Beyond prototypes: challenges in deploying ubiquitous systems

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

Beyond Prototypes:Challenges in DeployingUbiquitous Systems

Despite being written over 10 yearsago, many aspects of Mark Weiser’svision of ubiquitous computingappear as futuristic today as they didin 1991.1 The reasons for this appar-

ent lack of progress are manifold, and other articlesin this issue explore remaining technical problems inspecific areas of ubiquitous computing research. Wefocus on the technical and sociological challengesof creating systems that are ubiquitous in the

more general sense of the word—systems that extend beyond merelaboratory prototypes. Such sys-tems are deployed to the extentthat they become an integratedpart of our everyday lives. Onlywhen we have achieved this degreeof pervasiveness will Weiser’s

vision become reality. Unfortunately, we are still manyyears from creating such systems.

Here, we discuss significant research challengesthat have yet to be addressed. Central to docu-menting these challenges is recognizing the contextwithin which we are operating. So, we first describethe technical and social changes of the 1990s thatdirectly affected ubiquitous computing.

The decade of ubiquitous informationand communication

Since Mark Weiser wrote his seminal article, tech-nology has advanced along many dimensions. Inaddition to well-known developments in capabili-

ties such as processing power and storage in portabledevices, other significant—but perhaps less obvi-ous—developments have occurred. These develop-ments are likely to affect our ability to deploy ubiq-uitous computing systems. They include newtechnologies—such as the Global Positioning Sys-tem (GPS), smart cards, and radio frequency iden-tification (RFID) tags—and social developmentssuch as the increasingly widespread acceptance ofvideo surveillance in public places. However, thedecade’s most striking developments (with respectto ubiquitous computing) have undoubtedly beenthe emergence of the Web as a global informationand service resource and the widespread adoption ofdigital mobile telephony, letting users experiencenearly ubiquitous wireless communications.

The World Wide WebThe Web’s emergence has fundamentally changed

the way many people interact with computers. It hasalso created a culture that is substantially moreamenable to the deployment of ubiquitous com-puting environments than that which existed whenWeiser first articulated his vision.

Most obviously, the Web has created a nearlyubiquitous information and communications infra-structure. We can now access a huge wealth ofknowledge and services from almost any computer,including low-power mobile devices such as smartphones and PDAs. However, the Web has had other,more subtle effects on our culture.

First, the increased use of computers as portals to

The authors discuss the problems in creating the next generation ofwidely deployed ubiquitous computing systems and articulate currenttechnical and sociological challenges to inspire researchers in the field.

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

Nigel DaviesLancaster University and University of Arizona

Hans-Werner GellersenLancaster University

the Web has reduced many users’ sense ofattachment to a single computing device.In his article on calm technology,2 Weiserdrew parallels between personal comput-ers and automobiles, saying that both arespecial, relatively expensive items withwhich users form relationships. One exam-ple of this relationship between users andtheir personal computers is the commonpractice of anthropomorphizing a com-puter—for example, by naming it. Whilesuch relationships are likely to persist forsome period of time, many users now relatenot to their computer but rather to theirpoint of presence within the digitalworld—typically, their homepage, portal,or email service. So, for users who exten-sively use Web services and information,the computer that they use to access thesethings has become largely irrelevant. Manyusers commonly access the same point indigital space from several different devices(office or home PC, cell phone, PDA, andso forth) throughout the course of a typi-cal day. Consequently, for most users, com-puters themselves are becoming increas-ingly unimportant—what matters is theview a particular machine provides of thedigital world. In this sense, we are well onthe way to computers “disappearing” andusers being free to focus beyond them.

The Web has also accelerated the devel-opment of social and legal constructs fordealing with computationally rich envi-ronments. In particular, Web users havehad to contend with significant challengesto their privacy, primarily in the form oflogging technologies that can generate sub-stantial amounts of detail about a givenuser’s Web activities. This has raised sig-nificant concerns among many Web users.Commercial organizations, legislators, andprivacy groups are struggling to come toterms with new technology’s implicationsfor an individual’s privacy.

Mobile communications for the massesIn addition to creating a ubiquitous

information infrastructure, the last decadealso witnessed the widespread deploymentand adoption of digital mobile communi-cations, primarily in the form of the globalsystem for mobile communications.3 The

Internet’s growth, particularly in the US,sometimes overshadows the huge impactthat technologies such as GSM have had onmany consumers worldwide. There are cur-rently an estimated 800 million subscribersto mobile phone services, of which 65 per-cent use GSM-based networks (see www.gsmworld.com/news/statistics/index.shtml).During September 2001, these users sentover 23 billion SMS (short message service)messages—part of a steep upward trend asusage heads toward one billion messagesper day (see www.gsmworld.com/news/statistics/index.shtml). Furthermore, mod-ern handsets offer far more capabilities thanearly ubiquitous computing devices such asParcTabs in roughly the same form factor.A typical phone today might include simplePDA applications such as a calendar andto-do lists, games, text-messaging facilities,voice communications (of course), Webaccess, and even simple voice recognition.

Modern mobile phones have anotherimportant property that Weiser associatedwith ubiquitous computing devices: manyusers view them as a commodity that theycan find and use anywhere they travel. Animportant factor in forming this view is theprice. For most users, handsets are essen-tially free and are replaced relatively fre-quently. Weiser suggested that users canuse any ParcTab as if it were their own.Separating the handset from the subscriberidentity module card found in GSM sys-tems approximates this model of opera-tion. By inserting their SIM card into ahandset, subscribers can automatically usethe handset, placing and receiving calls asif it were their own phone. Of course, thisrepresents only a partial solution becauseusers typically own only one SIM card andhence can’t use multiple devices simulta-neously. Moreover, users must consciouslyinsert their SIM card into a handset—theycan’t just pick up and use any handset. Sim-ilarly, although phones are perceived ascheap, they are not usually left lyingaround for casual use in quite the wayWeiser described. Despite these shortcom-ings, using SIM cards demonstrates onceagain that, for many users, the end-systemis becoming less important than the accessit provides to the digital world.

