The unprecedented period of economic growth over the past few years was fueled in large part by the high technology sector. Although the New Economy has cooled some compared to a year ago, the digital revolution continues as innovations emerge on the scene.

Crafting these innovations, however, often takes years of research and development.

"We have numerous cutting-edge information technology projects coming down the pipe," says Raj Acharya, head of the computer science and engineering department. "The work on these innovations isn't just restricted to one department but is spread across many departments and disciplines in the College of Engineering."

Acharya says much of the major research effort is focused in five areas: wireless computing, power consumption, virtual reality, parallel computing, and bioinformatics.

Look Ma, no wires

Taking the country by storm in recent years, wireless technology has manifested itself in cellular telephones, pagers, wireless modems, and personal digital assistants (PDAs).

Experts say, however, that the capabilities available to users now represent merely the tip of the wireless iceberg. Some of the newest devices are even more "connected" than their predecessors. The newest cell phones, for example, offer users the ability to surf the Internet, while the latest wireless PDAs tout capabilities including real-time stock quotes, sports scores, and Internet shopping.

As these devices proliferate and demand for these services grow, wireless access to the Internet may slow or stop.

"The Internet has become more multiservice—services that the Internet was not designed to support," observes George Kesidis, associate professor of electrical engineering and computer science and engineering.

Kesidis is one of many researchers in the College focusing on what he describes as "Internet traffic engineering."

Kesidis's work centers on Bluetooth, a short-range, wireless data communications technology. Bluetooth's advantages, he says, lie in the fact that it is inexpensive, transmits on a public band, and doesn't need to be in direct sight of the transmitter or receiver.

The problem, Kesidis says, is the manner in which data is communicated. Typically, a "master" device talks with "slave" devices using a round-robin polling method. Although this works fine for normal communications, he says bursty traffic such as Internet access doesn't lend itself well to this scheduling scheme.

Kesidis's team is developing an alternative to the round-robin method called flexible polling. The proposed method's main advantage is its ability to handle asynchronous data traffic with "elastic" bandwidth needs by polling heavily trafficked devices more often. Tests performed by the team found that the new method had a higher throughput percentage than the conventional round-robin scheme. They also found savings in power consumption as devices with less traffic were polled less frequently.

Mohsen Kavehrad, professor of electrical engineering and holder of the William L. Weiss Chair in Information and Communications Technology, is also focused on wireless traffic, but of a different nature than Kesidis's Bluetooth work.

He and research associate Svetla Jivkova have been working on creating a wireless local area network that uses infrared light for transmissions. It is low powered but also highly reliable.
The task, however, is easier said than done.

Kavehrad explains there are two ways to transmit wirelessly. "Line-of-sight or point-to-point infrared signal transmission, which is used, for example, in television remote controls, is highly efficient at low power levels but suffers from the need for alignment between the transmitter and receiver. If someone 'shadows' or blocks the remote control beam while you're trying to change the channel, the signal can't get through.

"On the other hand, non-line-of-sight transmissions, which use a broad diffuse beam, suffer less from shadowing but usually forfeit the power efficiency, broadband, and low error rate values that infrared transmissions can offer."

Now, however, Kavehrad and his colleagues at Penn State's Center for Information and Communications Technology Research have developed a new link design that uses a multi-beam transmitter with a narrow field-of-view receiver. The system has a bit-error rate of only one error per billion bits and uses milliwatt transmitted power levels. Kavehrad says, "This error rate is unmatched considering the offered transmission capacity."

To use the Penn State signaling scheme, for example, to form a local area network for a group of computers in a room, each machine is equipped with a low power infrared source and a holographic beam splitter. The original low power beam is separated into several narrow beams, which strike the ceiling and walls at points that form an invisible grid throughout the entire volume of the room. Because the beams are also reflected at each of the strike points, they can be used to send or receive information.

Since the beams created by the splitter are narrow, narrow field-of-view receivers are used. Using a narrow field-of-view receiver makes it easier to filter out noise. In addition, receivers consisting of more than one element can ensure continued coverage when some of the transmitter's beams are blocked.

Traffic, however, isn't solely restricted to how information is transmitted between devices. Devices trying to access the Internet or databases may also run into traffic.

