues aside, significant hard-ware challenges exist, particularly for thesensing technologies that will enable proac-tive robotics. Vision is the canonical tech-nique for a robot to determine where it is,what objects are around it, and where theyare. The generalized vision problem re-quired to deal with dynamically changingphysical locations is quite difficult. Onesolution is to instrument a particular envi-ronment with sensors, beacons, or both toaid a robot in a particular locale—theequivalent of GPS for buildings. A high-accuracy indoor-positioning system wouldbetter allow an autonomous robot to func-tion within a building’s confines. Addi-tionally, tagging interesting objects in theenvironment would help a robot identifyand locate them. By exploiting infrastruc-ture-based computing, as we discussed ear-lier, robotics can make significant stridestoward being truly proactive and auto-nomous without solving the generalized

location problem mentioned in the previ-ous paragraph.

Hard problems & physicallimitations

Three hard problems have been the corechallenge of ubiquitous computing hard-ware and will likely continue to be far intothe future: size and weight, energy, and theuser interface. These problems transcendindividual points on the technology curve,partly because they are somewhat contra-dictory—as we’ll see, a solution in onespace greatly confounds that in another.Ongoing solutions to these problems,therefore, will have to come from “outsidethe box” and remove the fundamentalroadblocks.

Size and weightSize and weight significantly impede

ubiquitous computing because they con-tinually remind the user of the hardware’spresence. This limitation manifests itselfboth in mobile systems when the user mustcarry the device and during the setup andconfiguration of infrastructure-based sys-tems involving many pieces of equipment.Appropriately, the two main contributorsto a device’s size and weight stem from theother two fundamental problems: batteriesand the user interface.

The Itsy pocket computer (see Figure4)22 exemplifies this situation. It is just 70percent larger than its 2.2 watt-hour bat-tery and 320- × 200-pixel display. In addi-tion, these two components represent 60percent of the total weight without the casebut 43 percent with the case.

Overall, device sizes have been shrink-ing at an incredible rate owing to system-level integration. But devices are reachingthe point where they can’t get much smallerand still be usable or provide additionalbenefit. Moreover, such reductions mightincrease, rather than decrease, the cogni-tive load. For example, many new carsinclude a key fob for locking the doors andarming the alarm. If the fob were the sizeof a brick, few people would use it becauseit would be too large. Alternatively, if itwere the size of a penny, few people woulduse it owing to difficulty manipulating the

Figure 3. Sony’s Aibo robot dog.


buttons. In this case, the most suitable sizeand weight are inextricably tied to thedevice’s intended function.

EnergyThe challenge for any mobile system is

to reduce user involvement in managingits power consumption. Energy is a nec-essary resource for virtually all comput-ing systems, but any reference to itdetracts from a positive user experience.The degree to which an energy source dis-tracts the user depends on the intendedapplication, the hardware implementingthe application, and the energy source’scharacteristics.

Solutions to this problem fall into twoapproaches. The first is to reduce power con-sumption. Part of this approach consists ofdesigning energy-aware software that canidentify the hardware states that provide agiven service level and select those that aremost energy efficient. For instance, in sys-tems with a microprocessor whose energyconsumption is greater at high speeds, thesoftware can select the lowest speed possi-ble that still achieves the required task’s per-formance.23 The control software can alsomodify the quality of service it seeks todeliver.24 For instance, to save energy, thesoftware could reduce the frame rate, or size,of an MPEG movie, incrementally resultingin a corresponding loss of fidelity. Or, thesoftware could forward a voice utterance toa remote system for recognition rather thanexpending local energy on the task.24 Soft-ware systems are just beginning to addressthese issues concerning energy awareness.

