Researchers reengineer the wireless internet

University lab focuses on the toughest wireless networking problems

When you talk to the researchers at Rutgers University's Wireless Information Network Laboratory, you wonder how anyone actually communicates with radios. Or would want to.

That's because WINLAB is focused on research to solve some of the toughest problems facing large-scale wireless nets, from radio coexistence in crowded frequencies to re-engineering the internet for mobile traffic. Research efforts like those of WINLAB eventually will help enterprises deploy multiple radio technologies and enable their own mobile users to exploit new services delivered wirelessly.

WINLAB's reputation has been growing since its founding in 1989. The original focus, pioneering at the time, was focusing on digital CDMA radio resource management and power control, at a time "when a cell phone was a rare thing," says Director Dipankar Raychaudhuri. Much of that research is a basic feature in today's CDMA cell nets.

Now WINLAB focuses its research in three broad areas that together envision a world that is bound together by short- and long-range wireless nets, intelligent devices and sensors, and people on the go who increasingly will depend on these. The areas are the mobile internet, which will revamp today's internet architecture and protocols to include mobile users and wireless links; cognitive radios, which can seek out and use any available frequency; and "pervasive wireless," or embedded sensor nets wirelessly linked with Internet-based services.

It's kind of a like a radio think tank . . . with toys, from their own cellular base stations to custom-built programmable radios that can change their personalities to run on different wireless nets. And what they mostly think about here are all the stumbling blocks to realising this wireless future. You can get a flavour of their thinking from the titles of the 2007 theses from WINLAB's graduate students, such as "On the scalability of ad hoc networks" and "Cross layer network architecture for efficient packet forwarding in wireless networks."

Investing in future nets

Lots of folks are investing lots of money in overcoming those blocks. About 60% of WINLAB's roughly US$5 million (NZ$6.3 million) yearly budget (not counting faculty salaries) comes from federal and state research grants; in other words, your tax dollars. The remainder comes from company sponsors, currently about 15 of them. Some sponsors contribute to get early access to WINLAB research results, negotiating with Rutgers for intellectual property rights if the sponsor wants to use them.

Some sponsors actually get involved in the exploratory research, which has to satisfy WINLAB interests and requirements. "These projects have to involve [academic research] theses," says Ivan Seskar, associate director information technology for WINLAB. "It has to be research that we also are interested in."

The research is overseen by 25 to 30 faculty and staff, with most of the faculty being full-time Rutgers professors and all specialising in some area of wireless technology. About 50 PhD students at any time are working on the various research projects.

During the mid-'90s, WINLAB began looking at wireless technologies besides CDMA cellular. Raychaudhuri first came on board to oversee the Lab's wireless LAN research. After being named Lab director, he led the development of a new research agenda that focused on the complex, emerging world of distributed, pervasive, interlinked and co-existing wireless ecosystems. "We wanted to change the vertically oriented cellular architecture into a flat architecture, like the internet," he says.

That, as it turns out, is a tall order. WINLAB is only one of many academic research teams, laboratories and projects involved in this growing effort, but it plays an important role. The National Science Foundation's (NSF) Future Internet Design (FIND) programme is exploring "revolutionary architectures" for the internet, while a related programme, the Global Environment for Network Innovations (GENI), will create a large-scale network to experiment with the new protocols, services and technologies needed for such architectures. Wireless is a key element in both programmes, and WINLAB a key element in thinking about wireless in this context.

"The notion is that the internet needs to transform itself, so that a car [with an embedded radio], for example, can act as a radio router," says Raychaudhuri. "Today's internet doesn't recognise this: It's a scenario that's too dynamic for existing [internet] protocols."


In positioning itself to participate in this re-conceptualising of the internet, WINLAB launched in late 2005 the Open Access Radio Grid Testbed (ORBIT), which houses 400 programmable radio transceivers, each with multiple radio interfaces. They look like an army of helmeted yellow crabs hanging from a ceiling.

Funded by US$5.4 million over four years from the NSF, ORBIT lets researchers simulate, test and retest how experimental protocols perform in varied but controlled conditions. There's even an outdoor test range, mainly working with wireless in vehicles. Until now, says Seskar, who oversees ORBIT, there has been no common ground in simulating and testing new protocols. ORBIT creates a level playing field to advance wireless protocol design, which is a vital part of both the FIND and GENI projects.

Today's wireless nets often use custom protocols and rely on gateways to the internet. Cellular handsets and base stations introduce still more protocol layers. "With a flat architecture, you can eliminate all these and have the radios talk directly to IP," says Raychaudhuri. "We want devices talking to each other through just one or two layers of protocols, as the wired Internet does. That's a very important step in achieving our vision of a wireless world."

ORBIT can facilitate the creation of an internet that can handle the broken links and impaired connections that are common in wireless nets. "If your network is disconnected for a just a few seconds, the IP addresses are lost," Raychaudhuri says. "We need to be able to work with disconnects." IP protocols suffer heavily if radio links are impaired, he says.

Raychaudhuri says WINLAB forsees a future in which there is a "unified radio API," a standard interface that will allow any radio, whether wi-fi, WiMax or others, to simply plug into the re-architected internet. "Intel can create chips with both WiMax and wi-fi [capabilities] and they will just pick what [one] they want," he says.

Radios that think, and pervasive wireless

That freedom of radio choice also hinges on another major WINLAB research focus: cognitive radios, sometimes called software-defined radios. The goal is hardware that can be programmed to adopt different radio "personalities," using different media access control (MAC) layers.

The stakes are high because licensed spectrum is a scarce resource, Seskar says. If radios can find unused spectrum anywhere in numerous frequency bands, and avoid bands and channels currently in use, the available spectrum can be used much more efficiently. "In most areas, no one is using all of the broadcast TV spectrum, for example," he says. "Cognitive radios could see what's going on, and adjust accordingly to use it."

The final major focus for WINLAB is pervasive wireless sensor networks, which pose a special set of challenges, says Raychaudhuri. Unlike centrally planned and controlled cellular nets, created by a single vendor, "pervasive wireless is just a 'deployment that happens'," he says. "You have to deal with dynamic connections and disconnections." And often with a dense web of radios, sometimes in a small area, sometimes spread out over much larger ones. Recently, WINLAB projects have begun exploring the higher layers of the protocol stack, the applications that run above the MAC and physical layers of these nets.

The exploration fits with what Seskar says is a WINLAB mentality: a focus on the larger system. "Most researchers get bogged down in particular details," he says. "For example, in creating a wireless card that can do 400Mbit/s. But what if only one of these can be operational [in an area] at a time? You need to wrap what you're working on into the larger picture."

Join the newsletter!

Error: Please check your email address.

Tags Rutgers UniversityNetworking & Telecomms IDwinlabivan seskar

Show Comments