Last winter during a snowstorm, while edging his way home from IBM's Thomas J. Watson Research Center
in New York's Westchester County, Michael Greenwood had a brainstorm. "I saw all these teams of people trying to restore power," he recalls, "and I thought, this must be a nightmare." How can utility companies quickly assemble the large teams of workers needed to fix the lines? Greenwood, senior manager of pervasive computing solutions, knew that in the new age of choice opened up by the deregulation of the electric power industry, a dissatisfied customer could quickly become an ex-customer. By the time he pulled into his driveway 45 minutes later, he had outlined a solution in his head. Greenwood imagined a system that could quickly determine the size and makeup of a team appropriate for any emergency: how many engineers, how many tree climbers and so on. It would then automatically start recruiting the team based on who was available. Some would be paged, others sent email. Voice and text messages would arrive on smart cellular phones. An entire emergency response force could be rapidly assembled with the
click of a mouse.
Greenwood's vision on that snowy winter night quickly outgrew the confines of power lines
and grew into a system that is today known as TeamBuilder, which the City of Orlando,
Florida, has embraced to manage its response to disasters, and which is being considered
for applications ranging from health care to corporate operations ( see "Organizing in 'Pervasive Space' "). Ideas like Greenwood's are examples of what many at IBM believe is the next stage in the evolution of computers, pervasive computing.
The basic insight behind pervasive computing is that as processing power gets cheaper and more compact, the bulky and complex all-purpose desktop computer is giving way to a new generation of smart devices like personal digital assistants, screen phones, kiosks, smart cell phones and even intelligent credit cards. "These devices are going to liberate us from that box on our desks," says Mark Bregman, who leads IBM's year-old pervasive computing organization. "They are the logical extension of the networked world we live in, and of e-business."
The universe of pervasive computing is vast and expanding rapidly. In 1997, according to GartnerGroup, people around the world bought more than 100 million cellular handsets but only 70 million PCs. Hambrecht & Quist estimates that the market for smart portable devices that can communicate on the global information network is growing at 30 percent per year. By 2001 they will number 1.2 billion. By 2002 less than half of all Web-enabled devices sold will be PCs. "Pervasive computing is going to put sophisticated technology in our hands, literally," says Bregman.
The problem, of course, is to make all these devices, with their vastly different capabilities, work together seamlessly. Typically, a person will interact with smart devices throughout the day, in the home, on the road and at work. "Our vision is that people will be using pervasive devices to access all kinds of information, but as they move from device to device they want access to the same information," says Ajei Gopal, director of technology for the pervasive computing organization.
The space of pervasive computing is far too large for the as-yet small organization to conquer in one grand campaign. So Bregman and his colleagues have had to set priorities. Developing infrastructure, for example, will take precedence over devices. "Without the infrastructure," Bregman explains, "that cell phone against your head is just a piece of plastic." Moreover, Gopal adds, "We're focused on using solutions to specific problems to help us refine our message to get something to market relatively quickly."
To achieve this goal, Gopal's group has divided pervasive space into three broad categories: the home, the car and the mobile devices that move between these two domains. In the home, for example, consumers will access their personal information as well as information and shopping through screen phones, set-top boxes and other devices. In the car, the focus will be less on entertainment than on navigational information and services like directions to the nearest gas station or voice-enabled access to email. The car itself might be connected by the network to the dealer, who would remind the driver of scheduled maintenance or even deliver new engine software. Business travelers will retrieve detailed travel information on their Web-enabled cell phones and will be able to quickly make and update travel plans.
THE INFOPYRAMID
The billions of devices that will populate pervasive space will be almost as different in their capabilities as the people they serve. The PalmPilot™, a pocket-sized personal digital assistant, for example, has a small screen that displays no color and only limited shades of gray; pagers have even smaller, less capable displays. Automotive devices may have no display at all. Some devices will have high-bandwidth pipelines to the Internet; others will have to sip data through narrow, low-bandwidth connections. Some will be stuffed with memory and fast processors; others will be relatively dumb. But all will be expected to provide access to the same basic information.
IBM conceptualizes Internet multimedia content as a two-dimensional "infopyramid." One axis of the pyramid refers to the type of information -- text, image, audio or video. The other axis refers to the fidelity of the information. This concept allows the presentation of multimedia content to take into account such factors as the user's preferences and profile, the capabilities of the client device and the network bandwidth. Furthermore, the presentation for a given device need not be fixed: when batteries are running low, even a device with a high-bandwidth connection might switch to more compressed forms of data.