The second approach is to find alterna-tive and improved energy sources.Although battery energy densities are pro-jected to increase approximately 10 per-cent annually for the next three years,25 theenergy storage costs of batteries will likelyremain significant. This will lead us toexplore other technologies for storing, and

even generating, energy. For example, MITresearchers have exploited the human bodyas an energy source by constructing sneak-ers that use flexible piezoelectric structuresto generate energy (see Figure 5).26,27 Sim-ilarly, solar radiation, thermal gradients,mechanical vibration, and even gravita-tional fields all represent potential powersources for a mobile device.15

Additionally, storage technologies underdevelopment promise much greater energydensities than those of conventional bat-teries. In the near term, one of the morepromising technologies is fuel cells, partic-ularly direct methanol fuel cells.28 Puremethanol fuel offers an energy densityroughly 40 times that of a Lithium-ionpolymer battery, but 70 to 90 percent ofthis chemical energy is lost in conversion toelectricity. In the longer term, technologiessuch as MIT’s MEMS (microelectro-mechanical systems)-based microturbineand associated micro electric generatormight provide highly compact energysources with significantly longer lifetimes.

An alternative to acquiring energy is totransmit energy to a mobile device, reduc-ing its need for an autonomous powersource. This technique is difficult to dosafely over a long range, but it is applica-ble to the field of passive electronic tagging.For example, Radio Frequency Identifica-tion29 tags are inductively powered by thetag reader, typically up to a maximum ofone meter, employing load modulation totransmit their data back to the interroga-tor. Such passive tags have unlimited life-times, are smaller, and cost less; however,unlike battery-powered (active) tags, theycan communicate only over a short range

and cannot autonomously signal their pres-ence. The ability to transmit energy, whencombined with robotics, gives us the capa-bility for mobile computation that canrecharge itself.

User interfacesRendering user interfaces invisible is fun-

damentally difficult owing to the tradeoffbetween size and weight and usability.Reducing the size and weight will make thedevice less visible but might decrease itsusability. The degree to which these twocomponents matter depends on the specificproperties of the interface and of the appli-cation for which it is being used.

For example, to open or lock a door, auser might use a remote control contain-ing one button that unlocks and opens thedoor when pressed once and locks it whenpressed twice in quick succession. Becausethe single button serves multiple purposes,users will likely find it more demandingthan the original interfaces (the door knoband door lock). Alternatively, the remotecontrol could have two buttons, one foropening and one for locking and unlockingthe door. At the expense of increased sizeand weight, this solution substitutes a spa-tial differentiation (two single-functionbuttons) for a temporal one (a multiple-function button). These two solutions rep-resent the chief user-interface tradeoff thatmakes good user interface design for smallubiquitous devices fundamentally hard.

Today’s most popular user interfacesinclude buttons, keyboards, mice, point-ers, LCD panels, touch screens, micro-phones, and speakers. These elements aredesigned for high-rate information flow,

Figure 4. The Itsy pocket computer exemplifies the trend that the size andweight of energy sources and I/O devicesis dominating the size and weight ofmobile computers.



each suited to a specific application class.For example, a keyboard and display seemideal for writing a book, but a microphoneand speaker would probably be better forcommunicating with another person. Forother applications, such as alerting a userthat he or she has received mail, these userinterfaces are unnecessarily complex.Other, more unobtrusive, mechanisms suchas the ambientROOM,30 explore how tocommunicate low-latency, low-importanceinformation via interfaces that require lit-tle direct attention. For example, variancesin a projected image’s intricacy could indi-cate if there is unread mail, thereby mak-ing information available to the user unob-trusively through a directed glance.

Although these interfaces are distinctlyseparate from our bodies, the natural direc-tion for disappearing interfaces is for thetwo to blend together. Several researchersare exploring the possibility of interpret-

ing information from our neurons to let uscontrol computers and other machines byjust thinking about doing so.31 Early workinvolving a monkey with neural implantshas demonstrated the ability to gather suf-ficient information to let a robot mimic themonkey’s arm movement.31 Similarly,neural implants can feed information intothe brain, removing the need for humans togather information through their senses.This approach’s potential is suggested byrecent research employing cochlearimplants to help those with hearing loss tocommunicate better and become moreaware of their surroundings.32 For exam-ple, such neural implants could bring infor-mation (such as “you’ve got mail”) to theuser’s attention by fooling the brain intothinking a subtle but noticeable imagehas been projected onto the retina.

