Process Devised to Verify Safety of Assisted Driving Systems

(A. Kotok)

Computer scientists at Carnegie Mellon University in Pittsburgh have developed ways to verify the safety of driver assistance technologies, such as adaptive cruise control and automatic braking. Faculty member Andre Platzer and doctoral students Sarah Loos and Ligia Nistor will present their findings at the International Symposium on Formal Methods, on 22 June at the University of Limerick, Ireland.

Vehicle control systems promise to ease traffic on crowded routes and prevent accidents, but are, by design, highly complex. They must take into account both physical laws and the capabilities of the system’s hardware and software. But these systems add yet another layer of complexity: no single computer is ultimately in control. Each vehicle makes decisions in concert with other vehicles sharing the same road.

To verify the safety of assisted driving systems, Platzer’s team first developed a model of a car’s distributed control network. In that system, computers and sensors in each car combine to control acceleration, braking and lane changes, as well as entering and exiting the highway. They then used mathematical methods to formally verify that the system design would keep cars from crashing into each other.

The team showed that they could verify the safety of an adaptive cruise control system by breaking the problem into modular pieces and organizing the pieces in a hierarchy. The smallest piece consists of just two cars in a single lane. Building on that, they were able to prove that the system is safe for a single lane with an arbitrary number of cars, and ultimately for a highway with an arbitrary number of lanes. They were also able to show that cars could safely merge in and out of a single lane and then extended it to prove that cars could safely merge across a multi-lane highway.

Platzer cautioned, however, that the research has a major limitation: it only applies to straight highways. The problems of curved lanes, as well as sensory inaccuracy and time synchronization will be addressed in future work.

Read more: Engineers Developing Human Driving Model to Reduce Crashes

*     *     *