Source: Carnegie Mellon Engineering News: Spring 2003, p. 12-13

GM’s new $8 million research grant will help to put information technology into future vehicles in ways that could change the industry.

Perhaps you already think of your car as a “companion.” But when Ed Schlesinger uses the term, he’s talking about really bringing that vehicle to life with information technology--giving you a car that will know you and help you, inform and entertain you, and even take care of itself like a proper traveling companion.

That is the goal of the General Motors Collaborative Research Lab at Carnegie Mellon. Schlesinger, a professor of Electrical and Computer Engineering, co-directs the lab, which was launched three years ago with a GM grant of $3 million and is now expanding: GM has committed $8 million for the next five years.

Projects across campus range from research on deep enabling technologies to the building of prototype systems. Drawing from work now under way, here’s a hypothetical look at how your drive to a meeting in a nearby city might go in the “companion” car of the not-too-distant future: As you slide behind the wheel, the car recognizes you. The FaceCam system, mounted in the steering column, tells the seat, mirrors and dash display to reconfigure from the settings your spouse or teenager used the night before. The suspension and steering response adjust to suit your driving style.

The car also reads your destination from the calendar in your PDA and checks traffic reports from the Web. Since the freeway is jamming up a few miles ahead, the dash display suggests a detour. You are directed around the jam by a dynamic map that scrolls as you drive, showing only the street you’re on and the next turn to take, not the whole confusing street grid. Back on the freeway, you let your eyes linger on the roadside scenery. FaceCam notices that and generates a little rumble-strip sound to get your attention before you drift out of your lane.

Later, at your exit ramp, another system puts an incoming phone call on hold while you navigate some sharp curves and lane changes. Finally, approaching the downtown office tower where your meeting will be, the car senses that the parking garage is full. It directs you to a nearby lot where a transmitter is signaling “spaces available.” You walk into your meeting with time to spare.

There’s more: The car monitors itself for mechanical and electrical problems, and in many cases can temporarily “fix” itself well enough to get you home safely. Moving all such things from the conceivable to the realm of the feasible is the daily business of the GM Lab.

From Prototypes to New Business Models

“We’re getting more sophisticated,” Schlesinger says. “Our projects are evolving and we have a better understanding of the problems.” The FaceCam system--developed by ECE professor Tsuhan Chen, with his colleagues and students--is now installed in one of the four GM test cars on campus, along with other gear such as a floor-mounted GestureCam that will let you use hand gestures to control many in-vehicle systems. On a driving simulator, student researchers in the interdisciplinary Human-Computer Interaction Institute have tested prototypes of the dynamic map (which proves very easy to read) and the cell phone-interrupt system.

Moreover, as systems and technologies are combined, a host of new apps become possible. “How about a ‘teen tracker’?” asks Schlesinger. “When your car sees that it’s your teenager driving, maybe it won’t do certain things, like go faster than 55. Maybe it calls you if the car is going to certain places, or tells your kid: ‘It’s quarter to 11, time to head home.’”

GM of course will ultimately decide what kinds of features and services it brings to market. They’ll be developed further at GM’s own R&D center in Warren, Mich., with which the Carnegie Mellon GM Lab partners closely. Many factors will have to be considered, from privacy issues to the readiness of telecom infrastructure in the world outside the car.

Through it all, Schlesinger sees a big trend emerging: “Information technology will change the industry’s business model.” He explains that many new IT features might include services (like travel advisories) as part of the package. Others could be plug-in modules--like a controller to custom-adjust suspension and handling--or just software. “The car itself becomes a commodity, a platform for selling services and add-ons, like a cell phone or a PC,” Schlesinger says. “You’ll be going to the ‘car store’ in the mall every few months to buy new things that you’d like.”

The Deeper Challenges

Not yet, however. Cars at present are not designed for constant digital-age upgrading--either at the car store or in the factory. Thus a major research challenge for the Carnegie Mellon GM Lab is developing an electronic infrastructure and system architectures that are highly robust and adaptable. That will take a lot of doing on many fronts.

For instance, an in-vehicle wireless network is a key backbone feature. It would let you and your passengers do things like link up everyone’s portables and handhelds, or pipe music from an MP3 player into the car’s sound system--and upload or download anything at speeds limited only by external networks. But how does one deploy the antennas for optimum coverage in a car interior filled with ever-
changing configurations of people and stuff? Which wireless technology should be used? A team headed by ECE professor Dan Stancil is exploring these issues and learning daily.

Simply enabling engineers to design a smarter car is a challenge. Today’s cars already are crammed with electronics for everything from fuel-injection control to braking, and as ECE research engineer JoAnn Paul observes, “you’ve got a lot of programmable, heterogeneous microprocessors here.” Each has functionalities that can be embodied either in hardware or software--a design problem from the get-go, she notes, as the two media work and are modeled differently. It gets worse when system designers must spec out component designs to different suppliers. The result is an engineering Tower of Babel, with the lack of a common design language often making it hard to know what’s going on, let alone to work efficiently or optimize designs. Paul and ECE Professor Don Thomas are developing an approach they call MESH (Modeling Environment for Software and Hardware), a unifying meta-language that allows hardware/software codesign and would make the whole design process easier and more understandable.
Yet another problem: As onboard systems grow more complex, there are more pieces that can fail. And as ECE Professor Phil Koopman has said, “One of the worst things an auto company can do is leave you stranded by the side of the road.” For Koopman’s research team on dependable embedded systems, one of the major goals is “graceful degradation.” The idea is to design systems so that if a module malfunctions, or a mechanical part hangs up, a fail-safe mode kicks in to get you home: perhaps other systems or components can cover for the one that’s failed, performing its function in a different way. (Engine stuck at full throttle? Don’t spark every cylinder every time.)

The team’s work also has the related design goal of “graceful upgrade.” The new, improved version of any module should plug right into an existing system and be compatible--to avoid massive redesign, to simplify repairs, to give car owners the option of upgrading. Basic research in all these areas has potential applications far beyond the automobile. In fact, such research is typically co-funded by others [such as Defense Advanced Research Projects Agency (DARPA)] along with GM. “The beauty of the GM Lab is that it gives us a great platform for seeing if our ideas work in the real world,” says Schlesinger. “Maybe someday, these things we’re testing in the automotive environment will permeate design everywhere and have impact universally. That’s our mission from the academic point of view, and it dovetails nicely with GM’s mission here on campus.”