These interfaces form the personal-inter-action version of proactive computing,

where the computer and the real world aretightly integrated. Nonetheless, directneural interfaces embody tremendous risks,not the least of which is loss of humanautonomy. Clearly, this area requires muchmore research, but if we can overcome thechallenges, such interfaces would go a longway toward reaching Weiser’s vision.

Future challengesWe will soon be able to include com-

puter hardware into virtually every man-ufactured product, and provide a wirelessinfrastructure to let these devices commu-nicate directly or indirectly. But what willthey communicate about and in what pro-tocol? How will a user or user’s applica-tion know what devices to use for whatpurpose? How will a user know wheninvisible devices are present, functioning,and not compromising privacy? Whatactions can a user take if the situation is

unacceptable? These questions suggest thathardware, software, user interaction, andapplications all have unresolved issues thatwe must address before ubiquitous com-puting will truly reach Weiser’s goal ofimproving, rather than further complicat-ing, our lives.

Unresolved hardware issuesWe have already discussed some of the

challenges for our hardware platforms.Specifically, we will need to continue tomanage our devices’ power requirements.Making devices ubiquitous can’t be coupledwith the need to change batteries, or evenrecharge them, when thousands of devicesare involved. The solutions presented ear-lier only partially solve the problems; someare highly experimental and might not bepractical for unforeseen reasons.

Microprocessors can also continue evolv-ing. The need will persist for minimal

devices that use the baseline minimumpower for the job or occupy the smallestpossible volume. Processors will continueto shrink while increasing in capability andcapacity. However, new applications willdemand ever-greater processing capabilities,not the least of which will be those of com-munication and security. For example,implementing public key encryption with atypical microcontroller instruction set leadsto large code size, high power requirements,and slow performance.33 The challenge willbe to include appropriate primitives thatmake these operations more efficient in allthese dimensions while permitting the evo-lution of security algorithms.

As we crowd more and more frequencieswith wireless communications, we will needto make our radio systems adaptive. Deviceswill need to adjust their bandwidth require-ments on the basis of what other devices arein the radio neighborhood. Such a necessityhighlights the critical problem of system evo-lution. How will the development anddeployment of new devices affect the deviceswe already have deployed? Flexibility willbe needed, not only in terms of softwareupdates to adjust how a device is used butalso in terms of wireless communication.Software radios are promising in this sec-ond dimension; they let a device change howit uses the spectrum to be compatible withits neighbors.34

Disappearing softwareTo make hardware disappear, we also

need to make software disappear. Today’ssoftware is too monolithic and stovepiped,and is written with many assumptionsabout hardware and software resources.The ubiquitous computing environmentthat we must create will be much moredynamic. Devices, objects, and people willbe constantly moving around, creating an ever-changing set of resources—differentuser input/output interfaces, displays, andwindowing systems—that will be availableto our applications. Also, some devicesmight lose their ability to communicatewith others owing to interference or envi-ronmental conditions. How will we con-struct applications to operate in such envi-ronments?

A key element of ubiquitous computing

applications is knowing the precise

spatial–temporal relationships between

people and objects.


To do this, we must even further decou-ple our applications into small pieces ofcode, spread across ubiquitous hardware,that can come together as needed andexpect to have connections constantlyenabled and disabled. We must create datainterchange formats where the data notonly is self-describing but also can find itsway from the device that created it to itsdestination as autonomously and securelyas possible.35 We must develop mecha-nisms for devices to advertise their capa-bilities so that applications, whether in theinfrastructure or on portable devices, canbecome aware of and select from thepanorama of available resources.

Discovery services36,37 (for example, Uni-versal Plug and Play) to some extent alreadyenable such cyber foraging.38 These first-generation systems require highly capabledevices that can download code and enjoystable and relatively high-bandwidth con-nections. This approach will not scale tothe myriad sensing and processing devicesthat will surround us. They will employminimal computational elements and existin small communication cells with highlyvariable communication and processingproperties. This scenario requires that mul-tiple devices replace a single device. Adap-tation at this level is another challenge wemust tackle to achieve long-lived systemsthat can evolve gracefully.