Why the whole is more than thesum of its parts

Weiser’s vision still seems like science fic-tion, primarily due to a lack of integrationin existing systems. Consider the scenarioof Sal’s world that Weiser outlined in hisarticle (see page 24 in this issue); clearly,the technologies required are either alreadydeployed or could be deployed relativelytrivially.

For example, to realize the foreview mir-ror Weiser describes would require

• Equipping Sal’s car with a satellite nav-igation and information system

• Including some mechanism for detect-ing the availability of parking spacesnear her office building

• Including a detection or notificationmechanism that can highlight new shopsopening in her area

These things are already deployed in manyenvironments. Satellite navigation andinformation systems are standard in manyhigh-end cars; several different means existfor detecting parking availability (for exam-ple, we could tap into the parking lot’svideo surveillance camera and run an algo-rithm for detecting spaces); and we canprobably already identify new shops usinga neighborhood Web page or store guide.However, analyzing but a single scenario inSal’s world uncovers a host of issues asso-ciated with deploying ubiquitous comput-ing systems. The components from whichto construct such systems might already beavailable, but they are typically conceivedand operated independently, in the contextof their own restricted view of the world.

In the following discussion, we assumethat the “simple” problems associated withlarge-scale ad hoc software integration—such as naming, interface specification,fault tolerance, and configuration man-agement—have all been solved. We focuson the problems of creating integratedubiquitous computing systems.

Technical challengesThe first difficulty for anyone attempt-

ing to realize Weiser’s vision of the foreviewsystem is that it clearly has detailed knowl-

JANUARY–MARCH 2002 PERVASIVEcomputing 27

edge of Sal’s current task. In particular, itknows that Sal is driving to work and hencewill be interested in viewing parking spacesat her office. Assuming that Sal has notexplicitly told the system about her currenttask, it must deduce this information, per-haps based on the time, day, or directionthe car is traveling. Clearly, the possibilityof making erroneous deductions is high: theproblems associated with modern officepackages that attempt to predict user inten-tions illustrate the task’s difficulty, evenwhen the domain is extremely restricted.

Assuming that the system has deter-mined that Sal is interested in parkingavailability at her office, the next challengeis to make the correct associations betweenthe various components involved in pro-viding this information. This process is rel-atively easy for humans but extremely dif-ficult in software. For example, we quicklydetermined that we might locate availableparking using surveillance cameras, buthow could this ability to create new asso-ciations between components be realizedin a computer system? Moreover, suppose

that there are many parking lots near theoffice, and many cameras in each lot. Whatcriteria does the foreview system use todecide which parking lot to investigate(closest, cheapest, or most frequently used),and how does it determine which camerasapply to the task at hand (especially if thecameras are not static)?

Once the foreview system has found avideo image of a parking lot that appearsto contain a space, the next challenge is totranslate this information into drivingdirections or a map. This would requiredetermining the parking space’s precisegeographical coordinates—something thatwould probably imply knowing the posi-tioning of the parking lot’s cameras. Fur-thermore, if the cameras describe their field

of view using a different model of locationthan that used in the car navigation system(such as a symbolic model versus a geo-graphic model), the system must addressthe problem of mapping between differentlocation models. This area has seen somepromising research,4 but the problems areby no means solved.

Social and legal issuesEven if we assume that realizing the fore-

view application is technically feasible,numerous problems would still inhibit thesystem’s deployment. For example, sur-veillance data can obviously contain sen-sitive information such as the identity ofpeople at the scene, either directly throughrecognition or indirectly through vehicleidentification. Consequently, data protec-tion legislation applies (in European coun-tries, protection is decided on the basis ofDirective 95/46/ec5). So, the video surveil-lance component could only deliver imagesif it sought explicit consent from the peo-ple who might be identified—which clearlyis technically and socially infeasible.6

The obvious alternative would be for thevideo surveillance component to returnspecific information on parking availabil-ity—removing the possibility of other com-ponents accessing sensitive information butreducing the component’s general useful-ness in a ubiquitous computing environ-ment. However, designing such compo-nents would still require care to avoidpersonal data spills into public services. Forexample, a service that reports reservedspaces as being possible parking spaceswould provide implicit information on themovement of those spaces’ owners.

The reduction in generality also has sig-nificant consequences for ubiquitous com-puting system designers. For example, oneoften-cited application of a ubiquitous com-

puting environment is to help users locateobjects such as missing wallets or keys. Itis obvious to a designer how the compo-nents that were used to find a parking spacecould also help locate a user’s missing itemsif the user thinks he might have lost themwhen he parked his car. In this case, the ben-efit of a ubiquitous system operating as anintegrated whole as opposed to a set of sep-arate components is clear. However, suchgeneralization would be impossible if thesurveillance component offered only aninterface to locate parking spaces. Ofcourse, it is important to note that it is notjust infrastructure components thatrequire careful design to protect user pri-vacy. For example, if the foreview system’sstore-finder feature were implemented bymaking requests to a Web server providinglocal information, the foreview systemwould be providing detailed informationabout Sal’s movements to the server.

Economic concernsFinally, we observe that while the park-

ing-space application delivers value toSal—and she might well be prepared topay for such a service—the cost of pro-viding the application is distributedamong numerous components. Such com-ponents might include the video surveil-lance system and its operators and thecommunications provider that enables theinformation to reach Sal’s car. How dothese service providers recover their costs?Can one parking lot offer the informationfor free, hoping to recover the coststhrough increased use? How would weensure fair competition between neigh-boring parking lots? How would we billSal, and how much would she pay for theinformation? Developing effective busi-ness models for ubiquitous computingsystems will clearly be crucial to their suc-cess, yet at best, system designers poorlyunderstand this issue.

