Creating the Future of Mobile Computing
Like most users, Romit Roy Choudhury enjoys the freedom, information, and options his iPhone and other devices provide, but for him these instruments serve a more important role as research tools that help him envision future innovations in mobile computing. Roy Choudhury's academic life is dedicated to making those visions reality, and it's difficult to discuss his research with him without feeling like you're getting a privileged glimpse of the future.
Roy Choudhury, who was courted by numerous universities as he completed his graduate work at the University of Illionois at Urbana-Champaign, arrived at Duke in 2006 to begin a now rapidly growing wireless networking program. "I think Duke has a fantastic research program that is small but high quality, and that was very attractive to me," he says. Another positive was that he already had established working relationships with some Duke faculty. "When I came to interview here everything just clicked because the people are very friendly, very cooperative, and informal, and I really like that kind of atmosphere."
At the heart of Roy Choudhury's research is a perpetual quest to predict the directions that technology is heading or should head, to define the problems likely involved, then develop the solutions needed. The process involves understanding and forecasting a range of factors, including current technology trends, feasibility, human psychology and needs. "You stand at the horizon of technology and its up to you to decide where the next step should be," he says, "It requires a lot of envisioning and sometimes a lot of daydreaming." Internet connections as easy to come by as electric outlets, even in moving cars, and the ability to tap the collective wisdom of the world's billions of mobile phone users, are just a few of the potential advances he is working toward.
Once problems and goals are defined, Roy Choudhury works to develop the protocols, algorithms, and applications needed to move the field forward, along the way leveraging the advances of other groups, particularly those responsible for enabling hardware innovations.
Roy Choudhury's research balances both theoretical work and implementation of network applications, often crossing boundaries between fields. Accordingly, though his main appointment is in the Electrical and Computer Engineering department at Duke's Pratt School of Engineering, he also holds a secondary appointment in the Computer Science Department. His team of graduate students, drawn from both departments, reflects this diversity, as well as geographical diversity with members from around the world. The group, known as the Systems Networking Research Group, or SyNRG, also includes a number of active undergraduates that conduct their own research.
Roy Choudhury's work has been honored by both academia and industry. He has received fellowships and funding from Vodafone, Motorola, Cisco, and Verizon, to name a few. On the academic side, he received the Best Paper Award at the Personal Wireless Conference in 2003, and more recently the prestigious National Science Foundation CAREER Award in January of 2008. This honor is given to only a handful of junior faculty members "who most effectively integrate research and education within the context of the mission of their organization."
Like Lightbulbs to Spotlights: Optimizing Smart Antennas
One of Chouldury's longest running projects is the development of the algorithms and protocols needed to enlist smart antennas, already used in cell phone towers and other applications, for dramatically increasing the speed and security of mobile internet connections.
The "dumb" antennas in typical devices such as laptops and wireless routers send and receive signals in all directions just as a normal light bulb distributes light. But if instead those signals could be focused like a spotlight, it would dramatically increase connection speeds. As a by-product, security also increases because electronic eavesdropping would require placing yourself in the path of the focused signal. With the standard "light bulb" connection, the signal is broadly distributed and therefore much easier to find and tap.
When using smart antennas on large, hardwired cell towers, there are almost no limits to how much room the equipment can take up and how much power can be used. However, no such freedom is involved when equipping wireless devices with the fantastically innovative, compact smart antennas now becoming available. This means very different protocols and algorithms must be developed for making and maintaining connections without draining batteries too quickly, and this is where Choudury has played and continues to play a major role in advancing the field.
To explore the potential of emerging smart antenna technologies, Choudury, together with colleagues at the Advanced Telecommunications Research Lab in Japan, developed a test bed of laptop computers and receivers with smart antennas. The group ran tests with the computers inside and outside, and while moving and stationary. This and related work has shown that directional antennas can increase speeds by several times over compared to a standard connection, and that there may be potential for even faster speeds. But the work also confirmed or revealed a full range of challenges that Roy Choudhury's group continues to address.
One critical problem is the difficulty in programming computers and routers to reliably and efficiently find each other. When using a broad, dumb antenna, finding other computers to connect to is as easy as turning on an overhead light bulb to see who is in a room. However, the process is more difficult and time consuming if you are sweeping the room using a spotlight with a narrow beam--or a focused smart antenna. The problem becomes even more difficult if there are multiple people in the room swinging spotlights or signals from their antennas and trying to find a single target.