Interaction designAlthough new technologies are creating

new ways of interacting with our compu-tations, we could end up trading the prob-lems of one user interface for those ofanother. For example, wireless technolo-gies permit a device we are carrying tointeract with a public display. But whathappens when multiple potential users aresurrounding a display? Not only must the

software handle multiple users, but alsothe hardware must be able to accuratelyidentify and differentiate between the mul-tiple users. Addressing these issues usingwhat we’ve learned from the desktopmetaphor seems less promising becausethat interface has not been developed oroptimized for casual use by multiple users.

With computing capability in everyobject, users will want to take advantageof the devices they encounter throughouttheir day without worrying about owner-ship or security at every step. Consider howwe use paper and pencil: we can easily jotdown notes and later identify the authorbased solely on the handwriting. The note-taking process directly incorporates thisprocess; there is no explicit authenticationmechanism. Similarly, physical controlguarantees privacy: we put the paper in ourpocket and can hide it from others’ eyes.What metaphors will we have for the elec-tronic paper that can communicate withother devices? Without some kind of phys-ical icon, how do we seamlessly controlsharing content with other people?

A key element of ubiquitous computingapplications is knowing the precise spatial–temporal relationships between people andobjects. Such knowledge succinctly helpsspecify intent, an integral component ofuser interfaces. But the resolution of loca-tion systems needs to improve dramati-cally.39 Current commercial indoor loca-tion systems are either too coarse,operating primarily at the room level7 (forexample, the Versus Information System),or require prohibitively expensive infra-

structure,40,41 or are based on proximity(for example, electronic tags). We desper-ately need tags that can be located withina few millimeters, are cheap to create (forexample, by printing), and are completelypassive. Ideally, we would also want theability to detect that two tagged objects arephysically touching rather than just in closeproximity.42,43 Coming up with the tech-nologies that provide these capabilities,even if initially imperfectly, is another keychallenge. For more on this topic, see“Connecting the Physical World with Per-vasive Networks,” in this issue.

ApplicationsCurrently, most applications are based

on ownership of relatively large, multi-purpose hardware devices with only themost limited interaction with the physicalworld in which people live. We need to dis-cover and enable those most compellingapplications that will let us deploy the firsttruly ubiquitous systems. Yet, as we havealready discussed, the most difficult partof this will be to give these systems an evo-lutionary path, in contrast to today’sapproach of completely reengineering allthe component devices. Software engi-neering will take center stage in this effort.Clearly, we need a new set of abstractionsto make writing such applications possi-ble—abstractions that support softwaredeployment in separate modules on thou-sands of devices, cyber foraging, and hard-ware sharing. This is a different world ofsoftware development than we are accus-tomed to.

Figure 5. Energy-scavenging shoes usingpiezoelectric material, developed at MIT’sMedia Laboratory. With these shoes, normal walking motion can generate sufficient energy to broadcast an ID every three to five steps.27 Photo courtesy of the MIT Media Laboratory.



M any hardware componentsnecessary to build ubiqui-tous computing systemsare now available. Key

improvements since Weiser’s original visionof ubiquitous computing include wirelessnetworks; high-performance and low-power processors; high-quality displays;and high-capacity, low-power storagedevices. The progress toward building soft-ware systems to orchestrate these compo-nents has been less dramatic, with manyunresolved issues relating to user interfaces,security, privacy, and managing complex-ity. Consequently, the hardware compo-nents for many applications are reachingthe point where a user is less likely to bedistracted by the medium than by the inter-action with the controlling software. Andas a result, at this level, the hardware isbeginning to become invisible.

Of course, hardware could become tooinvisible. At some point, you might needto know what is real and what is simulated,and have some handle on an interface’slocation and components. However, tech-nological advances clearly will continue tobring us new hardware components that,if we wish, will function more invisiblythan before, until every aspect of an inter-face crosses a threshold that no longer hin-ders our senses. We will begin to seethrough it, just as we see through the ink ofprinted text and focus on the informationit contains. At that point, by definition, thehardware will have disappeared. Whenthat day comes, computer hardware willlikely be mediating in every aspect of ourdaily activities, and Weiser’s vision will bealmost complete.

