Share or print this article Print PDF
Stumble Upon Digg Delicious
Control Freaks

Engineers take embedded computing to new levels of precision in their quest for greater efficiency.

Control Freaksaving energy can be as simple as turning off the lights when you leave a room and as elaborate as the network of little machines in a modern control system. The typical automobile is controlled for fuel efficiency at several crucial points, including the fuel injection system, the exhaust, air intake, engine cooling, air conditioning and transmission. Among other things, sensors and computers control the mix of fuel and air, choose the most efficient gear and regulate the flow of engine coolant. Much more elaborate systems control jetliners and spacecraft, for stability and safety as well as efficiency. And the future promises much more of this technology, at higher levels of precision, interaction and complexity.

Researchers at UC Santa Barbara’s Center for Control, Dynamical Systems and Computation (CCDC) have long been working at the frontiers of control engineering. This field draws on the disciplines of mathematics, computer science and several branches of engineering – computer, chemical, electrical and mechanical. Started in 1991 by Petar Kokotovic, now an emeritus professor of electrical and computer engineering, the CCDC functions as an interdisciplinary meeting place for faculty, graduate students and visiting researchers. It’s “an umbrella that keeps us united,” says Andy Teel, a professor of electrical and computer engineering who is on the CCDC faculty and is a former director of the center. (The current director is Mustafa H. Khammash, a professor of mechanical engineering.)

Engine

Work at CCDC goes well beyond cars and fuel efficiency. The center’s research covers control problems of many kinds, from the hardware of MEMS (micro-electro-mechanical systems) to software and computational problems such as the development of algorithms for multi-layered “hybrid” systems such as the control of jetliners and the automation of freeways. But control-system theory is widely applicable. Solving a problem in one area can lead to solutions in others. In Teel’s case, his research (focused on feedback control algorithms for nonlinear and hybrid dynamical systems) has led him to such diverse applications as drug-treatment scheduling for HIV patients, mobile robots, optimization of car engine performance and aerospace work. Control theory can be applied just about everywhere in a technology-driven world.

Teel’s automobile work had its roots in research that Kokotovic was pursuing with the Ford Motor Co. when he retired. Along with a UCSB graduate student, Dobrivoje Popovic, and Ford researchers, Teel sought a way to find the optimal settings for adjustable parameters (such as fuel and air intake, exhaust and spark timing) that affect an engine’s fuel consumption, and to do so across the full range of engine speeds and torque outputs. For this extremely accurate tuning of a running engine, the researchers came up with algorithms based on extremum seeking (ES), a control method suited to dynamic, hard-to-model systems. The object was “to work out a system to map out the characteristics of engines,” Teel says. In practical terms, their research helped point the way to higher fuel efficiency by showing how better to manage the engine’s efficiency-related inputs in real time.

Research as Teel’s responds to the challenge of controlling environments that grow increasingly complicated as more hardware comes into play and more functions are automated. Control theory is not just about managing specific systems, but about deciding which of these systems need to be used at any time and in any situation. Teel points to modern jetliners as an example. They use different controls (autopilot, landing approach systems) for different phases of flight, and are also designed to switch automatically from one control to another, as well as monitoring warning systems, as circumstances change. Essentially, they do as much of the pilot’s decision-making as possible. For these hybrid systems, he says, the point is not so much to solve a particular problem “but trying to figure out what are the right problems to solve.”

“The more things you bring under control, the more you can control them for energy efficiency,” says Smith.

Roy Smith, a member of CCDC and a professor of electrical and computer engineering, is working in a similar vein. His research focuses on cooperative controls in interacting systems, specifically systems controlling spacecraft that detect planets beyond our solar system. He is using mathematical techniques that also can be applied to earthbound systems “that have multiple sensors that need to collectively share information and decide on a course of action.” That’s a good description of the computing embedded in a modern car, and Smith knows that side of automotive technology well. As far back as 1980 he was helping to develop microprocessors for real-time control of automobile engines for efficiency. Later, he worked on emissions control systems for marine engines, autos and trucks.

Smith says fuel-efficiency technology in cars could go further if sensors were deployed at more points in and around the engine – within each cylinder, for instance. The control of the air-fuel ratio now depends on sensors mounted mainly in the fuel injection and manifold, he says, which can only estimate the actual properties of combustion; sensors in the cylinders would measure the combustion directly. “This could buy you a 10% to 15% gain in efficiency,” Smith says – but maybe not without a push from government. The sensors would be costly (they need to have electronics rugged enough to work in high heat), and Smith sees the automotive industry as “very cost-sensitive.”

integrated circuit

Smith is also working with Forrest Brewer, a professor of electrical and computer engineering, on algorithms that enable control systems to use less energy in calculations. For his part, Brewer is developing new architectures for embedded computing to simplify the processing required by MEMS devices (such as sensors) and reducing their power needs. “The control systems themselves are not a big source of energy consumption,” says Smith. “But it would be easier to put those systems in many more places and get more precise control of physical systems” if the control technology is more efficient. And “the more things you bring under control, the more you can control them for energy efficiency.

diabetes