The Next Frontiers of Knowledge

School of Management Researchers Forge Ahead in Fields Ranging from 
Robotics to Breast Cancer Diagnosis

By Paula Felps

In today's fast-moving business environment, working on the edge of change 
often requires crossing borderlines that separate the present from the 
future. UTD School of Management (SOM) faculty researchers operate in 
these regions, redefining the boundaries of technology, pushing into New 
frontiers by combining today's innovations with tomorrow's visions. 
Putting a human touch into high- tech tools, the work of these SOM 
professors is helping redefine everything from how breast cancer is 
diagnosed to the way telecommunications networks are designed.

Improving Efficiency Through Robotics

After applying his mathematical scheduling theories to the development of 
robotic systems in Toronto and to the improved efficiency of manufacturing 
and service systems in Hong Kong and Grenoble, Chelliah Sriskandarajah is 
now in Dallas, blazing new trails in maximizing manufacturing productivity 
for the semiconductor industry.

Dr. Sriskandarajah, or "Sris" as his students call him, is associate 
professor in SOM's management information systems and operations 
management areas. His groundbreaking research has pioneered ways to 
achieve maximum efficiency in the simultaneous scheduling of jobs and 
sequencing of robot movements in robotic cells. He defines a robotic cell 
as a kind of family of machines served by robots, which all work together 
in a system controlled by computers that coordinate all the robots' 
movements.

"This is a fully automated machine system that is very precise," Dr. 
Sriskandarajah explains. "Each robot has to be directed toward a series of 
manufacturing steps and movement of parts. The number of combinations of 
possible movements by the robots within the cell is exponentially 
explosive, intractable even by computer. My research centers on developing 
optimization algorithms that can be used to discover the combination that 
gives the best throughput, or rate of production, for a given robotic 
cell."

Over the years, Dr. Sriskandarajah's research has drawn honors and notice 
from other scholars in his field. In 1999, he received a research award 
from the Natural Science and Engineering Research Council of Canada, and 
his 1992 research paper on ways to maximize efficiency within robotic 
cells has generated more than fifteen other papers and three theses from 
scholars in the field.

Currently, Dr. Sriskandarajah is working with Allen, Texas-based FSI 
International, a company that makes robotic cells, which are, in turn, 
employed in the manufacture of wafers used in the semiconductor 
manufacturing process.

Working to develop algorithms that can be used to increase the level of 
automation in manufacturing industries, especially in engineering and 
electronics, Dr. Sriskandarajah notes that his theories play a critical 
part in helping the United States compete in global markets.

"Our research in The UTD School of Management's PhD program is an 
important part of manpower development," he explains. "It plays an 
important role in the planning, designing, and implementing of advanced 
manufacturing systems to keep U.S. industry competitive."

Improving Breast Cancer and Other Diagnoses Through Isotonic Separation 

Research being conducted by Professors R. Chandrasekaran, Sungchul Hong, 
Varghese Jacob, and Young U. Ryu could make important strides in the 
diagnosis of breast cancer and other diseases. The SOM research team is 
looking at how physicians might use an isotonic separation technique - a 
method developed by the team to allow for more accurate classification of 
data.

"Techniques such as this can be used in situations in which classification 
is of critical importance, for example in medical diagnoses," observes Dr. 
Jacob. "In the medical field, typically a doctor collects information on 
symptoms from a patient and then makes a judgment on the patient's 
ailment. Since this data is typically stored, one can use the information 
to develop diagnostic systems that could possibly aid doctors in their 
evaluations of patients."

Noting that isotonic separation relies on historical data to help it 
classify new data, Dr. Ryu explains the SOM research team is using the 
isotonic separation technique to classify data on breast cancer symptoms. 
At this point in the project, the team's research seems to confirm that 
the isotonic separation technique can be applied to classify newly 
gathered medical data.

As Dr. Jacobs explains it, isotonicity addresses classification problems 
by establishing boundaries between groups of data. "A classification 
problem requires one to use observed information to classify an object 
into a category. For example, in the case of breast cancer, if one puts 
all the symptoms into classes, these symptoms either classify the patient 
as having breast cancer or not having this disease.

"Physicians usually make their diagnoses based on the presence or absence 
of certain symptoms and on certain measurements that are either too high 
or too low, outside the normal range. The isotonic separation technique 
uses the historical data to develop a boundary between the data that 
resulted in breast cancer and the data that did not. Once the boundary has 
been determined, the technique allows new data to be classified into one 
of the two categories," Dr. Jacob explains.

The technique could potentially be used to enhance diagnostic capabilities 
of doctors by allowing them to make their own diagnosis based on patients' 
symptoms, while at the same time, having patient data analyzed through a 
computer system that implemented the isotonic separation algorithm. If the 
two methods yield differing results, the doctor then may choose to pursue 
additional testing to discover the root of the medical problem. However, 
if the two separate methods come to the same conclusion, the patient has 
received a second opinion in a more expedient and less expensive manner.

Creating Hybrid Fiber Coaxial Cable Network Designs 

With an ever-increasing demand for broadband networks, telecommunications 
companies are looking for more cost-effective ways to provide more 
services at lower costs. In their research paper on Hybrid Fiber Coaxial 
(HFC) CATV network designs, Hasan Pirkul, Caruth Professor of Management 
Information Systems and dean of The School of Management, and Rakesh 
Gupta, assistant professor in the management science and information 
systems area, devised a way to optimize the cost of designing hybrid 
networks.

