Fraunhofer Institute: Faster Autonomy with AI

A team of scientists from the Fraunhofer Institute for Experimental Software Engineering IESE, the Fraunhofer Institute for Cognitive Systems IKS, and the University of York has now developed dynamic risk management as part of a reference security architecture. An important challenge for the acceptance of autonomous driving is to ensure the safety of road users without having to accept compromises in terms of speed. A team of scientists from the Fraunhofer Institute for Experimental Software Engineering IESE, the Fraunhofer Institute for Cognitive Systems IKS, and the University of York has now developed dynamic risk management as part of a reference security architecture.

Reducing speed is associated with less acceptance

Building self-driving cars that are superior to human drivers in terms of safety would, with today's technology, entail compromises in speed, leading to further declining acceptance of autonomous mobility. This was the result of a study by the Insurance Institute for Highway Safety, an American road safety organization that regularly publishes research findings on autonomous driving. Pilot studies by German automotive manufacturers also confirm passengers' perception that autonomous vehicles are usually slow and hesitant. A major challenge in the market introduction of autonomous systems is therefore to ensure safety without restricting speed and comfort to such an extent that acceptance wanes.

Layers of Protection Architecture for Autonomous Systems

In the project "LOPAAS" (Layers of Protection Architecture for Autonomous Systems), the Fraunhofer IESE, the Fraunhofer IKS, and the University of York, all key research institutions in the field of securing complex (software) systems, pursue the goal of enabling autonomous vehicles to drive faster and safer. The project's results are to be incorporated into technology transfer for standardization and norming. The partners pool their expertise to develop a reference safety architecture and argumentation for automated driving and autonomous systems.

AI combined with dynamic risk management

According to its own statements, Fraunhofer IESE brings its expertise in dynamic risk management, enabling autonomous systems to assess and control the risks of their action options in a situation-specific manner, while IKS focuses on trustworthy AI-based situation recognition and runtime monitoring of associated uncertainties. The University of York, in turn, contributes its expertise in systematically generating comprehensive and traceable safety arguments.

New Safety Concepts for Robotaxis and Highway Pilots

The project partners are developing innovative safety concepts for two major application areas: on the one hand for robotaxis and roboshuttles – self-driving cars for one or more passengers – and on the other hand for highway pilots integrated into private cars, software that can fully take over the driving and steering function on well-mapped highway sections under simple weather conditions. The safety concepts are being tested on specific usage scenarios of a highway pilot. With the digital “Safety Engineer,” the research teams are bringing a system on board that makes automated driving for the different use cases more efficient while guaranteeing safety. Adapted to the traffic situation, the digital Safety Engineer reacts individually and influences the driving behavior and driving experience of the person at the wheel. Dynamic risk management using AI allows for predictive driving, maintaining the required distances from other vehicles and preventing hard braking.

“Current approaches are based on worst-case scenarios to ensure optimal safety. They are based, among other things, on calculations of physical laws, i.e., how objects move. However, this leads to reduced speed of the vehicle. Another problem is the correct assessment of multiple risks that can occur simultaneously, for example, a pedestrian suddenly appearing to the left of the vehicle and a cyclist on the right side of the vehicle,” says Dr. Rasmus Adler, Program Manager “Autonomous Systems” at Fraunhofer IESE and Project Manager of “LOPAAS.” “The aim is to implement a risk understanding in the vehicles that does not calculate the worst case and thus overestimate all risks.”