WINLAB Looks to the Wireless Future

    by Douglas Dixon

WINLAB
Corporate Partners
Research Challenges
Smart Radios
Open Future

See also Next-Generation Wireless: LTE & WiMAX

We're standing in a big empty room in the WINLAB facility on Route 1 in North Brunswick at the Technology Center of New Jersey (www.winlab.rutgers.edu). Empty, that is, except for the 400 computers hanging from the ceiling.

Suddenly the ceiling powers on, and the room reverberates with the noise from the fans as the computers boot up, and we watch the status lights flicker across a 20 by 20 array of PCs, spaced 1 meter apart across the 80 by 70 foot room.

This grid of PCs is the ORBIT Lab -- the Open-Access Research Testbed for Wireless Networks (www.orbit-lab.org). Each node is a stand-alone PC with two 802.11 a/b/g wireless interfaces (plus some additional connections including Bluetooth). And the activity that we are seeing is from a researcher somewhere over the Internet, who is loading the grid with software to run an experiment. Each ORBIT PC includes a 1 GHz processor, 512 MB of RAM, 20 GB of local disk, two 100BaseT Ethernet ports, and two 802.11 a/b/g cards.

"ORBIT is hugely successful," says Wade Trappe, associate director at WINLAB and associate professor of electrical and computer engineering at Rutgers University. "This facility has 95 percent usage; It gets booted about 30 to 40 times a day. We have about 200 user groups worldwide. There are people logging in from Australia."

WINLAB

WINLAB, the Wireless Information Network Laboratory, is an industry-university cooperative research center for wireless networking, founded at Rutgers University in 1989. It's designed as an international resource for academics, industry, and government to experiment with new wireless networking technology.

        WINLAB Facility

WINLAB moved to its current facility in the Rutgers Technology Center just south of the Cook Campus in July 2005, with approximately 18,000 square feet of space, including the large ORBIT laboratory.

But why wireless experiments in New Jersey? The reason, explains Trappe, was that people that had been researching wireless networks did not have a common basis on which to compare different designs. Therefore, "you would go to a conference with a new mobile network protocol, and the speaker right after you said almost exactly the same thing with a separate protocol, and he beat you and you beat him." It just was too hard to run real experiments: "It's far easier to simulate a network with a thousand nodes than it is to build one."

As a result, says Trappe, "the National Science Foundation (NSF) put a push out for better scientific methodology. They created a program called networking research testbeds. The idea was to find something halfway between building a huge site, and simulating things. In the middle came this effort, at a moderate scale, and as a testbed where people could conduct experiments."

The idea, he says, used the "supercomputer model, a common community resource. You reserve your time, log in remotely, conduct your experiment, get your data, and go back to your lab. You can do the work at 2:00 a.m."

But how can a fixed grid of computers simulate mobile devices? First, the researcher can "warp the grid" by changing the nodes from which the experiment is transmitting and receiving, as if a device was in motion. Second, the testbed includes an array of noise generators that can be programmed to inject noise into the environment. "We can choreograph the noise," says Trappe, "raising the noise floor to change the signal-to-noise ratio and raise the interference, so you can see how this system deals with packet losses."

A similar local collaborative project is PlanetLab at Princeton University, which has some 866 nodes at 458 sites spread across the globe (www.planet-lab.org). "PlanetLab is about long-running Internet-based services," says Trappe. "The timescales we are interested in are much shorter: packet collisions on the scale of milliseconds."

ORBIT was seeded in 2003 by a $5.45 million, four-year grant from NSF. The project is a collaborative effort between several local university research groups: Rutgers, Columbia, and Princeton, along with industrial partners Lucent Bell Labs, IBM Research and Thomson.

The ORBIT testbed is integrated with PlanetLab to enable end-to-end wired plus wireless experiments. WINLAB also has ongoing collaborations with other universities from NJIT, Columbia, and Stevens, to Carnegie Mellon, University of Massachusetts, and UCLA.

Corporate Partners

WINLAB was founded in 1989. Its budget for 2007 was $5 million a year (up from $1.2 million in 2001). It currently has some 25 faculty and staff (from 10 in 2001), including faculty from the electrical and computer engineering and the computer science departments at Rutgers (www.ece.rutgers.edu). The work supports some 45 graduate students.