The task of tailoring information to this pyramid is known as transcoding, and it encompasses a wide variety of solutions. As an example, Watson researcher Rakesh Mohan notes that "video on the desktop won't translate easily to a PalmPilot." To solve this problem, his group is inventing ways to automatically select key frames from the video stream and deliver them as a low-bandwidth slide show. In some cases the device may forgo video and just use audio or closed-captioned information.
A high-resolution color image will overwhelm many portable displays. Richard LaMaire, manager of mobile networking at Watson, and his colleagues have developed techniques for "fuzzing" pictures by throwing away resolution or color information. A slider bar lets users select how much resolution they sacrifice or how long they wait for a download. Fuzzy faces may be good enough when reading the news, but parents will want -- and be willing to wait for -- sharper pictures of their children.
Similarly, Web pages are often too rich in graphics for some types of displays. IBM is exploring ways of taking this data and digesting it for delivery to portable devices. Some methods take Web information that was created with only the desktop in mind and converting it to forms more easily consumed by mobile gadgets. These techniques involve reformatting the data, eliminating graphic elements and even creating automatic summaries of the text. New Web languages such as XML will simplify the task of adapting Web pages for devices other than browsers (see "XML: the next big thing"). "XML is extremely important to pervasive," says Gopal, "because it lets you separate the rendering of the data from the semantics of the data. It is important because in some sense it represents a new lingua franca for storing self-describing data."
Converting information for different devices is easier if its creators have the different end users in mind. IBM is developing tools -- and promoting standards -- to allow authors to prioritize their information. A news outlet such as the CNN Web site could give headlines of major stories the highest priority, summaries of these stories the next priority, headlines of minor stories a lower priority still, and so on. You might download headlines to a pocket pager and summaries to a Web-enabled cell phone.
Researchers at IBM's Tokyo Research Laboratory are working on another aspect of this problem: how to generate customized Web pages on the fly. Their
project, known as Dharma (Dynamic generation of HTML Documents with Adaptive Rendering for Micro-Agents), lets Web applications automatically adapt their appearance to different devices, taking into account such variables as the communication speed and screen size of the device and the logical structure of the Web page. Dharma is already used by Japan's Sakura Bank, where it allows users of certain NTT cell phones to check their accounts and conduct transactions on the Web.
Another thorny issue is how to manage so many different computers. It is difficult enough for a company to coordinate a network of desktop computers that stay put. So besides thinking about how data is rendered, the pervasive computing organization is also working on means of managing the groups of users, the software on their devices and the security of those devices.
Bregman and his colleagues believe that the pervasive model -- enabling people to gain immediate access to information and services from anywhere -- is the future of computing, and are doing their best to make IBM a leader in shaping that future. "What I like is that we're in the early stages," says Bregman. "The leaders haven't yet been defined. There is a huge opportunity to help define this emerging market and position IBM as a leader."
Bruce Schechter is a Knight Science Journalism Fellow at MIT.
Organizing in pervasive space
In September 1998, Hurricane Georges, after wreaking a path of destruction across the Caribbean, was headed toward Orlando, Florida. Whenever a disaster threatens a city, hundreds of emergency workers must be called into readiness, bracing for the worst. For most cities, assembling and managing a disaster team is a huge and time-consuming undertaking, but for the City of Orlando all it took was a few keystrokes.
A city worker sat down at a terminal running TeamBuilder, an application built around Lotus Notes® and Domino™, opened a file titled "Hurricane Georges," loaded the hurricane template and pushed a button. Within seconds, dozens of phone, pager and email messages were automatically sent and an emergency response team was assembled. As team members checked in, they each
received a rundown of their duties, drawn from the
TeamBuilder database. And as the incident developed, TeamBuilder kept a detailed record of personnel and equipment used, greatly simplifying later accounting.
TeamBuilder, one of the first applications in "pervasive space," is a novel solution. "It not only leverages proven emergency management methodologies, but it helps
with both planning and execution -- which makes it unique in the industry," says Michael Greenwood, who
initiated the project and led the team that developed
the application.
When IBM approached Orlando with the idea of
automating emergency response, city officials were
intrigued. Since disasters like hurricanes occur infrequently, they felt that specialized management tools would cause problems; people would simply forget how they worked. But since TeamBuilder was constructed on Lotus Notes, it would not necessitate special training.
From his experiences in Orlando, Greenwood began to understand that the same methodologies used to organize a response to a hurricane can be applied to a wide variety of problems, such as contingency planning for Y2K problems, airport management, mergers and acquisitions, news and media. "A tool like this is perfect for helping implement the what-ifs," says Greenwood.
