Your browser is not supported. Please update it.

10 April 2015

Delphi secures Audi’s zFAS contract, finishes coast-to-coast autonomous trip

After introducing the world to Jack, the self-driving A7 saloon back at CES in January, Audi has announced that its zentrale Fahrerassistenzsteuergerät (zFAS –Central Controller Driver Assistance, ‘all functions, one unit’) system will be developed by Delphi – the OEM that just drove an autonomous Audi coast-to-coast.

Audi says that the zFAS control system was developed in collaboration with TTTech, Mobileye, Nvidia, and Delphi. For the hardware, the zFAS system uses an Nvidia Tegra K1 SoC and Mobileye’s EyeQ3 processor, with Delphi responsible for providing the complete board to Audi. TTTech’s contribution goes unnamed in the press release, but the company’s website says it is providing the Ethernet communication that links the multitude of sensors and cameras in the car to the zFAS unit – tucked away out of sight in the luggage compartment, about the side of a laptop.

zFAS is a different approach to the application than most, according to Audi, which says that most current driver assistance systems are managed by spatially separated controllers. With zFAS, all the mechanisms are controlled by a single board, combining the sensors, electronics hardware and the software architecture into its central system.

The goal of this centralized architecture is to more quickly compute the sensor information that the car gathers from its near-field ultrasonic sensors, a front-facing radar with a range of up to 250m, rear radar sensors for traffic monitoring, top view cameras, a front facing laser scanner with an 80m range, and a wide-angle front facing camera for object recognition. Audi says GPS gives it positional data with a resolution of 2cm.

The zFAS unit aims to compute that data and export it to the car’s systems in order to autonomously drive safely in real-time. The data is used to compute a complete model of the car’s immediate surroundings, which is then used by the different driver assistance systems that are fitted to Audi’s range. The modular design claims to be scalable as well as future-proof.

Audi has also said it is working with “leading suppliers such as Bosch, Continental, Valeo and Delphi on the sensors and actuating elements, such as braking and steering systems.” The objective here is to develop common standards while offering customers advanced systems during the transition to fully automatic driving – an incremental process if ever there was one, thanks to the myriad of legal challenges and clarifications that need to be made in order to make driverless cars street-legal.

Last of the announcements was Audi’s ambition to enable the cars to continuously learn as they drive, using cellular networks and LTE where available to backhaul the zFAS data to the cloud. The backend system will then analyze the data, using the machine learning and AI technologies that are becoming increasingly popular in the enterprise, before using the same cellular network to send updates and information to the car.

Also this week, Delphi rigged up an Audi SQ5 estate to drive some 3,400 miles from San Francisco to Manhattan, arriving at the New York International Auto Show. Taking a route through the south of the country, largely to avoid the cold, the SQ5 had a Delphi engineer at the wheel at all times to comply with the myriad of state laws that govern autonomous vehicles.

Currently, only four states have defined laws that govern autonomous driving. The Delphi trip passed through two of them, California and Nevada, but not Florida and Michigan – the other two. Delphi explained that some negotiating was required to allow them to pass through the other 13 states on its trip, but the fact that none actually prohibited the experiment is a good sign.

The technological progress is far outpacing the societal and legal changes that need to accommodate autonomous cars, but the world seems to be slowly catching up with the technical possibilities. Insurance might cry havoc in Congress, but it’s hard to argue with the forecasts for the reductions in accidents and fatalities that autonomous cars could bring about – never mind the bottom-line benefits that can be afforded by removing human drivers from the logistics industry, for example.

Delphi’s Roadrunner, as it was christened, used a slightly different set up to Audi’s Jack, with six long-range radar units, three cameras, six lidar units (laser-based radar), and Delphi’s localization system – most of which were powered by Mobileye. Road works, heat, and rain were the biggest obstacles – and conversely the most valuable sources of data for the Delphi team, which completed the trip without fault; exceeding the expectations of Delphi’s CTO Jeff Owens.

While the inner-city driving was still handled by a human, the freeway driving was fully autonomous at the push of a button. Given that active cruise control and early warning systems are already standard on some cars, the next iteration of these systems will be very similar to Roadrunner’s functionality – which will allow drivers to activate an autopilot mode once safely on a freeway, letting the car take control for long periods of relatively predictable traffic and helping reduce accidents caused by driver fatigue.

Around 98% of the driving was autonomous, and the trip captured around 3TB of data that Delphi will be pouring over, as it looks to improve on its own offerings. For now, the Audi deal appears to be its biggest future order in the autonomous space, and the team will be preoccupied with trying to work out why Roadrunner gave an extra-wide berth to tractor trailers.