ACKNOWLEDGMENTS

The thoughts and views expressed in this articlewere not formed in isolation. We thank the manycolleagues who have worked alongside us on ubi-quitous computing projects and, in doing so, haveshaped our views in this exciting research area. Particularly, we thank Sunny Consolvo, Gunner Dan-neels, and the IEEE Pervasive Computing editorialboard and supporting reviewers for their comments;Jim Kardach, Graham Kirby, Muthu Kumar, MuraliSundar, and Paul Wright for the presentation of thepersonal server prototype; and Bob Mayo and BillHamburgen for their contributions to our discussion

on the hard problems associated with ubiquitouscomputing. And a special thank you to Mark Weiser,in memory, for his visionary ideas in the early 1990s.

REFERENCES

1. K. Fishkin et al., “Embodied User Interfacesfor Really Direct Manipulation,” Comm.ACM, vol. 43, no. 9, Sept. 2000, pp. 75–80.

2. M. Weiser and J. Seely-Brown, “The Com-ing Age of Calm Technology,” Beyond Cal-culation: The Next Fifty Years of Comput-ing, P.J. Denning and R.M. Metcalfe, eds.,Copernicus, Heidelberg, Germany, 1998.

3. D.A. Norman, The Invisible Computer,MIT Press, Cambridge, Mass., 1998.

4. R. Want et al., “An Overview of the ParcTabUbiquitous Computing Experiment,” IEEEPersonal Comm., vol. 2, no. 6, Dec. 1995,pp. 28–43.

5. C. Kantarjiev et al., “Experiences with X ina Wireless Environment,” Proc. UsenixSymp. Mobile & Location-IndependentComputing, Usenix Assoc., Berkeley, Calif.,1993, pp. 117–128.

6. S. Elrod et al., “Liveboard: A Large Interac-tive Display Supporting Group Meetings,Presentations, and Collaborations,” Proc.1992 Conf. Human Factors in ComputingSystems (CHI 92), ACM Press, New York,1992, pp. 599–607.

7. R. Want et al., “The Active Badge LocationSystem,” ACM Trans. Information Systems,vol. 10, no. 1, Jan. 1992, pp. 91–102.

8. T. Truman et al., “The InfoPad MultimediaTerminal: A Portable Device for WirelessInformation Access,” IEEE Trans. Com-puters, vol. 47, no. 10, Oct. 1998, pp.1073–1087.

9. B. Schilit, N. Adams, and R. Want, “Con-text-Aware Computing Applications,” Proc.Workshop Mobile Computer Systems andApplications, IEEE CS Press, Los Alamitos,Calif., 1994, pp. 85–90.

10.G. Moore, “Cramming More Componentsonto Integrated Circuits,” Electronics, vol.38, no. 8, 19 Apr. 1965, pp. 114–117.

11. J.W. Toigo, “Avoiding a Data Crunch,” Sci-entific American, vol. 282, no. 5, May 2000,pp. 58–74.

12.A. Fox et al., “Integrating InformationAppliances into an Interactive Work-space,” IEEE Computer Graphics & Ap-plications, vol. 20, no. 3, May/June 2000,pp. 54–65.

13.Embedded Microcomponent Forecast and

Trends through 2004, MPUs, MCUs, DSPsand Cores, Gartner Group, Stamford,Conn., 22 Jan. 2001.

14.R. Want and B. Schilit, “Guest Editors’Introduction: Expanding the Horizons ofLocation-Aware Computing,” Computer,vol. 34, no. 8, Aug. 2001, pp. 31–34.

15.A. Chandrakasan et al., “Design Consider-ations for Distributed Microsensor Sys-tems,” Proc. 1999 IEEE Custom IntegratedCircuits Conf., IEEE Press, Piscataway, N.J.,1999, pp. 279–286.

16. J. Hill et al., System Architecture Directionsfor Network Sensors, Proc. 9th Int’l Conf.Architectural Support for ProgrammingLanguages and Operating Systems (ASPLOS2000), ACM Press, New York, 2000, pp.93–104.