WINLAB continues to receive some 80 percent of its funding from federal grants, but also has developed a corporate sponsorship program. It is currently working with some 15 industry sponsors, including Intel, Qualcomm, Alcatel-Lucent, and Toyota.

"So far it has been pretty successful," says Trappe. "We average about 10 to 15 sponsors per year." WINLAB offers two levels of sponsorship, full and associate (at $70,000 and $40,000 annual membership fee). The full level, says Trappe, gives the company a sponsored project. "We will work with them to define a project of common mutual agreement. We're not doing development for them; we find something that's good from a research point of view, and which also benefits them. We then map out a year-long project." Students also benefit from support for thesis research and internships at sponsor companies.

While some companies want to control the intellectual property developed through this program, WINLAB prefers to use an open-source model for its software, so that the fruits of research can be shared. "We are an academic organization," says Trappe. "Licensing is too difficult."

WINLAB originally worked with the cellular companies, but now, says Trappe, "things have changed, and we are now finding nonstandard type companies, not wireless companies, coming to us. We still have companies like Intel. But we are increasingly finding companies like Toyota, Alpine, and DaimlerChrysler. What they are facing is that they do not have wireless knowledge, but wireless is key to giving them some differentiation in their market."

For example, he says, "the automotive companies are interested in wireless sensors within the car, wireless communication to roadside beacons, and car-to-car communications." For collision avoidance, two cars could be exchanging information telling each other to adjust the brakes, or signaling that "at this point in the road I sensed there is a low coefficient of friction, so the information that the road is icy is sent to cars behind you."

        Wade Trappe

Describing himself as "a Texan by birth, and an applied mathematician by training," Trappe spends most of his time "in the strange limbo land between mathematics and engineering." He received his undergraduate degree in mathematics from the University of Texas at Austin in 1994, and his masters and then Ph.D. in applied mathematics and scientific computing from the University of Maryland.

While at Maryland, Trappe came to New Jersey as an engineering intern with Dialogic Corp. (a subsidiary of Intel) in Parsippany in 1997 and 1998, and then joined the department of electrical and computer engineering and WINLAB at Rutgers after his graduation in 2002.

Research Challenges

To satisfy the demands for 4G broadband wireless services, "we need to take a fresh look," says Dipankar Raychaudhuri, director of WINLAB and professor of electrical and computer engineering at Rutgers University, and move to a "simplified architecture," built on IP networking.

    See also Next-Generation Wireless: LTE & WiMAX

        Dipankar Raychaudhuri

Raychaudhuri held corporate R&D positions in the telecom/networking industry before coming to Rutgers and WINLAB in 2001. He received his bachelor's degree in electronics and electrical communications from the Indian Institute of Technology, Kharagpur in 1976, and his masters and Ph.D. in electrical engineering from SUNY Stony Brook in 1978 and 1979. He has held executive positions at RCA Laboratories, NEC USA C&C Research Laboratories, and Isospan Wireless, a San Jose company where he was as chief scientist.

To reach this goal for next-generation wireless, Raychaudhuri identified some key challenges that WINLAB is addressing:

- Delivering megabytes per second to each wireless device, and therefore building system capacity to handle gigabytes of bandwidth -- essentially providing today's fiber bandwidth to tomorrow's mobile devices.

- Achieving more efficient use of the available spectrum, while co-existing with shared bands and remaining compatible with legacy equipment.

- Taking advantage of information about the current location of devices to deliver location-aware services (like GPS mapping) and context-aware content to mobile devices.

- And doing all these while still supporting security and privacy in wireless network services.

Raychaudhuri expects the number of wireless devices to continue to explode with the need to connect objects in the environment (sensors, machines, automobiles, etc.) with the network using wireless interfaces. The research at WINLAB has therefore moved beyond cellular telephone to address the demands of this next generation vision of mass-market "pervasive computing" with ubiquitous mobile wireless devices.

