i;;ii!~!;;!il; ¸ 5 ¸ SCIENCE I

~computmg enhances computer use by ~any computers ava le ~hroughout the physical e p, while making them effectlvely invisible to the us ~ticle explains what is new ~and different ~about the science involved in ubiquit~i~s . ¸¸¸!ill! brief overview of ubiquit( is computing, :ough a series o~ e pli drawn from des of comp '~;~ence:i ~trdware com- ponei cols, inter~ ion substrates and pens), applid ions, privacy, ~[ on iutmg offers a and~xc 'itm r ..... acrGi the spectrum :omputer ~ince we>~work : at .~erox Pal~ Alto Research .~a few pla~ have benin work on this

ilil;~ ~'n computing environment in which ly interacting with hundreds of nearby ~ter~ected com puters. The goNis to achieve the most elI~ iive kind of technology, that whih is essentially invisible the user. To bring compu~ee~ to ~his point while ~er will require radically new kinds of com- ~d shapes to be available to each person. I ~ld "Ubiquitous COmputing" (Ubicomp) ~ method for ubiquitous computing is stan- ~al : the construction of

• ~es of the necessai'y infrastructure i ent ;bug the viability of the systems l~,e~eryday ug~ td ~ colle~ues serving as gui~a pigs. This is ...... -I{-E.ILL.

an important step toward ensuring become effectively invisible by taking world, leading to another "desktop that our infrastructure research is over the human sensory and affector model" [2], I wanted to put the new robust and scalable in the: face of the systems [19]. VR is extremely useful kind of computer out in this world of details of the real world. in scientific and enter- concrete information conveyers. The idea of ubiquitous computing tainment, and will be very significant Since these written artifacts occur in first arose from contemplating the for those niches. But as a tool for many different sizes and shapes, with place of today's computer in actual productively changing everyone's malay different qualities, I wanted activities of everyday life. In particu- relationship to computation, it has the computer embodiments to be of lar, anthropological studies of work two crucial flaws: first, at the present many sizes and shapes, including tiny life [14, 22] teach us that people pri- time (1992), and probably for de- inexpensive ones that could bring marily work in a world of shared sit- cades, it cannot produce a simulation computing to everyone. uations and unexamined technologi- of significant verisimilitude at rea- The physical world comes in all cal skills. The computer today is sonable cost (today, at any cost). This sizes and shapes. For practical rea- isolated from the overall situation, means that users will not be fooled sons our ubiquitous computing work however, and fails to get out of the and the computer will not be out of begins with just three different sizes way of the work. In other words, the way. Second, and most impor- of devices: enough to give some rather than being a tool through tant, it has the goal of fooling the scope, not enough to deter progress. which we work, and thus disappear- user--of leaving the everyday physi- The first size is the wall-sized interac- ing from our awareness, the com- cal world behind. This is at odds with tive surface, analogous to the office puter too often remains the focus of the goal of better integrating the whiteboard or the home magnet- attention. And this is true through- computer into human activities, since covered refrigerator or bulletin out the domain of personal comput- humans are of and in the everyday board. The second size. is the note- ing as currently implemented and world. pad, envisioned not as a personal discussed tsar the future:, whether Ubiquitous computing is explor- computer but as analogous to scrap one thinks of personal computers, ing quite different ground from per- paper to be grabbed and used easily, palmtops, or dynabooks. The charac- sonal digital assistants, or the idea with many being used by a person at terization of the future computer as that computers should be autono- one time. The cluttered office desk the "intimate computer" [12], or mous agents that take on our goals. or messy front-hall table are real-life "rather like a human assistant" [25] The difference can be characterized examples. Finally, the third size is the makes this attention to the machine as follows: Suppose you want to lift a tiny computer, analogous to tiny in- itself particularly apparent. heavy object. You can call in your dividual notes or Post-it notes, and Getting the computer out of the strong assistant to lift it for you, or also similar to the tiny little displays way is not easy. This is not a graphi- you yourself can be made effort- of words found on book spines, light cal (GUI) problem, but lessly, unconsciously, stronger and switches, and hallways. Again, I saw is a property of the whole context of just lift it. There are times when both this not as a personal computer, but usage of the machine and the attri- are good. Much of the past and cur- as a pervasive part of everyday life, butes of its physical properties: the rent effort for better computers has with many active at all times. I called keyboard, the weight and desktop been aimed at the former; ubiquitous these three sizes of computers position of '.screens, and so on. The computing aims at the latter. boards, pads, and tabs, and adopted problem is not one of "interface." The approach I took was to at- the slogan that, for each person in an For the same reason of context, this tempt the definition and construc- office, there should be hundreds of is not a multimedia problem, result- tion of new computing artifacts for tabs, tens of pads, and one or two ing from any particular deficiency in use in everyday life. I took my inspi- boards. Specifications for some pro- the ability to display certain kinds of ration from the everyday objects totypes of these three sizes in use at real-time data or integrate them into found in offices and homes, in par- PARC are shown in Figure 1. applications. (Indeed, raultimedia ticular those objects whose purpose is This then is Phase I of ubiquitous tries to grab attention, the opposite to capture or convey information. computing: to construct, deploy, and of the ubiquitous computing ideal of The most ubiquitous cu .... t infnr_ invisibility.) The challenge is to cre- mational technology el ate a new kind of relationship of peo- artifacts is the use of writ ple to computers, one in which the primarily words, but in, computer would have to take the pictographs, clocks, and lead in becoming vastly better at get- of symbolic communical ting out of the way, allowing people than attempting to repr to just go about their live,;. objects inside the virtu; In 1988, when I started PARC's work on ubiquitous computing, vir- tual reality (VR) came the closest to enacting the principles we believed important. In its ultimate envision- ment, VR causes the computer to