17. J.M. Kahn, R.H. Katz, and K.S.J. Pister,“Next Century Challenges: Mobile Net-working for ‘Smart Dust,’” Proc. ACM/IEEE Int’l Conf. Mobile Computing andNetworking (MobiCom 99), ACM Press,New York, 1999, pp. 271–278.

18. J.M. Rabaey, “Wireless beyond the ThirdGeneration—Facing the Energy Challenge,”Proc. 2001 Int’l Symp. Low Power Elec-tronics and Design (ISLPED 01), ACMPress, New York, 2001, pp. 1–3.

19. J. Crowley, J. Coutaz, and F. Berard, “Per-ceptual User Interfaces: Things That See,”Comm. ACM, vol. 43, no. 3, Mar. 2000, pp.54–64.

20.H. Ishii and B. Ullmer, “Tangible Bits:Towards Seamless Interfaces between Peo-ple, Bits and Atoms,” Proc. 1997 Conf.Human Factors in Computing Systems (CHI97), ACM Press, New York, 1997, pp.234–241.

21.D. Tennenhouse, “Proactive Computing,”Comm. ACM, vol. 43, no. 5, May 2000, p.43.

22.W.R. Hamburgen et al., “Itsy: Stretching theBounds of Mobile Computing,” Computer,vol. 34, no. 4, Apr. 2001, pp. 28–36.

23.M. Weiser et al., “Scheduling for ReducedCPU Energy,” Proc. 1st Symp. OperatingSystems Design and Implementation,Usenix, Berkeley, Calif., 1994, pp. 13–23.

24. J. Flinn, Extending Mobile Computer Bat-tery Life through Energy-Aware Adaptation,PhD dissertation, tech. report CMU-CS-01-171, Computer Science Dept., Carnegie Mel-lon Univ., Pittsburgh, 2001.

25.Y. Kim, “Capacity Enhancement Technol-ogy of Samsung Soft Case Li-ion Batteries,”Proc. Power 2001: 9th Ann. Int’l Conf.Power Requirements for Mobile Comput-


ing, Wireless Electronic Devices & Busi-ness/Consumer Applications, 2001.

26.T. Starner, “Human-Powered WearableComputing,” IBM Systems J., vol. 35, nos.3–4, 1996, pp. 618–629.

27.N. Shenck and J. Paradiso, “Energy Scav-enging with Shoe-Mounted Piezoelectrics,”IEEE Micro, vol. 21, no. 3, May/June 2001,pp. 30–42.

28.S. Thomas and M. Zalbowitz, Fuel Cells—Green Power, Los Alamos Nat’l Lab, LosAlamos, N.M., 1999; www.lanl.gov/energy/est/transportation/trans/pdfs/fuelcells/fc.pdf.

29.K. Finkenzeller, The Radio Frequency Iden-tification (RFID) Handbook, John Wiley &Sons, New York, 1999.

30.H. Ishii et. al., “AmbientROOM: Integrat-ing Ambient Media with ArchitecturalSpace,” Proc. 1998 Conf. Human Factorsin Computing Systems (CHI 98), ACMPress, New York, 1998, pp. 173–174.

31.A. Regalado, “Miguel Nicolelis: Brain-Machine Interfaces,” Technology Rev.,Jan./Feb. 2001, pp. 98–100.

32.M. Nicolelis, “Actions from Thoughts,”Nature, vol. 409, no. 6818, 18 Jan. 2001,pp. 403–407.

33. A. Perrig, “SPINS: Security Protocols forSensor Networks,” Proc. 7th Ann. ACMInt’l Conf. Mobile Computing and Net-working (MobiCom 2001), ACM Press,New York, 2001, pp. 189–199.

34.V. Bose, D. Wetherall, and J. Guttag, “NextCentury Challenges: RadioActive Net-works,” Proc. ACM Mobile Computing andNetworking (Mobicom), ACM Press, NewYork, 1999, pp. 242–248.

35. M. Esler et al., “Next Century Challenges:Data-Centric Networking for Invisible Com-puting—The Portolano Project at the Uni-versity of Washington,” Proc. MobiCom1999, ACM Press, New York, 1999, pp.256–262.