Hybrid HFC networks combine coaxial cable, used in older cable television 
networks, with fiber optic cable, which has a much higher data-carrying 
capacity, to deliver voice, video, and data to consumers. The research 
details how service providers can design high-capacity cable network 
connections, using the combination of fiber and coaxial cable, at a lower 
cost than in the past.

"Everyone now is asking, 'How do I design my network for the least amount 
of money?' It's something that so many companies are having to look at," 
Dr. Gupta says. "This research provides some of those solutions."

Their findings, published in the May 2000 issue of the European Journal of 
Operational Research, are applicable to large networks while still 
maintaining the networks' high quality of service. Where that information 
goes now remains to be seen, although it is being eyed by major cable 
providers.

"We've completed the technique and published the paper, and now it's 
available for others to use," Dr. Gupta says. "Presumably, companies who 
use these types of networks will be interested in what it can offer."

Management of Hybrid Fiber and Coaxial Networks 

Picking up where Dr. Pirkul and Dr. Gupta's research left off, Dr. Ray 
Patterson, assistant professor in the management science and information 
systems area, worked with Professor Erik Rolland from the A. Gary Anderson 
School of Management at the University of California to take that research 
to the next level. Their research, to be published later this year in the 
journal Operations Research introduces a new process called Adaptive 
Reasoning Technique (ART) for general problem solving.

"It's important for us to find low-cost solutions for broadband access," 
Dr. Patterson explains. "Companies like AT&T, @Home, and others in the 
industry are currently spending vast amounts of money to lay cable to 
bring broadband access to people's homes. [Improving methods] to route 
cables was a big step, and that's what Rakesh Gupta and Hasan Pirkul did 
with their research."

Dr. Patterson's ART approach to problem solving relies on the combination 
of memory and trial and error to learn from past mistakes, rather than 
trying to create new problem-solving techniques by adding more problem 
specific facts to the mix. In technology, that means the computer is fed a 
problem, then learns about different solutions until it is able to hone in 
on the solution most applicable to the individual situation. Dr. Patterson 
says ART can be applied to a variety of problems, even those not in the 
telecommunications realm.

"The management of networks is a hot topic, but the real contribution that 
came out of this research is the development of a heuristic technique 
using the adaptive reasoning technique to find optimal, or near-optimal, 
answers for any large or complex problem," he says. "We're very encouraged 
by it, because it can be used any place you have mathematically 
intractable problems."

Targeting Markets in the Software Industry 

The use of different editions of a certain product is seen in virtually 
every market, from "light" incarnations of our favorite foods and 
beverages to special editions of automobiles. Low-end and higher-cost 
versions of similar products allow manufacturers to target different 
consumers in the global marketplace. In recent years, such target 
manufacturing, called segmentation, has been used in the packaged software 
industry as well, but new research conducted by UTD's Srinivasan 
Raghunathan indicates that it's not always the best option for companies 
or customers.

Dr. Raghunathan, assistant professor in the management science and 
information systems area, has researched the use of segmentation in the 
packaged software market and found that the software distribution arena 
differs from most markets, where segmentation has proved highly 
successful. His research points out that the software market is 
distinguished from others in two very specific ways: software is extremely 
cost-effective to produce, and it also has the built-in option to offer 
upgrades. Therefore, trying to introduce different versions of the 
software doesn't always yield the same benefits as introducing new 
versions of other products. When the spread between high-end and low-end 
segments is small, such as is found with word-processing software, it is 
better to introduce a single version of the software, his research shows.

"On the contrary, in the case of specialized software, such as database 
and Web server, the difference between the segments' valuations is so high 
that the software manufacturer will be better off splitting the software 
into multiple editions and marketing a different edition for each 
segment," Dr. Raghunathan explains.

He also has discovered that vendors can earn higher profits simply by 
announcing their strategies for the future. Dr. Raghunathan's findings 
suggest that an announcement gives consumers more confidence in the 
vendor's future, and they will make purchases accordingly.

Knowledge Sharing in Organizations 

Finding a way to harness and utilize the vast range of employees' 
knowledge is a challenge that faces most companies today. Throughout the 
world, organizations are investing significant amounts of money on 
knowledge management efforts in hopes of better corralling the diverse 
range of intellect that is available through participants.

Sumit Sarkar, SOM associate professor of management information systems, 
and Sury Ravindran, assistant professor of management information systems, 
are working on a research project that provides ways for a company to 
encourage information sharing among its workers. Among the elements the 
researchers are grappling with is a way of gauging available knowledge and 
creating incentives for sharing that knowledge.

"There are alternatives that a company has available to motivate 
individuals to participate in knowledge sharing," Dr. Ravindran explains. 
"What we're looking at with this research is the best way to do that. If 
there are experts in a certain area, and you also have a lot of rookies, 
there are a couple of alternatives companies can look at."

The first alternative is for companies to take a "survival of the fittest" 
approach, giving more knowledge and power to those who already have it, 
and eliminating the weaker, less-knowledgeable players. That may build a 
strong organization initially, but it is rather shortsighted for companies 
seeking long-term goals, since the knowledge won't be cultivated in those 
who are rising through the ranks.

"They may also be missing out on some alternative knowledge that the 
weaker ones have," Dr. Ravindran acknowledges. "So perhaps a better 
approach is to make the weaker ones stronger, and then all the people in 
the organization become an expert in their domain."

Using an analytical modeling approach, Dr. Ravindran and Dr. Sarkar 
have derived conditions under which the different approaches would be 
effective. They are working on empirically validating their findings.