Roy Choudhury's team has also identified a related problem they refer to as deafness, where one smart antenna is facing away from the smart antenna of a computer that is trying to find it. If the seeking computer fails to find a signal, it has no way of knowing if there is an error preventing connection or if its target's antenna is simply pointed the wrong direction.
To deal with the range of problems that complicate initial connection between smart antennas, Roy Choudhury has been working out algorithms and potential protocols that combine the use of smart and dumb antennas. The general scheme is analogous to turning on that overhead light, seeing whose in the room, then turning it off with you and your target then able to easily aim your spotlights at each other. Another possibility they are exploring is using an antenna whose directionality is somewhere between smart and dumb antennas, making the connection process simpler.
The solutions that Roy Choudhury and others devise for these smart antenna challenges will require significant changes to existing, commonly used connection protocols such as TCP and the IEEE standards that govern mobile computing. One route is to modify existing protocols, a simple but limiting option. The other possibility is complete replacement of existing protocols. Various groups are pushing potential solutions, and Roy Choudhury says the process is as much a political game as a scientific one. "It is indeed competitive," says Roy Choudhury, "but certain solutions are being selected and slowly moving into the standards."
The Need for Speed: Coordinating Multiple Connections
Since coming to Duke, Roy Choudhury has been expanding the scope of his research to address a number of problems related to smart antennas. One of these newer interests is called time division multiple access, or TDMA. This technology could solve the problem of multiple people trying to access the same wireless network. Commonly used protocols tend to involve computers competing for the bandwidth they need to transmit. TDMA would instead establish schedules for when various computers would get to transmit, increasing the efficiency of, for instance, a local area network (LAN). Roy Choudhury also conducts theoretical work to determine the maximum speed and efficiency physically possible for networks using smart antennas, with the goal of then working backwards to create protocols that achieve these maximums, instead of trying to maximize the performance of existing networks and protocols.
In related work, Roy Choudhury is also developing the algorithms needed to enable an entirely new way of training different wireless routers to concurrently send packets of data across the internet. Right now if two or more people are connecting through separate routers, those routers have to send packets of data in rotation, which slows the process.
The problem is like trying to listen to multiple people speak at the same time. Most people would ask everyone to be quiet so that we can listen to one person at a time. Current routers are something like those speakers, but new hardware under development offers another possibility. With proper algorithms this equipment should be able to analyze signals being sent by a nearby router to determine how those transmissions interfere with its own. After a period of analysis during which the more standard round robin system might be used, it could begin suppressing that interference. This would allow the different routers to send packets in parallel, a solution something like one of the speakers in the crowd talking louder or in some other way making it possible for the listener to focus on that single voice, tuning out the cacophony of others in the room.
"This can fundamentally improve the capacity and performance of wireless networks," says Roy Choudhury, "because it breaks the bottleneck that exists now with devices in the vicinity of each other."
Mesh Networking: Imagine WiFi Everywhere
One of the most intriguing possibilities potentially enabled by new [smart antennas] is true mobile computing with access to a network almost anywhere. Roy Choudhury envisions wireless connection in the industrialized world becoming in a sense even more ubiquitous than electricity.
One of the most likely technology contenders for enabling this vision is known as mesh networking. The idea is that computers in parking lots, moving cars, or otherwise currently unconnected realms would send and receive data from a network of wireless routers strategically placed throughout an area, for instance on lampposts and rooftops. The critical element is that these points would not have to themselves be connected to the internet. Instead, packets of information would jump from one wireless router to the next until it eventually reaches a wired access point where it makes it onto the internet.
Roy Choudhury has already developed algorithms and protocols needed to drive a mesh network testbed on the Duke campus. Working closely with the Duke IT department, he is planning to equip buses with smart antennas and outfitting certain bus routes with mesh networked wireless routers. This will give bus riders the chance to experience the very first truly mobile network the planet has ever seen.
Even if mesh networks become widespread, there may still be places where you can't connect, but one solution is already under development. The concept is called delay tolerant networks. Wireless devices could be programmed to accept packets of data from other devices and then transmit those packets later that day once a wireless connection was reached. If you needed to send that last email out from a plane, for instance, you might be able to transmit your data to the devices of people on the ground crew, who will soon enter the airport. There, their devices will connect, perhaps getting your email out hours before you could have when you landed in California.
Roy Choudhury's work in this area is focused on developing the algorithms for determining which device or devices your computer should send its information to so that the likelihood of successful transmission is maximized. He is exploring the idea that devices could exchange information in such a way that your device could choose the best candidates.
The Virtual Information Telescope: Realizing the Full Potential of Mobile Phones
Given their popularity, saying that mobile phones will become more useful in coming years would hardly be a revelation. But figuring out just how to tap their potential takes a little more thought, and it's a topic that captivates Roy Choudhury.