Experiences in deployingubiquitous systems

Many systems highlight developmenttrends that we generally associate with thedrive toward ubiquitous computing, suchas the provision of continuous service, con-

28 PERVASIVEcomputing http://computer.org/pervasive


Developing effective business models for

ubiquitous computing systems will clearly be

crucial to their success, yet at best, system

designers poorly understand this issue.

textualized use, and ad hoc collaborationof components. Here, we review some ofthese systems—ones that, in one sense oranother, have been deployed in environ-ments beyond strictly controlled lab set-tings. First, we discuss the pioneering workat Xerox Parc and at AT&T Labs in Cam-bridge,7 both of which emphasize deploy-ing infrastructures to facilitate ubiquitouscomputing applications. In contrast, wethen discuss Guide and Cooltown,8 whichfocus more on creating actual ubiquitouscomputing applications and deployingthem in large user communities (a similarproject is Classroom 2000;9 see “TheHuman Experience” in this issue for moreinformation on that project). Finally, weexamine the MediaCup10 project at Karl-sruhe, which explores the issue of deploy-ment in everyday artifacts.

Xerox Parc’s ubiquitous computingexperiment

Weiser’s articles on ubiquitous comput-ing usually presented not just ideas butdescriptions of systems designed, devel-oped, and deployed as part of Xerox Parc’sUbicomp project. Obviously, these systemswere bound by the technology of the early1990s. Because the Ubicomp project couldnot push computing and networking intoeveryday objects to the extent envisioned,it instead approached the diversity ofdevices in a future world by designing andimplementing computers representing dif-ferent scales—from inch- through yard-sized devices.

Devices with characteristics similar tothose prototyped have since become com-mon: clear parallels exist between inch-scale ParcTabs and phone and PDA prod-ucts; foot-scale Pads1 and reading appli-ances and tablet PCs; and the yard-scaleLiveboard and interactive white boardproducts. However, 10 years ago, devel-oping ParcTabs, Pads, and Liveboardsrequired tremendous effort, becauseresearchers had to construct the entire sys-tem (network, communication, user inter-face, and applications) from scratch. Notsurprisingly, the technical challenges ofdeveloping the individual devices meantthat the overall project fell short of address-

ing integration issues, and developmentssuch as the ParcTab and the Liveboard,which were tremendously influential intheir own right, remained mostly separate.

Among the systems explored in the Ubi-comp project, the ParcTab particularly influ-enced ubiquitous computing researchbecause it introduced two major featuresthat represent a departure from desktopcomputing: continuous service and contex-tualized use. The device’s lightweight designenabled continuous service in combinationwith permanent connectivity to remoteapplications, using a wireless infrared net-work. The network’s cell-based nature sup-ported localization of a ParcTab device,enabling context-aware applications.11 Theentire ParcTab system was deployed in alarge office environment, with first 25 andlater 50 infrared cells, to study the use andevolution of applications with a communityof more than 40 people.12 The experiment’sscale facilitated an understanding of designand user issues that demonstrations or sim-ulations could not have achieved.

Of the many ParcTab observationsreported, we relate just one here to under-line the value of deploying working sys-tems. Researchers identified email accessearly on as a particularly compellingParcTab application. It was instantlyappealing in demonstrations and popularwith users when first introduced to the sys-tem. However, in the deployed system,wherever a ParcTab infrared cell existed,there was also a powerful workstation thatmade a much better platform for readingemail. The only exceptions to this rule wereconference rooms; in these, all the confer-ence participants (typically about 20 peo-ple) would try to use their ParcTabs simul-taneously over the limited bandwidth ofthe shared cell, making them too slow forcommunication.13 The lesson learned isthat understanding a new artifacts utilityrequires serious deployment; experiencegathered in demonstrations and lab trialscan be misleading.

From the Active Badge to sentientcomputing

The Active Badge project, conducted inthe early 1990s at the Olivetti Research Lab

(now AT&T Laboratories) in Cambridge,England, heavily affected ubiquitous com-puting research because it implemented thefirst significant indoor positioning system.The system used wearable badges to emitbeacons and networked sensors to detectbadges and locate their wearers.14,15 Inter-estingly, its development was not driven bya long-term vision such as ubiquitous com-puting but by a very practical concern: howto locate people in an office complex forthe purpose of routing phone calls.Addressing a clearly perceived problem, theActive Badge brought a new level of utilityto users that facilitated widespread deploy-ment (various universities and research labsthroughout Europe and the US deployedover 1,500 badges and 2,000 sensors, withthe largest single system at Cambridge Uni-versity having 200 badges and 300 sensorsin daily use). Once deployed on this scale,the Active Badge system helped new appli-cations emerge, including compellingexamples of the departure from personaldesktop-bound computing toward ubiqui-tous computing.

One such example was Teleporting,which supported mobile users by trans-porting the interface of their remote appli-cations to the nearest available terminal.16

The teleporting concept supports the visionthat services become directly attached topeople rather than to a particular machine,rendering the service ubiquitous and mak-ing access to any particular machine lessrelevant. This links back to the effect ubiq-uitous Web access has on the tie betweenthe user and personal computer. For tele-porting, X Windows provided the under-lying machine independence, but the prin-ciples remain the same.

Teleporting and other early experimentswith context-aware systems used room-scale information generated by ActiveBadges, but it soon became obvious thatmany applications require finer-grained 3Dlocation and orientation information. Thiscorresponds to Roy Want and his col-leagues’ analysis of in-building location asa key problem in ubiquitous computing.17

The challenge is to achieve a degree of spa-tial resolution much closer to the humanperception of space, so we can treat the


physical environment’s current state ascommon ground between computers andtheir users.

