zFAS processes information gathered by the sensors to create a detailed impression of the car’s surroundings. An online connection routes this data to a cloud‑based IT backend. There, the data is processed using machine learning algorithms and artificial intelligence before it is transmitted back to the car. The car thus continually expands its intelligence as it drives, ceaselessly improving its performance in complex situations. At the same time, other users of this technology can also benefit from this accumulation of information.

The piloted driving showcar explores the fusion of the car’s architecture with the display and control concept to create a novel unit. The entire front of the instrument is executed as a three-part touch display. This exploration of the human-machine interface is important, Vukotich says, because enabling a satisfactory relationship between the human driver and the automated capabilities of the car will be critical to the realization of the piloted driving technology.

Audi has already progressed to the point at which is has confidence in its piloted driving technology capabilities in highway scenarios; the pre-CES drive from San Francisco to Las Vegas is only one proof point is a series of demonstrations. The zFAS, the central control unit, is interchangeable between implementations, Vukotich said.

We have shown the ability to do [piloted driving on the highway]. The things that we use to do it are close to series production. We are also already using certain components in production models.

—Alejandro Vukotich

What is extremely critical, Vukotich noted, is the managing the relationship between the driver and the car; there will not be a car that drives itself all the way.

We will start with highways. Those are pretty simple compared to the cities, where you have crossings and bikers and pedestrians. There will always be a point where the car the will not be able to drive. You are always having a dialogue between the driver and the car.

—Alejandro Vukotich

NVIDIA, one of Audi’s technology partners for the zFAS, has just announced its DRIVE line of automotive computers, based on the new NVIDIA Tegra X1 mobile super chip, which is built on NVIDIA’s latest Maxwell GPU architecture and delivers more than one teraflops of processing power, giving it more horsepower than the world’s fastest supercomputer of 15 years ago.

That type of processing power, Vukotich suggested, will be required for delivering piloted driving capabilities in the city, which is characterized by unclustered, unstructured situations requiring significant machine learning that requires a great deal of computing power.

However, one of the main issues with piloted driving, Vukotich adds, is the time required to hand control back from the car to the human: about 10 seconds, according to studies. While on highways, and especially in traffic jams, that lag time doesn’t create significant problems; in city driving, that is not the case.

We have to maintain safe driving; we have to make sure that we anticipate the 10 seconds. If there are power losses, communications problems, we have to deal with it. The problem is not the car, it’s the human being.

—Alejandro Vukotich

While some of that 10-second issue might be addressed with HMI refinements, much will depend of the regulations that emerge governing the application of piloted driving and the accompanying safety requirements, Vukotich said.