He envisions a not too distant future where all of the mobile phones around the world become a virtual information telescope, with the phones acting in concert like a giant lens. The full vision includes a number of components that would enable new ways of gathering information, for instance about where you are, where you're going, or what people are doing in places of interest.
Phones are already equipped with sensors that can provide useful information, including cameras, GPS, and microphones, and next generation devices will add ever more sophisticated sensors, measuring air quality or other parameters. The trick will be figuring out how to amass and access all the information that could potentially be gathered.
At the basic level, volunteer users might provide multimedia information about the places they visit-- photos, video and a recorded message about a beach they enjoyed, for instance, or a restaurant they hated. Such "microblogs" can then be stamped with location information and sent to a central server that organizes the information according to locations on Google Maps. If you're heading to a restaurant or beach, you can go to the map and download what's available.
Microblogging is not new, but Roy Choudhury is working to take the concept to the next level. One idea, which he refers to as "Sticky Notes in the Air", involves virtually floating microblog information at a relevant location. As you pass a restaurant, your phone could automatically download comments from past patrons. Or, while you're in a museum, your phone might download comments from experts about a particular painting you've stopped to admire--all automatically as you pass through a location.
To make the virtual mobile phone telescope even more useful, though, users will have to be able to move beyond just hoping someone has posted the information they want. They'll need to be able to request specific information. With properly developed location technology for phones, someone could send out a request for specific information to a server that would then distribute the request to phones at the relevant location. Various credit and rating systems could be used to encourage people to respond, and to ensure that their responses are accurate. Some questions could be answered without an explicit response. For instance, whether a particular location has WiFi access could be determined by checking for live connections at the location. Similarly, an asthmatic might check to see if the smog is too thick for a visit to Manhattan by requesting information from sensors on phones in the area. The Roy Choudhury team has already created a testbed microblog to explore the Sticky Notes concept with Duke student volunteers.
Many new uses for mobile phones will require efficient, accurate technologies for determining location, an area Roy Choudhury is also exploring. Though many phones already have GPS systems, using them quickly drains batteries, so other options are needed. The SyNRG group is working on advanced algorithms that use statistical analyses to predict location using information from the cell tower sor WiFi access points used. Such localization technologies would be far less energy intensive and so should offer good short-term alternatives to GPS.
Just a few years ago some might have said that no one would ever be willing to be tracked, to answer questions posed by total strangers, or take the time to voluntarily leave information for strangers. However, the dramatic increase in the popularity of social networking sites, along with other signs such as enthusiastic reception of these concepts at key recent conferences, suggests otherwise. It's now clear that at least some of the planet's 2.5 billion mobile subscribers are ready to use and contribute to the virtual information telescope Roy Choudhury is helping create.
Mobile Devices as Medical Tools
Other ideas for the [virtual information telescope] stretch beyond the social and recreational, and Roy Choudhury is collaborating with the Duke Medical School on a number of possibilities. One possibility would be to use locator technologies to track hand washing in a hospital. Failure to wash hands is a major problem in hospitals because it can lead to the spread of potentially life-threatening infections. However, direct monitoring of an individual would create privacy issues reminiscent of George Orwell's Big Brother.
So, SyNRG is exploring a system for monitoring groups as a whole. Mobile phone localization technology could, for instance, enable a system where nurse visits to sinks are tracked and compiled for a report. This could measure overall compliance, as in what percentage of nurses visited the sink after working with a patient, but would include no information about what an individual did.
Another potential medical application would be to have volunteers in global disease trouble spots give permission to be tracked or to have their past movements analyzed. This information could help researchers determine where someone contracted a disease such as the Asian bird flu to better focus eradication efforts.
Undergraduate Motivation: There's Nothing Like Competition
Besides teaching graduate courses on wireless networking and mobile computing, Roy Choudhury also relishes the chance to teach, work with, and motivate undergraduates through courses designed specifically for them. His philosophy is that teaching should be integrated with research and that industry recognition should be used as an incentive for learning. Recognizing that such motivation can not only help students learn better, but also make a class more enjoyable, Roy Choudhury has set up a unique program thanks to initial sponsorship from Cisco Systems, with growing interest from other companies.
Students in a recent undergraduate course took on semester-long projects to develop and implement full-scale protocols. Duke faculty and Cisco researchers then judged these projects, with three winners. "It's a good way to get recognized by the industry," says Roy Choudhury, "and fantastic for the resume."