In work that has evolved from their ear-lier location-aware systems, researchers atAT&T Laboratories Cambridge havedeveloped the Active Bat, an ultrasoniclocation system that achieves fine-grained3D positioning (see Figure 1).18 The Bat isdeployed throughout the lab’s three-flooroffice building, with all 50 staff using itcontinuously. With 95 percent of the sen-sor readings being correct to within 3 cmin 3D, the Bat is by far the most accuratedeployed indoor positioning system. TheBat system serves as infrastructure for fur-ther research into sentient computing,exploring applications that become possi-ble when computers have a fine-grainedmodel of the physical environment inwhich they are used.7 The sentient com-puting system maintains a software modelof objects in the real world that containsup-to-date information about location andstate. The system further supports a pro-gramming model in which developers canspecify spaces to monitor and operationsto execute based on spatial relationships.

Although the work of the Cambridge-based lab continues to be at the forefrontof enabling location as a resource for ubiq-uitous computing applications, it has gen-erated public controversy. In particular,introducing the Active Badge triggeredOrwellian interpretations of future workenvironments, and tracking people con-

tinues to be a sensitive topic. This reactionto location technologies is an importantreminder that understanding user concernswill be critical to obtaining sufficient accep-tance to ensure successful deployment ofubiquitous computing systems.

Lancaster’s Guide systemThe pioneering work at Parc and at

AT&T Labs required developing home-grown infrastructures; hence, the projectswere primarily bound to research environ-ments. However, the increasing availabil-ity of readily deployable mobile computinginfrastructures and affordable outdoorlocation technology components (such asGPS) led to a series of systems that investi-gated ubiquitous computing beyond theoffice setting. In 1996, systems began toemerge that studied mobility, continuousservice provision, and contextualized usein application domains such as tour guidesand navigation. Examples include Cyber-guide,19 the Touring Machine,20 and Lan-caster’s Guide.21 The Lancaster Guide proj-ect was a comparatively small and focusedresearch effort, but it involved a large fieldtrial with real users, providing insight intoissues concerning deploying ubiquitouscomputing systems to the wider public.22

The research team at Lancaster designedthe Guide system to provide visitors to thecity of Lancaster with the type of infor-mation normally found in a tour guide, butcontextualized on the basis of the visitor’sinterest and movement around the city.

They based the system on a distributed,dynamic information model, augmentingstandard hypertext with geographic infor-mation and navigation elements, as well ascontext-sensitive active components. Usersinteract with the system using tablet PCsas end-systems, connected to informationservers through an 802.11 network de-ployed around the historic city’s majorattractions. This network covers most butnot all of the inner city. Its cell structurealso provides end-systems with coarse-grained (approximately 200 m accuracy)location information based on their cur-rent cell identifier. Despite the coarse gran-ularity, integrating location nonethelesssupports an interesting information navi-gation model. Users effectively navigate theinformation space both explicitly throughthe Guide user interface and implicitlythrough their movements in the real world.

In 1999, the Lancaster team deployedGuide and evaluated it in a field trial forapproximately four weeks. Sixty touristsvisiting Lancaster used the system—peo-ple of all age groups, mostly without anyprevious experience with information nav-igation systems such as the Web.

The design and use experience gainedfrom the Guide project uncovered variousissues that related to the interplay of infra-structure and user interaction. For exam-ple, in the user interface, the team had toconsider the fact that the positioning sys-tem is not available everywhere a touristmight go. They chose to involve the userin disambiguation of the system stateshared between the human and computer(see Figure 2). This relates closely to theissue of dealing with ambiguity in predic-tion-based user interfaces investigated atGeorgia Tech.23 The general lesson to belearned from systems such as Guide is thatuser interaction tends to be governed bythe capabilities of the infrastructure, which,in ubiquitous computing environments, is



Figure 1. (a) The AT&T Active Bat systemand (b) the Bat wireless tag device. Figure courtesy of M. Addlesee, R.Curwen, S. Hodges, J. Newman, P.Steggles, A. Ward, and A. Hopper.7

more prone to changes in composition andavailability than in traditional applicationsettings.

The Cooltown projectAlthough the Web infrastructure was

never conceived as a general distributedsystems platform, its ubiquity and versa-tility have made it an attractive choice forlarge-scale application deployment. Notsurprisingly, researchers have investigatedcombining the Web infrastructure withubiquitous computing concepts. For exam-ple, Accenture’s CStar group has developedvarious demonstrators for augmentedcommerce, integrating Web-based e-com-merce services with context-aware clienttechnology and tagged real-world envi-ronments. (One example is the Pocket Bar-gain Finder, which integrates a barcodescanner and wireless Internet access toretrieve online quotes for products browsedin real-world shops.24) In a more generalway, Hewlett-Packard Labs’ Cooltownexplores opportunities arising from the con-vergence of Web technologies, wireless net-working, and new kinds of devices on theclient and server sides.25

The Cooltown project is developing aninfrastructure to support Web presence for“people, places, and things,” extending the

Web’s ad hoc nature and mechanisms forlocating and linking resources into the realworld of physical entities.25 Cooltowntechnology addresses physical integrationwith embedded server technology, virtualrepresentations of physical places (calledplace managers), real-world embeddedURLs (for example, emitted by beacons),and new interaction techniques such as e-squirt, a real world equivalent to drag-and-drop. HP internally deployed the technol-ogy to create office environments in whichmobile users could use resources (printers,projectors, and so forth) served by placemanagers (maintaining Web portals tolocations) and into which users can“squirt” URLs as handles to any kind ofinformation. Interestingly, users have notchanged the way they do things (for exam-ple, they don’t bring URLs rather than doc-ument copies into meetings), even whenthe benefits seem obvious. HP has alsoexternally deployed components of theCooltown system in a science museum,enabling the combination of physical andvirtual exploration.

From an overall systems perspective,Cooltown presents an interesting counter-point to many other ubiquitous computingsystems by placing the user in the controlloop. For example, other proposed plat-

forms suggest that dynamically requestedresources should be discovered automati-cally, using service discovery protocols. Incontrast, Cooltown involves the user inresource discovery—for example, using adhoc compilation of descriptions and linksto available resources in the Web page of aparticular location (each place managermaintains a page for a given location). Theargument for user involvement is that it avoids having to design systems that can carry out task analysis or form com-plex associations between components—processes that we have already observed areextremely difficult to achieve in computersystems.