Ali Hurson, professor of computer science and engineering, says problems arise when a device of limited processing capabilities and resources tries to sift through massive amounts of data through a wireless connection. Hurson's solution to the problem is the concept of the mobile data access system (MDAS). Using wired and wireless connections, MDAS can access heterogeneous data sources such as news, weather, stock information, and the World Wide Web.

This is accomplished by superimposing a multi-database system over a wireless-mobile environment. Working like a super-search engine, the multi-database gives users access to multiple databases with a single query and integrates the results.

The beauty of MDAS, Hurson says, is its ability to manage query traffic so that gridlock doesn't happen when more than one user wishes to access the same database. Using a concurrency control algorithm called V-lock, two users can access different parts of the same database. Hurson explains V-lock uses semantic information to examine a query and makes an "educated guess" when it directs traffic.

Power plays

But network traffic isn't the only thing working against handheld wireless devices.

"If a device is mobile, then it has a limited energy supply, a limited input/output capacity, and limited form factor," states Mary Jane Irwin, distinguished professor of computer science and engineering.

A research team including Irwin is examining ways that embedded devices can be designed more efficiently. Embedded systems have, as the name implies, everything they need built into them. Irwin says these systems not only include laptop computers, cell phones, and PDAs, but also home appliances such as microwaves, consumer electronics such as VCRs, and micro-controllers such as in cars and the space shuttle.

The team consists of Vijaykrishnan Narayanan, assistant professor of computer science and engineering; Mahmut Kandemir, assistant professor of computer science and engineering; Anand Sivasubramaniam, assistant professor of computer science and engineering; and Irwin. Each team member brings to bear a different area of expertise in embedded systems, including hardware, software, code compiling, and operating systems.

"We're looking at power consumption at all levels in order to optimize the system as a whole," Irwin explains.

The team has already created an application called SimplePower that allows designers to quickly and accurately estimate energy consumption in both hardware and software.

The hope for the project, Irwin says, is to get designers to be more energy aware as they put future devices on the drawing board.

Sight beyond sight

Another area of research in the College receiving a great deal of visibility (pun intended) is computer vision.

Rangachar Kasturi, professor of computer science and engineering, and Rajeev Sharma, associate professor of computer science and engineering, are developing image processing methods for computer vision.

Experts believe computer vision and pattern recognition will play a vital role in the security and intelligence fields. Kasturi says proposed technologies such as fingerprint recognition, face recognition, and retinal scans all rely on computer vision. He says the University's Computer Vision Laboratory is working on video analysis systems to help intelligence agencies extract more information from video intelligence.

Using computer vision, an engineering research team consisting of Lee Coraor, associate professor of computer science and engineering; Octavia Camps, associate professor of electrical engineering and computer science and engineering; and Kasturi, has built an on-board flight system to warn pilots about impending mid-air collisions. Funded by NASA, the system was successfully tested on an Air Force test aircraft.

"The work is being extended to warn helicopter pilots about potential hazards such as electric lines during low altitude flights," Kasturi states. "This technology might also enhance helicopter flight safety during search and rescue operations."

Sharma has used computer vision technology for other applications, developing an interactive computer map called iMAP. The iMAP uses computer vision to recognize gestures and combines it with speech recognition software to create a new interface people can use.

Interestingly, Sharma says the idea for gesture recognition came from watching forecasters on the Weather Channel and how they interacted with the television map.

Creating this speech/gesture interface gives people a new method of accessing information.

"You're trying to access the database behind the map," Sharma explains. "You say something like, 'Show me my dorm,' and it'll display all the dorms. Then you point and say, 'This one,' and it'll go to that one."

Sharma has already improved upon the original iMAP technology, creating a new system called eOz.tv. The system was recently set up in the HUB to promote the fall career fair and for passersby to play with.

"The technology has evolved to become more robust and stable," says Sharma of his new creation. "It immerses the user in the interface."

Unlike the iMAP, eOz.tv doesn't just track the user's hand—it tracks the user's entire body. People using the eOz.tv system step in a circle on the ground, and the computer tracks the person. The new system uses improved computer vision and deals better with environments that are crowded or poorly lit.