76 July 1993/Vol.361 No.7 ClOMMUN|q~AVlOMS OF THIE ACM -RoE-A-L* learn from a computing environ- small analog/digital converter for the processor can actually time the pulse ment consisting of tabs, pads, and pressure-sensitive screen, and analog widths in timing loops. Our boards. This is only Phase I, because sense circuitry for the IR receiver. current design has insufficient stor- it is unlikely to achieve optimal invisi- Interestingly, although we have been age, and we are increasing the bility. (Later phases are yet to be de- approached by several chip manu- amount of nonvolatile RAM in fu- termined.) But it is a start down the facturers about our possible need for ture tabs from 8K to 128K. The tab's radical direction, for computer sci- custom chips for the Tab, the Tab is goal of casual use similar to that of ence, away from emphasis on the not short of places to put chips. The Post-it notes puts it into a design machine and back to the person and display size leaves plenty of room, space generally unexplored in the his or her life in the world of work, and the display thickness dominates commercial or research sector. play, and home. total size. Off-the-shelf components are more than adequate for explor- Pad Hardware Prototypes ing this design space, even with our The pad is really a family of note- New hardware systems design for severe size, weight, and power con- book-sized devices. Our initial pad, ubiquitous computing has been ori- straints. the ScratchPad, plugged into a Sun ented toward experimental plat- A key part of our design philoso- SBus card and provided an forms for systems and applications of phy is to put devices in everyday use, X-Window-system-compatible writ- invisibility. New chips have been less not just demonstrate them. We can ing and display surface. This same important than combinations of ex- only use techniques suitable for design was used inside our first wall- isting components that create experi- quantity 100 replication. This ex- sized displays, the liveboards, as well. mental opportunities. The first ubiq- cludes certain techniques that could Our later untethered pad devices, uitous computing technology to be make a huge difference, such as the the XPad and MPad, continued the deployed was the Liveboard [6], integration of components onto the system design principles of which is now a Xerox product. Two display surface itself. This technol- X-compatibility, ease of construction, other important pieces of prototype ogy is being explored at PARC, as and flexibility in software and hard- hardware supporting our research at well as other research organizations. ware expansion. PARC are the Tab and the Pad. While it is very promising, it is not yet As I write this article, at the end of ready for replication. 1992, commercial portable pen de- Tab The Tab architecture incorporates vices have been on the market for The ParcTab is a tiny information a careful balance of display size, two years, although most of the early doorway. For user interaction it has a bandwidth, processing, and memory. companies have now gone out of pressure-sensitive screen on top of For instance, the small display means business. Why should a pioneering the display, three buttons positioned that even the tiny processor is capa- research lab build its own such de- underneath the natural finger posi- ble of providing a four-frames-per- vice? Each year we ask ourselves the tions, and the ability to sense its posi- second video rate, and the IR band- same question, and so far three tion within a building. The display width is capable of delivering this. things always drive us to continue to and touchpad it uses are standard The bandwidth is also such that the commercial units. The key hardware design prob- lems affecting the pad are physical size and power consumption. With several dozens of these devices sitting around the office, in briefcases, in pockets, one cannot change their bat- teries every week. The PARC design uses the 8051 chip to control detailed interactions, and includes software that keeps power usage down. The major outboard components are a • ,RoE.A-L o