The MediaCup experienceTo cite Mark Weiser and John Seely

Brown, “Ubiquitous computing is funda-mentally characterized by the connection ofthings in the world with computation.”26 Inmany deployed ubiquitous computing sys-tems, this connection is established indirectlythrough the approximate location of physi-cal objects. A more direct approach, appliedin Cooltown and many other projects, is toattach pointers from everyday objects to enti-ties in the computational world, using, forexample, RFID tags.27 Taking this approachone step further, computation might be com-pletely embedded in everyday objects to turnthem into permanently connected residentsin a ubiquitous computing world. TheMediaCup project at the University of Karl-sruhe is an experimental deployment ofeveryday objects activated in this sense.10

The guiding principle in the MediaCupproject was to augment objects with a dig-ital presence while preserving their origi-nal appearance, purpose, and use. Thefirst objects prototyped were coffee cupsequipped with a low-power microcon-troller, embedded sensors, and wirelesscommunications (see Figure 3). The em-bedded technology lets the cups sense theirphysical state and map sensor readingsautonomously to a domain-specific model


Figure 2. Involving the user indisambiguating the shared state in theGuide system.

of the cup. This object model is broad-casted at regular intervals over the wirelesslink to establish the object’s digital pres-ence. Karlsruhe first deployed the aug-mented coffee cups in September 1999 inan office environment. Since then, a smallpopulation of cups, typically five to six ata time, has been in everyday use. The cupsare not personalized but are simply ashared resource in the lab’s kitchen. Whilethere are a few staff who are dedicatedusers, the cups are mostly used occasion-ally by lab members and visitors.

The MediaCup deployment is a long-term experiment aimed at understandingopportunities that might arise from thedigital presence of mundane objects. Eventhough the MediaCup was designed with-out any specific use or application in mind,as the project evolves, applications that useMediaCup information have emerged. Aninitial application simply visualized cupsand their state on a 2D map of the officeenvironment. This was followed by moreembedded applications such as door platesthat used the aggregation of hot cups in aroom to infer and indicate meetings, andwrist-worn PCs that issue warnings if theuser picks up a cup with coffee or tea that’sstill too hot to drink.

AnalysisFrom a technical perspective, many of

these projects have faced common prob-lems in developing deployable prototypes.For example, energy concerns heavilyinfluenced designs for the Active Badge,Active Bat, ParcTab, and MediaCup. More-over, these concerns affected many aspectsof the systems’ designs, highlighting theintegrated nature of ubiquitous computingimplementations (for example, decisionsregarding the temporal accuracy of sensorreadings were influenced by energy con-siderations but have a significant impacton the types of application that can be sup-ported). A further example of the closecoupling between aspects of ubiquitouscomputing system design can be seen in theway in which projects such as Guide sup-port variations in networking connectivityin both the underlying protocols and theuser interface. For the research community,

this implies that we will need forums to dis-cuss projects and results that cut across var-ious computer science topics.

In terms of deployment, only throughexperience can we understand the cost-versus-benefit issues in ubiquitous com-puting systems. For example, the ActiveBadge system implies a cost for the badgewearer (some inconvenience, loss of pri-vacy), so the badge must offer a benefit (andnot just a communal benefit). For example,to boost the number of badge wearers foran active badge system deployed at Lan-caster, applications were designed thatspecifically benefited the wearer. Specifi-cally, we offered a collection of user-sensi-tive public monitors that provided tailoredinformation to users as they approachedthe screen (though we eventually had toremove the monitors from the building’scorridors because they did not comply withsafety regulations).

Probably the most significant lesson,however, is the importance of deployingprototypes—not just to evaluate their util-ity but also to explore ideas, discover newviewpoints, and unearth unexpected issues.Design and usage data is crucial to advanc-ing our understanding of future ubiquitouscomputing systems. However, obtainingsuch experience is nontrivial. To achievethe levels of deployment required for real-

istic user trials often involves researchersaddressing design tradeoffs in areas suchas cost, functionality, and robustness thatare more commonly the focus of productdevelopment teams.

Research challengesDeploying ubiquitous computing sys-

tems provides insight into their design anduse, but deploying systems beyond the lim-ited existing prototypes will require signif-icant progress toward integration. Here,we consider major research challenges wemust overcome before achieving this goal.This list is not exhaustive; it simply pro-vides illustrative examples of the require-ments that supporting integration placeson components of ubiquitous computingsystems.

Component interactionTo gain leverage from the substantial

work carried out in the distributed systemscommunity, future components of ubiqui-tous computing systems must clearly fol-low the same basic principles as open dis-tributed systems. They should be designedand implemented in an open and extensi-ble manner, letting us combine componentsto form applications unforeseen at the timeof their deployment.

Technically, this implies obvious featuressuch as open interfaces and support forintercomponent communication.28,29 How-ever, deployed ubiquitous computing sys-tems will require assurances from theircomponents in terms of metrics such as per-formance, security, and reliability. Further-more, the characteristics of ubiquitous com-puting environments place demands onexisting platforms that the platforms havenot been designed to address. For example,many existing platforms perform poorlywhen applied to the type of saturated com-puting environments Weiser described.30

(“System Software for Ubiquitous Com-puting” in this issue discusses designingsoftware components for integrated ubiq-uitous computing systems.)

Adaptation and contextual sensitivity

The environment in which a ubiquitous



Figure 3. The MediaCup.

computing component functions is subject tochange.31 Such changes might be promptedby variations in resource availability as aresult of failures or the deployment of newservices or by variations in patterns of usageor mobility. The importance of adaptation iswell understood in the field of mobile com-puting.32 However, it is significantly morecomplicated in ubiquitous computing sys-tems, where there is a need to respond to a much larger set of contextual triggers. At a component level, components will berequired to adapt internally. More impor-tantly, we might also have to substantiallyreconfigure applications involving multiplecomponents.