Sharma describes eOz.tv as an "infotainment center." Students who used the system in the HUB could not only get information on some of the companies recruiting at the Bryce Jordan Center but also play games.

"This new system is very immersive as well as interactive," he says.

Virtual reality exists elsewhere on campus, too.

At various locations throughout the University are FakeSpace RAVE systems, eight-foot by eight-foot monitor "walls" connected to powerful computer workstations. Users don goggles to create 3-D illusions out of the large displays' stereoscopic images.

But you won't find people using these systems playing games. The systems are devoted entirely to research and can be connected to the ultra high speed Internet2, allowing for collaboration between colleagues across the country.

In addition to the goggles, RAVE users are equipped with a wand, which acts like a three-dimensional mouse," explains Paul Plassmann, assistant professor of computer science and engineering. "With the wand you can poke data, change conditions and parameters, and see what happens in real time."

Until the advent of virtual reality technology, researchers could only perform complex computations on clusters of PCs. Systems like the RAVE allow them to not only compute data, but visualize and analyze results, says Lyle Long, professor of aerospace engineering. RAVEs let people watch complex data unfold in 3-D or 4-D (3-D plus time) simulations.

"You can tweak the data and the parameters that control the experiment," Plassmann says. "You learn by doing things. You build intuition and get a better understanding."

Parallel potential

With or without virtual reality, parallel computing remains a potent tool for researchers. The College already offers a graduate minor in high speed computing, and one faculty member in the College is trying to make parallel computing even faster than the mind-boggling speed of today's machines.

"Everything you can do, you want to do it faster," says Padma Raghavan, associate professor of computer science and engineering. She says the need for speed isn't a frivolous endeavor.

"Almost all simulations are done on computers. Therefore, the more power you can get, the better the simulations," she explains. "There are some things that can't be solved today."

For example, Raghavan says we don't have the ability to accurately model how our nuclear weapons stockpile deteriorates over time.

Her own work is looking at simulating nanotechnology and whether it can be effectively used in future computers.

"Can you make a transition junction out of carbon? Can you take fifty atoms and make them into a ‘T' and not have it disintegrate? Will it have the properties of a transistor?" Raghavan asks. Because fabrication is extremely expensive, she says it's crucial that the simulations answer these questions.

"Once we can prove these ‘T' junctions work and mass produce them, they'll be much closer to reality than quantum computing or biocomputing," she says.

Potential applications for this new technology include language processing, multimedia processing, and weather modeling, Raghavan says.

DNA and CPUs

Research in computer science and engineering isn't limited to commercial applications or Internet use, however. Engineers are also using computers to unravel the genetic mystery behind humans and animals.

This field, called bioinformatics, utilizes computer technology to compare the genomic DNA sequences between two mammals.

Webb Miller, professor of computer science and engineering, is focusing his efforts on comparing the human genome with that of mice.

"Humans and mice have the same number of genes within one percent of each other," he says. "Lots of people will be surprised by how large regions of the human genome are identical to the mouse genome. There are regions in the two that are strikingly similar, but we don't know what they do."

Miller says that the process of evolution doesn't change all of the DNA in an organism at once. "Evolution doesn't impact all areas of the genome as uniformly as we thought. It's more complicated than anyone thinks," he says.

He goes on to explain, "The DNA sequence that doesn't change is like a little red flag that says, ‘This piece is doing something useful for the organism.' Because it's doing something useful, it'll be resisted to evolutionary change."

The trick, Miller says, is to isolate the regions of the genome that have not changed during the course of evolution. Using software he developed, he is comparing the genomes of humans and mice in hopes of finding these unchanged regions.

There is a payoff in comparing humans to mice. "If you can find the corresponding genome in the mouse, then you can start doing experiments."

Ain't seen nothin' yet

Acharya says the innovative work happening in computer science and engineering is merely the beginning. In 2003, the department is slated to pack its bags and move into the new IST Building (for more on the building, click here).

"We're looking at new facilities and, more importantly, new opportunities to collaborate with the faculty in the School of Information Sciences and Technology," he says. "Then you'll see some remarkable innovations!"

—Curtis Chan, Barbara Hale, A'ndrea Messer, and Bridget O'Brien