36.W. Adjie-Winoto et al., “The Design andImplementation of an Intentional NamingSystem,” Operating Systems Rev., vol. 34,no. 5, Dec. 1999, pp. 186–201.

37. J. Waldo, Jini Architecture Overview, tech.report, Sun Microsystems, Palo Alto, Calif.,Jan. 1999.

38.M. Satyanarayanan, “Pervasive Computing:Vision and Challenges,” IEEE PersonalComm., vol. 8, no. 4, Aug. 2001, pp. 10–17.

39. J. Hightower and G. Borriello, “LocationSystems for Ubiquitous Computing,” Com-

puter, vol. 34, no. 8, Aug. 2001, pp. 57–66.

40.M. Addlesee et al., “Implementing a SentientComputing System,” Computer, vol. 34, no.8, Aug. 2001, pp. 50–56.

41.A. Ward, A. Jones, and A. Hopper, “A NewLocation Technique for the Active Office,”IEEE Personal Comm., vol. 4, no. 5, Oct.1997, pp. 42–47.

42.K. Partridge et al., “Fast Intrabody Signal-ing: Empirical Measurements of IntrabodyCommunication Performance under Var-ied Physical Configurations,” Proc. 14thAnn. ACM Symp. User Interface Software

and Technology (UIST 2001), ACM Press,New York, 2001, pp. 183–190.

43.B. Vigoda and N. Gershenfeld, “TouchTags:Using Touch to Retrieve Information Storedin a Physical Object,” Extended Abstractsof Computer-Human Interaction 1999,ACM Press, New York, 1999, pp. 264–265;www.media.mit.edu/physics/publications/papers/99.01.touchtag.rtf.

For more information on this or any other com-puting topic, please visit our Digital Library athttp://computer.org/publications/dlib.

the AUTHORS

Roy Want is a principal engineer at Intel Research. His interests include ubiquitouscomputing wireless protocols, hardware design, embedded systems, distributedsystems, and automatic identification and microelectromechanical systems. While atOlivetti Research, he developed the Active Badge, a system for automatically locatingpeople in a building. As part of Xerox PARC’s Ubiquitous Computing program, he ledthe ParcTab project, one of the first context-aware computer systems. At PARC, Wantalso managed the Embedded Systems group. He received his BA and PhD in computerscience from Churchill College, Cambridge University. Contact him at Intel Corp., 2200

Mission College Blvd., Santa Clara, CA 95052; [email protected]; www.ubicomp.com/want.

Trevor Pering is a research scientist at Intel Research. His research interests includemany aspects of mobile and ubiquitous computing, including usage models, powermanagement, novel form factors, and software infrastructure. He received his PhDin electrical engineering from the University of California, Berkeley, with a focus onoperating system power management. He is a member of the ACM. Contact him [email protected].

Gaetano Borriello is a faculty member of the University of Washington’s Departmentof Computer Science and Engineering. He is on a two-year leave to establish a newIntel research center adjacent to the University of Washington campus. His researchinterests are in the design, development, and deployment of computing systems withparticular emphasis on mobile and ubiquitous devices and their application. His mostrecent research accomplishments include the development of the Chinook design system for heterogeneous distributed embedded processors. He received his BS in electrical engineering from the Polytechnic Institute of New York, his MS in elec-

trical engineering from Stanford University, and his PhD in computer science from the University of California, Berkeley. He received an NSF Presidential Young Investigator Award in 1998 and a Universityof Washington Distinguished Teaching Award in 1995. Contact him at the Dept. of Computer Science &Eng., Univ. of Washington, Box 352350, Seattle, WA 98195-2350; [email protected].

Keith I. Farkas is a senior member of the research staff of Compaq Computer’sWestern Research Lab. His research interests include microprocessor design and software and hardware techniques for managing and optimizing computer systemenergy consumption. He received his PhD from the University of Toronto. He is amember of the IEEE and ACM. Contact him at [email protected].

disappearing hardware - cs.cmu.edujasonh/courses/ubicomp-sp2007/papers/04-disap… · however,...

Documents