Major research areas for WINLAB therefore include:

- Sensor networks -- Providing robust wireless connectivity with widespread use of ultra low-power and low-cost devices. The challenges with these devices include limited processing speed and transmission power, intermittent connectivity, and low speed communications -- they can't talk directly to a cell tower or other common central hub. These issues also are being explored by technologies including ZigBee for industrial remote control (IEEE 802.15.4, www.zigbee.org) and UWB for multimedia (WiMedia Ultra-Wideband, www.wimedia.org).

- Vehicular networks -- Cooperative communication between near-by cars driving on a road, for information, safety (cars are breaking up ahead), and confidence (parking). "One hundred million cars will add Internet connectivity in the next 10 years," says Raychaudhuri.

- "Ad-hoc" networks -- The larger research problem of building these kinds of rugged "infrastructure-less," or peer-to-peer, networks that are self-organizing and work without central coordination points like cell towers or network hubs. These "mesh" networks must be able to dynamically adjust as devices move though an area, adding and removing nodes as they become available, keeping track of the network topology and interconnections between devices, and finding "multi-hop" end-to-end routing connections to reach a distant node by sending a message though several intermediary nodes.

- Wireless security -- The openness of wireless networks also creates research challenges for security, including authenticating messages to detect spoofing and anomalous traffic from imposters, and defending from radio interference attacks by enhancing resistance to jamming and denial of service attacks. This is Trappe's research focus. "When it comes to the security research community," he says, "most of the work is about provably guaranteeing that you have security, whereas the practicality is whether you have good enough security to stop 95 percent of the bad guys." For example, his research group has developed clever techniques to take advantage of low-level information about the radio transmissions to enhance security, by extracting device-specific signatures from the frequency, delay, strength, and other properties of its broadcasts.

Smart Radios

The need for more efficient use of the wireless spectrum drives research into cognitive radio, software-defined radio systems that can reconfigure to communicate on whatever spectrum is available, and with whatever protocol is required.

Raychaudhuri showed examples of how wireless spectrum is currently used in different cities in the U.S., with less that 6 percent of the available spectrum actually occupied. In addition current devices already have multiple radios -- your laptop may have three (Bluetooth, 802.11 WiFi, and cellular data). What is needed is not more radios, but "smart radios," with fast scanning for empty slots of spectrum, agility to use them, and proper etiquette to share with others.

For example, says Trappe, "users like the Department of Defense want mesh type networks with the ability to find an open frequency, that won't interfere with other networks. They're interested in waveform agility, so they can change their communications to interfere less with the indigenous population. The networks can begin to negotiate among themselves."

WINLAB has begun integrating cognitive radio equipment into its ORBIT testbed. Ten GNU Radio boards have already been installed, and the team is developing baseline cognitive radio software so that researchers can experiment with cognitive radio protocols. A new research project will install some 64 nodes, including its own prototype cognitive radio hardware.

"Cognitive radios are not easy for researchers," says Trappe. "You need skills like FPGAs [programmable chips], DSPs [arithmetic signal processors], and protocols [low-level networking software]. Not what an assistant or associate professor could necessarily have their hands on. We have integrated our cognitive radio projects into the ORBIT testbed, so you can just log in and use them."

Open Future

By using open software (based on Linux) and hardware, says Raychaudhuri, the point is to "create an open platform that can be used for experiments and research. ORBIT is an open wireless box. We do not require you to use any particular protocol. It can be programmed by the experimenter, and you can get repeatable results, and it's all accessible over the Internet. You can try out your protocol, your own routing, your own security. It's a clean slate to try out new ideas."

"ORBIT is our contribution to making this vision of the future of the Internet happen. says Raychaudhuri. "It is a very exciting future for networking. Ten years from now you may not recognize many of the protocols and boxes that make up the network. We went through a tough time in the beginning of the 2000's; network and wireless companies were not doing well. It's now a time of a lot of innovation and change for our industry, and for young people it's a great opportunity to get involved. I have only optimistic things to say."

Portions originally published in the U.S.1 Newspaper, June 11, 2008
   www.princetoninfo.com/index.php?option=com_us1more&Itemid=6&key=06-11-2008%20WINLAB&more=1