WALNUT CREEK — Without Karl Hedrick, your smart car wouldn’t be so smart after all.
The 70-year-old nonlinear control systems expert operates below the average person’s radar as UC Berkeley’s James Marshall Wells Professor of Mechanical Engineering and director of Berkeley’s Vehicle Dynamics Laboratory. When he’s not roaming the world leading short courses or international transportation committees, speaking about one of his 110 journal publications and co-authored books or accepting industry awards, Hedrick commutes in a 15-year-old Infiniti.
In the Walnut Creek home where he and his wife Carlyle raised their three daughters, Hedrick is on sabbatical, but about to depart for China to give a lecture.
Clearly, the former college tennis player rarely stops moving. “I was good enough to lose to some very good players in three U.S. Open tournaments,” he says, laughing, in an interview. Despite understanding combustion torque like nobody’s business, he insists, “I’m no gearhead.”
But venture into the highest echelons of academia at the University of Michigan, Stanford, MIT, Cambridge and UC Berkeley (all places he’s been a student or instructor) and you will hear substantive praise.
“Karl’s research in the late 1980s and the early 1990s set the foundation for automated (autonomous) driving,” Masayoshi Tomizuka wrote in an email. Tomizuka is associate dean of engineering at Cal, and has worked with Hedrick since 1988. He said communication systems within vehicles, the training of industry leaders and even the national economy have been enormously impacted by his colleague’s work.
Specifically, Hedrick’s four decades of research are arguably the reason those fancy sensors and fly- or drive-by-wire systems — embedded in airplanes, Google’s self-driving car, medical robotics, drones and less-advanced technological wonders like an average late-model car — all manage to operate functionally. His fascination with stability, bifurcation, manifolds, “Popov” criterion and a host of other theorems may be more science than sexy, but no one with a GPS or cruise control is complaining.
Least of all Hedrick, who says, “If it was just math involved, just studying a system, it wouldn’t excite me at all. It’s problem solving — finding real solutions, that’s my work.” After banging around Germany in his 4-cylinder ’67 Porsche 912, playing tennis and contemplating going pro or teaching 4-year-olds the game, Hedrick got his Ph.D. in aeronautics from Stanford. The Department of Transportation and automotive companies were investing in high-speed ground systems research in the 80s, and Hedrick was on solid funding ground. Applying nonlinear control — a term for systems that don’t follow a “double the input, double the output” -type trajectory — Hedrick’s focus moves from high-speed rail to automotive to unmanned vehicles and highway systems.
“I’d be happy to apply nonlinear control to diabetes research, the stock market — any subject,” he said. Technology in development since the 50s, sensors that are cheap and the huge computational power of today’s cars, he said, are having an impact on autonomous vehicle systems all at once.
“But we have to make it work in the real world; in dust storms, nighttime driving. The important thing is, when do we allow the human to not be in charge? I think that’s 20 years from now.”
Hyundai sends researchers to UC Berkeley, and Hedrick recently drove a heavily modified (but not self-driving) Hyundai Genesis for a week. With a camera detecting lane markings, the steering wheel vibrated when he crossed the center line. Adaptive cruise control set the distance between cars, adjusting for changing speeds; side-view cameras issued audible warnings.
“The safety aspect is getting huge,” Hedrick says, “but the user interface isn’t there yet. I drove it for a week and I wasn’t over the curve for setting the controls. It was like trying to find the right station on the radio, I was all over the place.”
Hedrick said creating algorithms for what a human might do two seconds into the future, especially after morning coffee or a dinnertime martini take effect, means collaborating with experts in psychology, artificial intelligence and computer science.
“We have to figure out the best way for humans and robots to work together. A robot is good at finding a bomb, but not at detecting the difference between a bomb and a paper bag,” he says. “Humans don’t like to be told by a robot what to do, so we have to figure out how robots get the mission done, but stay out of the way. Fusing the systems … that will take years.”
—