The need to manage these configurationchanges in ad hoc ubiquitous environmentsposes significant problems. One approach toaddressing these problems—one that hasreceived considerable attention of late—is to use reflective middleware platforms.33

Although researchers have generally focusedon using reflection to support adaptation inmobile environments, its application to ubiq-uitous computing systems is an obviousextension.

Appropriate management mechanismsand policies

As the number of deployed componentsincreases, system management will likelybecome increasingly problematic. While wewant zero-configuration, low-maintenancesystems, the reality is that substantial sys-tem management will still likely be re-quired. We might expect future componentsto support standardized management inter-faces enabling, for example, tasks such asconfiguration management over a widerange of components. A particular chal-lenge for ubiquitous computing systems isthat this management is unlikely to occurwithin the context of a single administra-tive domain. Indeed, for many components,the administrative domain might changedynamically—for example, depending onthe proximity of different users or devices.The combination of requirements for low(or zero) administration, multidomainmanagement, and support for rapid recon-figuration will likely raise new challengesfor system management.

Component association and taskanalysis

In studying Weiser’s scenarios of a ubiq-uitous computing world, we clearly get thesense of some form of intelligence work-ing on the user’s behalf to coordinate theactions of components in the infrastruc-ture. Two areas in which this is particularlyevident are the system’s ability to accu-rately determine a user’s task and intentionand its ability to develop associationsbetween components to assist the user inthese activities. Achieving these objectivesin anything other than extremely limiteddomains is an unsolved problem.

Viable economic models and supporting infrastructure

One often-cited reason for the relativelack of success in deploying ubiquitouscomputing systems is that none of theapplication scenarios seem likely to gener-ate significant revenue. Users might pay tolive in a ubiquitous computing world, butit is difficult to imagine them paying a lotof money for any one application or fea-ture (this of course assumes that the searchfor a single “killer” application is unsuc-cessful). Consequently, the cost of deploy-ing and operating a given componentmight need to be recovered in the form ofmany small contributions from applica-tions that use the component. This willrequire support from components in termsof billing and accounting at a level previ-ously unseen in widespread distributedsystems.

User interface integrationAs the number of applications operating

in a ubiquitous computing environmentincreases, we’ll need coordination betweenthese applications to ensure we can providea reasonable user interface.18 This coordi-nation might range from traditional areassuch as arbitrating screen usage to new chal-lenges such as deciding which applicationmay use the intensity of the light in a roomto communicate with the user. While userinterface designs for ubiquitous computingsystems are covered elsewhere in this issue,it is important to stress the lessons learnedfrom the Guide system: the infrastructure’s

capabilities directly affect the user interface.So, support of the user interface becomes anissue for all components, at whatever levelof the system they operate.

Social, legal, and technical solutionsto privacy and security concerns

Support for adequately handling per-sonal data challenges both legislators andubiquitous systems developers. Tradition-ally, data protection legislation has tendedto prohibit any capture and storage of per-son-related data and has only allowedexceptions bounded by a clearly definedpurpose, at the end of which data recordshad to be deleted. This approach to pri-vacy protection is inadequate for a mod-ern, open information society. For exam-ple, to demand that sensitive data bedeleted after its use is clearly out of syncwith the Internet. Most important, legisla-tion must acknowledge that person-relateddata has become a currency in the infor-mation economy.

Here lies a core problem for developerswho need to create systems that betteraddress privacy issues. Currently, usersdon’t fully understand how the electronictrails they create can be used, so they can-not understand their personal data’s value.A key challenge for future ubiquitous sys-tem designers is to empower users to eval-uate the tradeoff between protection of pri-vacy and access to improved service.Meanwhile, legislation must contribute bydefining the boundaries within which suchtrade-offs may occur.

The challenges we’ve describedfocus on integration. However,although integration is key tofuture ubiquitous systems, we

might also design such systems with thenotion of interference in mind. By way ofanalogy, consider the current situationregarding electromagnetic interferencefrom devices. An appropriate authoritymust certify most electronic equipment toconfirm that it does not cause undue inter-ference to neighboring electronic systems.This certification lets consumers purchase


products without worrying about negativeside effects on their existing equipment.

We expect the same concerns will arisewhen consumers start purchasing anddeploying components of a ubiquitouscomputing environment. However, in thiscase, we’ll need to extend the concept ofinterference beyond the notion of physicalinterference to include interference at thelogical (software) level. For example, if twoubiquitous computing applications haveconflicting requirements in terms of infra-structure services, they will need to resolvethis conflict, ideally without user interven-tion. It seems reasonable to assume thatfuture ubiquitous computing componentswill need to be both integration-friendlyand interference-free for wide-scale deploy-ment. How this capability is tested willlikely be a significant research challenge inits own right.

ACKNOWLEDGMENTS

We thank Keith Cheverst, Tim Kindberg, MarcLangheinrich, and all the reviewers for their help inpreparing this article.

REFERENCES1. M. Weiser, “ The Computer of the 21st Cen-

tury,” Scientific American, vol. 265, no. 3,Sept. 1991, pp. 66–75 (reprinted in thisissue, see pp. 19–25).

2. M. Weiser and J. Brown, “Designing CalmTechnology,” PowerGrid J., v 1.01, July1996; also see “The Coming Age of CalmTechnology,” Beyond Calculation: The NextFifty Years of Computing, Springer-Verlag,New York.

3. M. Rahnema, “Overview of the GSM Sys-tem and Protocol Architecture,” IEEEComm., Apr. 1993, pp. 92–100.

4. U. Leonhardt, Supporting Location-Aware-ness in Open Distributed Systems, doctoraldissertation, Dept. of Computing, ImperialCollege London, 1998.

5. European Commission, Directive 95/46/ec,Official J. L 281, 23 Oct. 1995, pp. 31–50;http://europa.eu.int/eur-lex/en/lif/dat/1995/en_395L0046.html.

