T. John Koo, assistant professor of computer engineering, has been recognized with a CAREER award from the National Science Foundation (NSF).
The Faculty Early Career Development awards are considered NSF’s most prestigious honor for junior faculty members. Koo will receive $400,000 over five years to support his efforts to pioneer a new approach that will help engineers do a better job of designing the wide array of “smart devices” which contain microchips that are spreading rapidly throughout modern technology.
According to NSF, CAREER awards support exceptionally promising college and university junior faculty who are committed to the integration of research and education. The awards recognize faculty members, new in their careers, who are most likely to become the academic leaders of the 21st century.
Not counting personal computers with their printers and peripherals, the average household already contains some 40 to 50 tiny smart devices, and experts predict that this number could grow tenfold in the next decade or two. Today, microchips are embedded in cell phones, washers and dryers, refrigerators and microwave ovens, televisions and stereos, remote controllers and garage-door openers.
These “information appliances” bring many obvious benefits: Without them phones would all still be tethered to wires, dryers wouldn’t turn off automatically when the clothes are dry; CDs and DVDs wouldn’t exist; and people would still have to get out of the car to open the garage door. In the case of automobiles, which come loaded with as many as 100 microchips in high-end models, they improve fuel efficiency, help maintain control of the car in dangerous conditions and deploy airbags to protect the occupants. Not only the home and the automobile are rapidly becoming dependent on such devices but they are spreading through aviation, medicine, manufacturing, the military – virtually every corner of society.
But the trend also has a dark side. The marriage of digital processors and sensors with mechanical systems—what engineers call “embedded hybrid systems”—produces devices that are much more complex than the purely mechanical or electrical systems that they are replacing. In addition, they interact with their environment and each other to a great extent. So there are more things that can go wrong. Consequently, hybrid systems are considerably more difficult to design and test.
That is where Koo comes in. “Traditionally, we have had mechanical engineers and physicists focus on the analog aspect of a design, and we have had electrical engineers and computer scientists who focus on its discrete, digital aspects,” Koo says. “But now, since these things are coupling together, we need to look at the interaction between them more closely, and we need a new system science to do that.”
To help provide a basis for this new science, Koo is combining two mathematical methods – “muti-resolution analysis” and “level set methods” – to improve the design of hybrid systems. Multi-resolution analysis can create a mathematical model of a hybrid design that is detailed enough to be predictive but efficient enough to run on standard workstations. The level set method predicts how the model will evolve, including how it will react in varying environmental conditions. When combined, the two techniques should have the capability to predict how a system will perform under a wide variety of conditions.
Current simulation techniques can predict how a cell phone design, for example, will perform under specific combinations of important factors, such as heat, humidity, battery charge, signal strength and distortion. Koo’s approach, by contrast, holds the promise of predicting how a design will work in all combinations of these variables, making it much easier to pinpoint situations where the design is most likely to fail.
To double-check the accuracy of the design tools that he is developing, Koo is creating a “test bed” that allows him to closely monitor the behavior of real-life hybrid systems and compare it to his software’s predictions. The two systems he has selected for this purpose are a reconfigurable electronic power circuit and a multi-vehicle control system. The power circuit is a DC-to-DC converter, a component that is used in a wide variety of electronic devices to match the voltage coming from the power source to that which is needed by the electronic components. The vehicles he is using are model helicopters, which are among the most difficult vehicles to control autonomously.
“My hope is that these tools will help us to finally realize the dream of safe and reliable autonomous control systems for cars and other vehicles, smart structures that can respond dynamically to earthquakes and severe storms and advanced biomedical devices,” Koo says.