design our own pad hardware: giving extreme flexibility. For in- mance has dominated the need for First, we need the correct balance stance, changing the power control low-power consumption in processor of features--this is the essence of functions and adding high-quality design. However, recognizing the systems design. The commercial de- sound were relatively simple FPGA new requirements of ubiquitous vices all aim at particular niches, bal- changes. The MPad has both IR (tab computing, a number of people have ancing their design to that niche. For compatible) and radio communica- begun work in using additional chip research we need a rather different tion built-in and includes sufficient area to reduce power rather than to balance, particularly for ubiquitous uncommitted space for adding new increase performance [16]. One key computing. For instance, can the circuit boards later. It can be used approach is to reduce the clocking device communicate simultaneously with a tether that provides it with frequency of their chips by increas- along muh-iple channels? Does the recharging and operating power and ing pipelining or parallelism. Then, support multipro- an Ethernet connection. The operat- by running the chips at reduced volt- cessing? What about the potential for ing system is a standalone version of age, the effect is a net reduction in high-speed tethering? i[s there a the public-domain Portable Common power, because power falls off as the high-quality pen? Is there a high- Runtime developed at PARC [28]. square of the voltage, while only speed expansion port sufficient for about twice the area is needed to run video in and out? Is sound in/out and The Computer Science of at half the clock speed. ISDN connectivity available? Op- Ublcomp Power = CL*Vdd2*f tional keyboard? Any or,e commer- To construct and deploy tabs, pads, where CL is the gate capacitance, cial device tends to satisfy some of and boards at PARC, we found our- Vdd the supply voltage, these needs, ignore others, and selves having to readdress some of and f the clocking frequency. choose a balance of the ones it does the well-worked areas of existing satisfy, optimizing its niche, rather computer science. The fruitfulness This method of reducing power than ubiquitous com]?uting-style of ubiquitous computing for new leads to two new areas of chip design: scrap computing. The ba)lance for us computer science problems justified circuits that will run at low power, emphasizes communication, system our belief in the ubiquitous comput- and architectures that sacrifice area memory, multimedia, and expansion ing framework. for power over performance. The ports. The following subsections "as- second requires some additional Second, apart from balance of fea- cend" the levels of organization of a comment, because one might sup- tures are the requirements for par- computer system, from hardware to pose one would simply design the ticular features. Key among these are application. One or two examples of fastest possible chip, and then run it a pen emphasis, connection to re- computer science work required by at reduced clock and voltage. How- search envJironments such as Unix, ubiquitous computing are described ever, as Lyon illustrates, circuits in and communication emphasis. A for each level. Ubicomp is not yet a chips designed for high speed gener- high-speed (>64KB/sec) wireless coherent body of work, but consists ally fail to work at low voltages. Fur- capability is not built into any com- of a few scattered communities. The thermore, attention to special circuits mercial devices, and they do not gen- point of this article is to help others may permit operation over a much erally have a sufficiently high-speed understand some of the new re- wider range of voltage operation, or port to add such a radio. Commercial search challenges in ubiquitous com- achieve power savings via other spe- devices generally come with DOS or puting, and inspire them to work on cial techniques, such as adiabatic Penpoint, and while we have devel- them. This is more akin to a tutorial switching [16]. oped in both, they are nor our favor- than a survey, and necessarily selec- ite researc]h vehicles because they tive. The areas included are hard- Wireless lack full access and customizability. ware components (e.g., chips), net- A capable of accom- The third factor driving our own work protocols, interaction sub- modating hundreds of high-speed pad design,; is ease of expansion and strates (e.g., software for screens and devices for every person is well be- modification. We need full hardware pens), applications, privacy, and yond the commercial wireless sys- specifications, complete operating computational methods. tems planned for the next 10 years system source code, and the ability to [