6. M. Langheinrich, “Privacy by Design—Prin-ciples for Privacy-Aware Ubiquitous Sys-tems,” Ubicomp 2001: Ubiquitous Com-puting, Lecture Notes in Computer Science,Springer-Verlag, Berlin, 2001, pp. 273–291.

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

8. K. Cheverst et al., “Experiences of Devel-oping and Deploying a Context-AwareTourist Guide: The Lancaster GUIDE Pro-ject,” Proc. ACM MobiCom 00, ACMPress, New York, 2000. pp. 20–31.

9. G.D. Abowd, “Classroom 2000: An Exper-iment with the Instrumentation of a LivingEducational Environment,” IBM Systems J.,vol. 38, no. 4, 1999, pp. 508–530.

10. M. Beigl, H.W. Gellersen, and A. Schmidt,“Mediacups: Experience with Design and Useof Computer-Augmented Everyday Objects,”Computer Networks, vol. 35, no. 4, Mar.2001, pp. 401–409.

11.B.N. Schilit, N.I. Adams, and R. Want.“Context-Aware Computing Applications,”Proc. IEEE Workshop on Mobile ComputerSystems and Applications, IEEE CS Press,Los Alamitos, Calif., 1994, pp. 85–90.

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

13. R. Want, “Ten Lessons Learned about Ubiq-uitous Computing,” PowerPoint presentationfrom invited keynote (at the Dagstuhl Semi-nar on Ubiquitous Computing), Sept. 2001;www.inf.ethz.ch/vs/events/dag2001/slides/roy-lessons.pdf.

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

15.A. Harter and A. Hopper, “A DistributedLocation System for the Active Office,”IEEE Network, vol. 8, no. 1, Jan./Feb. 1994,pp. 62–70.

16.F. Bennett, T. Richardson, and A. Harter,“Teleporting—Making Applications Mo-bile,” Proc. IEEE Workshop on MobileComputer Systems and Applications, IEEECS Press, Los Alamitos, Calif., 1994, pp.82–84.

17.R. Want, M. Weiser, and E. Mynatt, “Acti-vating Everyday Objects,” DARPA/NISTWorkshop on Smart Spaces, 1998; sandbox.xerox.com/want/papers/aeo-nist-jul98.pdf.

18.A. Harter et al., “The Anatomy of a Con-text-Aware Application,” Proc. 5th Ann.ACM/IEEE Int’l Conf. Mobile Computingand Networking (MobiCom 99), ACMPress, New York, 1999, pp. 59–68.

19.S. Long et al., “Rapid Prototyping of MobileContext-Aware Applications: The Cyber-guide Case Study,” Proc. 2nd Intl. Conf.Mobile Computing and Networking (Mobi-Com 96), ACM Press, New York, 1996, pp.97–107.

20. S. Feiner et al., “A Touring Machine: Proto-typing 3D Mobile Augmented Reality Sys-tems for Exploring the Urban Environment,”Personal Technologies, vol. 1, no. 4, Dec.1997, pp. 208–217.

21.N. Davies et al., “Caches in the Air: Dis-seminating Information in the Guide Sys-tem,” Proc. 2nd IEEE Workshop MobileComputing Systems and Applications(WMCSA 99), IEEE Press, Piscataway, N.J.,1999, pp. 11–19.

22.K. Cheverst et al., “Developing a Context-Aware Electronic Tourist Guide: Some Issuesand Experiences,” Proc. 2000 Conf. HumanFactors in Computing Systems (CHI 2000),ACM Press, New York, 2000, pp. 17–24.

23. J. Mankoff, S.E. Hudson, and G.D. Abowd,“Interaction Techniques for Ambiguity Res-olution in Recognition-Based Interfaces,”Proc. UIST, 13th Ann. ACM Symp. UserInterface Software and Technology, ACMPress, New York, 2000, pp. 11–20.

24.A.V. Gershman, J.F. McCarthy, and A.E.Fano, “Situated Computing: Bridging theGap between Intention and Action,” Proc.3rd Int’l Symp. Wearable Computing (ISWC99), IEEE Press, Piscataway, N.J., 1999, pp.3–9.

25.T. Kindberg et al., “People, Places, Things:Web Presence for the Real World,” 3rd Int’lIEEE Workshop Mobile Computing Sys-tems and Applications IEEE CS Press, LosAlamitos, Calif., 2000.

26.M. Weiser and J. Brown, “Designing CalmTechnology,” PowerGrid J., vol. 1.01, July1996.

27.R. Want et al., “Bridging Real and VirtualWorlds with Electronic Tags,” Proc. ACMSIGCHI, ACM Press, New York, 1999, pp.370–377.

28.The Common Object Request Broker:Architecture and Specification, Revision2.5., The Object Management Group, Need-ham, Mass., Sept. 2001.

29. Jini Architectural Overview, white paper,Sun Microsystems, Palo Alto, Calif., 1999.

30.A. Friday, N. Davies, and E. Catterall, “Sup-porting Service Discovery, Querying andInteraction in Ubiquitous Computing Envi-ronments,” Proc. 2nd ACM Int’l WorkshopData Engineering for Wireless and MobileAccess (MobiDE 2001), ACM Press, NewYork, 2001.



31.M. Satyanarayanan, “Pervasive Computing:Vision and Challenges,” Personal Comm.,vol. 8, no. 4, Aug. 2001, pp. 10–17.

32.R. Katz, “Adaptation and Mobility in Wire-less Information Systems,” IEEE PersonalComm., vol. 1, no. 1, 1st Quarter 1994, pp.6–17.

33.M. Roman, F. Kon, and R.H. Campbell,“Reflective Middleware: From Your Deskto Your Hand,” IEEE Distributed SystemsOnline, vol. 2, no. 5, 2001; http://com-puter.org/dsonline/0105/features/rom0105_print.htm.

