Driver assistance systems are systematically developing in the direction of automated driving. Reflecting this trend, the topic was also the dominant theme at the “Driver Assistance Systems Conference” in December, for which TÜV SÜD Academy and the Technical University of Munich (TUM) had brought together 240 experts from industry, research and regulatory bodies.
The conference's subtitle of “Introduction of highly automated driving” indicated the issues that are currently spurring on the industry. Technological progress already makes it possible for driver assistance systems to carry out some driving tasks independently. The first cars now on the market meet the standard of "level 3 automation". If this level three out of five possible automation levels is approved, these cars will be able to drive autonomously on motorways up to a speed of 120 km/h. However, a driver must still be in place as a fallback, i.e. to respond in case of the need to intervene. While addressing more advanced forms of automated driving, the experts also discussed the topic of “teleoperated driving”. In this scenario, operators located at remote command centres take over using remote control and steer cars safely through traffic when the autopilot reaches its limits.
However, to ensure that these solutions function safely at all times, they first need to undergo thorough testing and homologation. “Testing and approval are pivotal for automated driving and autonomous vehicles”, said Prof. Dr.-Ing. Lothar Wech from Ingolstadt University of Applied Sciences at the conference held on TÜV SÜD's premises. Simulation will play a major role in future testing and approval practice. This opinion was shared by other conference participants. Using simulation methods, test organisations can assess up to 50 million scenarios in a single day – a feat that is impossible to match or even approach with physical vehicle testing.
Vehicle electronics will increasingly be able to mimic human behaviour. “Neural networks”, a phrase heard frequently at the conference, designates a type of artificial nervous system which enables machines or computers to learn independently. Neural networks go in the direction of artificial intelligence (AI), which, as several speakers at the conference warned, should still be used “with care”. A talk on supervised learning for capturing and emulating human driving behaviour presented a possible first step in this direction. Another talk in the same field addressed the virtual assessment of driver assistance systems, taking into account human driving behaviour.
TUM presented its IMAGinE project, which is aimed at developing cooperative advanced driver assistance systems that can even introduce friendly gestures to automated driving. A car steered purely by computer logic, for example, would never give way to another car if it has right of way. By contrast, a cooperative driver assistance system can certainly learn to do so or, for example, permit other road users to change lanes – as human drivers do.
The experts at the conference, held at TÜV SÜD's premises, discussed these almost human characteristics of automated cars as well as the technologies behind them. In the future, sensors including cameras and radar scanners will increasingly transmit their data to the vehicles' central on-board computers instead of their own control units. The first cars to incorporate such computers are already available. According to the representative of one car manufacturer, “non-exclusive” platforms will also become more common in the future. The same technology will be available to several companies, including competitors; this will slash costs and make ultra-modern systems available for a larger number of vehicles.
The talks and discussions about sensors give an indication of the extent to which processing can be of value for the data supplied by cameras and radar signals. While some lecturers reported that crash barriers were still a source of interference for radar scanners, on the other hand, these new devices can now even bounce the radar signal under a car in front and scan the road ahead. Camera images recognise pedestrians and other road vehicles, even in difficult lighting conditions in which they probably would have been overlooked by human drivers.
According to the experts at TÜV SÜD Academy’s conference, these systems developed with autonomous driving in mind will come in useful even while cars are still steered by human drivers. Automated driving has thus proved to be an enormous boost for driving assistance systems.
Note for editorial teams: For high-resolution photos please feel free to contact [email protected].
Press contact: Vincenzo Lucà