Prof. Stiller, will we soon be able to stop worrying about getting home after a party because intelligent cars will do the driving for us?
Stiller: It’s still going to take a couple of decades for cars to drive themselves in all situations. On the way to that goal, intelligent cars will automatically mitigate more and more critical situations and help drivers drive more safely. We’re already familiar with systems like the electronic stability program (ESP) that recognize a skidding situation and then apply to brakes to specific wheels to stabilize the vehicle.
Just what makes automobiles intelligent? Is it a matter of software technology and sensors?
Stiller: That’s a philosophical question. What makes people intelligent? I think it’s the ability to experience your own environment, to understand it, to imagine the effects of different actions in the environment, and to be able to decide on the best alternative. The same kind of thinking would apply to automobiles. It’s already possible to equip the driving environment with video, radar, and light detecting and ranging (LIDAR) sensors – albeit at high cost. For the most part, we can already calculate how a vehicle reacts to our actions – braking, accelerating, and steering. The ability of a computer to understand the environmental information is a significant technological challenge. Part of the challenge is making the computer aware of the limits of its perception. For example, other vehicles can obstruct the view of other drivers, so the car must be able to recognize such a situation reliably and then adjust its own actions. Enabling a vehicle to predict the behavior of other drivers is another major task for us.
What is the most important ability that a car needs to solve complex traffic situations on its own?
Stiller: The key is the ability to recognize these situations reliably and early. When that happens, the car can automatically perform the appropriate maneuvers.
When will the technology advance to that point?
Stiller: In my estimations, we need at least 20 more years for secure, automatic driving in all situations. The first intelligent driving functions have already been implemented in mass-production vehicles. Many automobile manufactures already offer video, LIDAR, or radar sensors that provide information to keep the vehicle in the proper lane and to support headway control. Things have already developed to the point that in some, clearly recognizable situations – like encountering a traffic jam on a highway – a car can automatically brake before an accident occurs. Manufacturers have not yet dared to build automatic intervention into steering – even though that would often be a more efficient strategy for avoiding accidents when a vehicle travels at great speed. But before they can enable a function in a vehicle, manufactures must be sure that the onboard computer cannot make bad decisions that would themselves lead to accidents.
What are the most significant difficulties in developing intelligent vehicles, and which technology is most promising right now?
Stiller: I see the greatest and most fascinating challenge in teaching a car to understand its environment. In my opinion, the most significant potential is found in analyzing video data. After all, anyone can drive a car with this information.
What is your vision of an intelligent car?
Stiller: Someday, I want to sit in a car that drives without accidents and that coordinates its behavior in traffic with other cars to optimize safety, traffic flow, and environmental effects.
What can the cognitive automobiles developed in your research area do already?
Stiller: Our research combines the University of Karlsruhe, the Fraunhofer Institute for Information and Data Processing (IITB in its German abbreviation) in Karlsruhe, the Technical University of Munich, and the Military University in Munich. Our vehicles already capture most of the three-dimensional geometry of a scene: they see other cars and lanes. They can also perform a variety of driving maneuvers autonomously. For example, they can independently plan a route to a specific destination and then drive the route. They can look for a parking spot, create a plan for parking, and execute the plan. In the initial simulations, they have been able to draw simple, logical conclusions about their environment, but we’re just at the beginning in this regard. For example, a car can already conclude that another car will probably change lanes because it’s traveling behind a slower-moving vehicle and the lane to its left is empty.
In November, your research area will participate in the Urban Challenge competition in the United States with an intelligent vehicle. What are the greatest challenges in such a race, and how is your team, AnnieWAY, technically equipped for such an event?
Stiller: The competition has automobiles without operators drive through a typically urban environment. They have to perform driving tasks and follow traffic laws like granting the right-of-way at intersections and merging into flowing traffic. We created a team made up of students, doctoral candidates, and scientific employees of the universities in Karlsruhe and Munich. Thanks to the Cognitive Automobile special research area, we have a vehicle equipped with the required sensors and computers. The equipment includes video, LIDAR, and radar sensors along with precision GPS. The onboard computer is able to control the brakes, gas pedal, and steering.
What are your chances?
Stiller: Our team impresses me. It can do everything, including winning. We’ve already gone through several elimination rounds and are now one of the 43 best teams among the original 100 competitors. That’s already better than we expected, but we’re playing a game of David versus Goliath. Our competitors from the United States have often received millions of dollars from the hosts and sponsors. In many ways, that helps the other teams. We’d like to see the same kind of involvement from German industry that our American competitors enjoy.
But we think of the competition as a sporting event. We don’t keep our research secret. We discuss it with our competitors, who also look at research advances as the ultimate goal. We’ve already won because of the scientific incentives linked to the competition and the give-and-take with top American researchers.