Carnegie Mellon and General Motors: Looking to the Future

For more than a century, Americans have enjoyed a love affair with their cars. Although vast strides have been made in terms of safety, comfort, and reliability, there’s still room for improvement—and nobody knows that better than the world’s largest automaker. So when General Motors wanted to advance research on next-generation vehicle technologies, they looked to Carnegie Mellon for help.
   
The Perfect Fit

"GM is an aggressively forward-thinking company and Carnegie Mellon is a leader in software expertise," said Raj Rajkumar, professor of electrical and computer engineering, who sees the partnership as a perfect match between industry and academia.

Rajkumar believes the future of driving technologies is in computing. The average car being sold today, for example, houses between 30 and 40 computers (with high-end models in the 70-plus range), which monitor or enable systems such as anti-lock brakes, cruise control, traction control, fuel injection, stereos, and more. As technology improves and costs drop, more and more computing power and features will be added.

Rajkumar is a leader in the field of cyber-physical systems (CPS), which integrate computing and communication capabilities with the dynamics of physical and engineered systems. Cyber-physical systems include many of a car's computer systems. These systems use software to perform computations and networks to communicate among computing elements. They also use sensors to sense driver inputs and automobile/road conditions. Finally, they actuate physical entities such as brakes, skid control or fuel injection systems. All these operations must be carried out in real-time, safely, and reliably.

According to Rajkumar, cyber-physical systems will revolutionize automobile technologies in many ways. As one example, wireless networks enable us to communicate from virtually anywhere and all computers have the ability to talk to each other via networks.

"Imagine driving down the street, your garage door calling you to remind you that you left the door open, and being able to close it from wherever you are at," he said.

But the end-goal is much more ambitious: General Motors and Carnegie Mellon recently announced a new collaborative research lab (CRL) to develop driverless vehicles. The Autonomous Driving CRL will be headed by Rajkumar and Dr. Nady Boules, director of the Electrical & Controls Integration Lab at GM R&D.

Autonomous Driving CRL

The benefits of autonomous driving are plentiful. Because human error is a primary cause of automotive accidents—a driver can be distracted, tired, sleepy, angry or even drunk—driverless vehicles can perhaps prevent thousands of roadside deaths per year. Although some drivers may be skeptical about relinquishing control to a computer system, in reality much of how we drive now relies on computing systems. "Autonomous driving isn't really about giving up control," Rajkumar adds, "it's about making driving safer, less stressful and more convenient."

Senior citizens, for example, must often quit driving at a certain point, relying instead on family and friends to help them get around, which could lead to lowered self-esteem as well as decreased social activities. Autonomous vehicles could alleviate those problems and restore senior independence. Similarly, autonomous vehicles could safely drive teenagers to school or practices.

Of course, seniors and teenagers aren't the only ones who will benefit from autonomous driving. "Imagine being virtually chauffeured safely in your car while doing your e-mail, eating breakfast and watching the news," said Larry Burns, GM vice president of R&D and Strategic Planning.

A History of Collaboration

Carnegie Mellon was chosen as the site for the Autonomous Driving CRL because of a history of successful collaboration with General Motors. In 2000, GM contributed $3 million over 3 years to the university to establish the Information Technologies Collaborative Research Lab.  The relationship was subsequently extended for 6 more years with $7.5 million in support. Rajkumar joined the lab as co-director in 2004. Dr. Boules is GM's co-director of this laboratory as well.

The Information Technologies CRL investigates the evolution from cars being electromechanical to incorporating embedded systems. It's a transformation that's happening fast and one that's assisting drivers along the way.  Navigation devices with Global Positioning System receivers, for example, can work as active traffic management devices by sending and receiving information about the car's speed and location and displaying traffic patterns, which can be used to adjust routes and avoid delays.

While the two labs are closely related, researchers at the Information Technologies CRL are developing technologies that will assist drivers without autonomous driving. For instance, vehicle-to-vehicle (V2V) networks can add wireless capabilities to cars—similar to WiFi networks—that let them to talk to each other and help prevent rear-end collisions. Safety measures such as these will be much improved with autonomous capabilities.

Rajkumar envisions even more applications emerging from vehicle-to-vehicle networks. Data sharing, he said, will become a common practice. Just as people can sync up their laptops, PDAs, cell phones, and MP3 players, so too will drivers and passengers be able to access music and movies from other cars, gas stations, and rest stops, using V2V networks.

Taking it to the Road

With breakthrough technologies emerging from the Information Technologies CRL, General Motors and Carnegie Mellon were on pace to push the collaborative research even further. So in 2007, when the Defense Advanced Research Projects Agency (DARPA) announced an Urban Challenge, the two institutions jumped on board.

Carnegie Mellon already had a solid record at the 2005 DARPA Grand Challenge, a field-test to accelerate research and development in autonomous vehicle technologies specific to battlefield conditions. Carnegie Mellon's Sandstorm and H1ghlander finished in second and third place, respectively. But the 2007 Challenge took that test to an urban environment, motivating GM to become Tartan Racing's primary sponsor.

"They sent us some of their best researchers, the best equipment possible, the best technicians, and showed good friendly support for all the teams, not just Carnegie Mellon," Rajkumar said of GM's sponsorship.

Boss, an autonomous Chevy Tahoe, beat 10 other vehicles to complete the 55-mile course about twenty minutes ahead of the second-place finisher. The win was especially significant for GM and the university because the Urban Challenge mimicked a real-world environment, complete with traffic signals, other vehicles, and intersections. For their efforts, the DARPA awarded Tartan Racing $2 million.

A Team Effort

Dr. Red Whittaker, who led the 2005 Carnegie Mellon team, also led the Tartan Racing team, along with Chris Urmson, John Dolan, Paul Rybski and others. Dr. Urmson was the Director of Technology for the overall Tartan Racing team, Dr. Dolan led the subgroup that designed Boss's behaviors under different driving situations, and Dr. Rybski led the subgroup that extracted information from Boss's multiple sensors. In light of their accomplishment, Urmson, Dolan, and Rybski will play key leadership roles in the new Autonomous Driving CRL.

Among the most promising technologies being developed at the CRL is computer vision and pattern recognition. Lead architect Tsuhan Chen, professor and associate department head of electrical and computer engineering, is working on exploiting cameras as smart yet inexpensive devices.

Tests have found that cameras can detect road signs up to two seconds before drivers, particularly when they are partially obstructed by trees or brush. In time, the cameras will be used to locate pedestrians, food signs, and more. Cameras can also monitor the driver and detect signs of fatigue, which could trigger the car to make necessary adjustments ensuring safety.

Priya Narasimhan, associate professor of ECE, is undertaking one of the most significant tasks in making next-generation vehicle technologies practical. Her areas of expertise include diagnostics and prognostics, or recognizing and predicting problems before the car even starts moving.

There are, of course, still many hurdles to overcome before we see the car of the future on the road. Issues of privacy and security must be addressed and the technology must be affordable. But with Carnegie Mellon and General Motors behind the wheel, you can be sure that future isn't so far off.

--Doug Phillips