CMU gets $10 million for complex system study
Electronic and computer systems are similar to human bodies. Both are complex systems with many variables that must work for the system to function properly.
With that perspective in mind, a research group based at Carnegie Mellon University plans to develop computer modeling programs to analyze pancreatic cancer and atrial fibrillation as well as aerospace and automotive control systems to understand how they work and what goes awry when they don't.
The National Science Foundation has awarded CMU and a group of research universities a five-year $10 million Expeditions in Computing grant to launch IMDECS -- the Institute for Model Discovery and Exploration of Complex Systems. It will be based at CMU's new Gates Center for Computer Science.
Edmund M. Clarke -- the 2007 winner of the Turing Award, the computer science equivalent of the Nobel Prize -- is the project's principal investigator and also serves as institute director. He has assembled a research team of 19 all-star computer scientists and medical researchers with multidisciplinary skills.
The institute will use new computer modeling and analysis systems to probe the cause of pancreatic cancer and the common heart-rhythm problem, atrial fibrillation. It also will study embedded computer systems critical to safe operation of aircraft and automobiles.
Dr. Clarke said biological, aerospace and automotive systems can be reduced in similar fashion to abstract models conducive to computer analysis.
"If we could take a look at flight control systems for commercial aircraft and find errors and fix them, that would be a success," he said. "If we find a chemical reaction in the pancreas that causes pancreatic cancer, that would be very rewarding."
The goal, he said, is to produce useful tools within the five-year grant period.
Dr. Clarke is co-inventor of Model Checking, the most widely used technique for detecting and diagnosing errors in complex hardware and software design. It considers every possible situation to determine whether it's consistent with design specifications. Such analysis can reveal inconsistencies that can cause malfunctions. But MC is limited by the size of systems it can analyze.
Abstract Interpretation, developed by Patrick Cousot of New York University and an institute team member, doesn't attempt to analyze the entire system, but instead develops a simplified approximation of the system before analyzing it. AI can be used to analyze large, complex systems, such as the million lines of software code in Airbus A380's primary flight control system. But it has less precision than Model Checking.
The institute will attempt to combine the systems into one system, MCAI 2.0.
"What's innovative about this is the claim that the same technique can be used for the cellular level as is true for the embedded [computer] system," Dr. Clarke said.
One goal will be to model pancreatic cancer, the fourth-leading cause of cancer deaths in the United States and Europe, to determine how the cancer develops and how to detect it at a treatable stage.
Modeling also could enable physicians to predict the onset of atrial fibrillation, which is responsible for 15 to 20 percent of all strokes.
Embedded computer systems in cars currently lack any real means to interact with each other. The research project could help figure out how best to integrate those systems to work more harmoniously together.
The size and complexity of embedded software technology in the aerospace industry pose similar problems.
"Professor Clarke has truly assembled a dream team for this important new initiative," CMU President Jared L. Cohon said. "Computational modeling and simulation have become critical to discoveries in almost every scientific discipline, so finding new ways to build and explore these models will pay research dividends for years to come."
First Published August 19, 2009 12:00 am