For more information on this or any othercomputing topic, please visit our DigitalLibrary at http://computer.org/publications/dlib.


the AUTHORS

Nigel Davies is an associate professor at the University of Arizona and a professor ofcomputer science at Lancaster University. He has participated actively in the mobilecomputing research community. He has managed numerous projects at Lancaster, andhe helped create its Mobile Computing group. In recognition of his work in establishingLancaster as a major research center for mobile computing, he was awarded a personalchair in the Computing Department in 2000. He holds a BSc and PhD in computer sci-ence, both from Lancaster University, UK. He is an associate editor in chief of IEEE Perva-sive Computing and an associate editor of IEEE Transactions on Mobile Computing. Con-

tact him at the Dept. of Computer Science, Univ. of Arizona, Gould-Simpson 721, PO Box 210077, Tucson,AZ 85721-0077; [email protected].

Hans-Werner Gellersen is professor for interactive systems at Lancaster University. Heobtained an MSc and PhD in computer science from the University of Karlsruhe. Hisresearch interest is in ubiquitous computing and context-aware systems. He is activelyinvolved in the formation of a ubiquitous computing research community, has initiatedthe HUC/Ubicomp conference series, and leads European research collaborations in theCommission’s program on future and emerging technologies. He is editor of Personaland Ubiquitous Computing. Contact him at the Computing Department, Lancaster Uni-versity, Bailrigg, Lancaster LA1 4YR, UK; [email protected].

PURPOSE The IEEE Computer Society is the world’s

largest association of computing professionals, and is

the leading provider of technical information in the

field.

MEMBERSHIP Members receive the monthly

magazine COMPUTER, discounts, and opportunities

to serve (all activities are led by volunteer mem-

bers). Membership is open to all IEEE members,

affiliate society members, and others interested in

the computer field.

B O A R D O F G O V E R N O R STerm Expiring 2002: Mark Grant, Gene F.Hoffnagle, Karl Reed, Kathleen M. Swigger, RonaldWaxman, Michael R. Williams, Akihiko Yamada

Term Expiring 2003: Fiorenza C. Albert-Howard, Manfred Broy, Alan Clements, RichardA. Kemmerer, Susan A. Mengel, James W. Moore,Christina M. Schober

Term Expiring 2004: Jean M. Bacon, RicardoBaeza-Yates, Deborah M. Cooper, George V.Cybenko, Wolfgang K. Giloi, Haruhisha Ichikawa, Thomas W. WilliamsNext Board Meeting: 10 May 02, Portland OR

I E E E O F F I C E R SPresident: RAYMOND D. FINDLAY

President-Elect: MICHAEL S. ADLER

Past President: JOEL B. SYNDER

Executive Director: DANIEL J. SENESE

Secretary: HUGO M. FERNANDEZ VERSTAGEN

Treasurer: DALE C. CASTON

VP, Educational Activities: LYLE D. FEISEL

VP, Publications Activities: JAMES M. TIEN

VP, Regional Activities: W. CLEON ANDERSON

VP, Standards Association: BEN C. JOHNSON

VP, Technical Activities: MICHAEL R. LIGHTNER

President, IEEE-USA: LeEARL A. BRYANT

EXECUTIVE COMMITTEEPresident: WILLIS K. KING*University of HoustonDept. of Comp. Science501 PGHHouston, TX 77204-3010Phone: +1 713 743 3349 Fax: +1 713 743 [email protected]

President-Elect: STEPHEN L. DIAMOND*

Past President: BENJAMIN W. WAH*

VP, Educational Activities: CARL K. CHANG *

VP, Conferences and Tutorials: GERALD L. ENGEL*

VP, Chapters Activities: JAMES H. CROSS†

VP, Publications: RANGACHAR KASTURI†

VP, Standards Activities: LOWELL G. JOHNSON

(2ND VP)*

VP, Technical Activities: DEBORAH K.

SCHERRER(1ST VP)*

Secretary: DEBORAH M. COOPER*

Treasurer: WOLFGANG K. GILOI*

2001–2002 IEEE Division VIII Director:THOMAS W. WILLIAMS

2002–2003 IEEE Division V Director:GUYLAINE M. POLLOCK†

Executive Director: DAVID W. HENNAGE†

*voting member of the Board of Governors

COMPUTER SOCIETY WEB SITEThe IEEE Computer Society ’s Web si te, at http://computer.org, offers information and sam-ples from the society’s publications and conferences,as well as a broad range of information about techni-cal committees, standards, student activities, andmore.

COMPUTER SOCIETY O F F I C E SHeadquarters Office

1730 Massachusetts Ave. NW Washington, DC 20036-1992Phone: +1 202 371 0101 • Fax: +1 202 728 9614E-mail: [email protected]

Publications Office10662 Los Vaqueros Cir., PO Box 3014Los Alamitos, CA 90720-1314Phone:+1 714 821 8380E-mail: [email protected] and Publication Orders:Phone: +1 800 272 6657 Fax: +1 714 821 4641E-mail: [email protected]

European Office13, Ave. de L’AquilonB-1200 Brussels, BelgiumPhone: +32 2 770 21 98 • Fax: +32 2 770 85 05E-mail: [email protected]

Asia/Pacific OfficeWatanabe Building1-4-2 Minami-Aoyama, Minato-ku,Tokyo 107-0062, JapanPhone: +81 3 3408 3118 • Fax: +81 3 3408 3553E-mail: [email protected]

E X E C U T I V E S T A F FExecutive Director: DAVID W. HENNAGEPublisher: ANGELA BURGESSAssistant Publisher: DICK PRICEDirector, Volunteer Services: ANNE MARIE KELLYChief Financial Officer: VIOLET S. DOANDirector, Information Technology & Services:ROBERT CAREManager, Research & Planning: JOHN C. KEATON

11-FEB-2002

Documents

Beyond prototypes: challenges in deploying ubiquitous systems