High-speed autonomy done safely 

An important challenge for the acceptance of autonomous driving is to ensure the safety of road users without a loss of speed.

With today’s state-of-the-art technology, building self-driving cars that are safer than human drivers would result in a loss of speed and comfort and further decrease the acceptance of autonomous mobility. This was the result of a study by the Insurance Institute for Highway Safety, a US traffic safety organisation that regularly publishes research on autonomous driving.

Pilot studies by German automakers also confirm passengers’ perception that autonomous vehicles are mostly slow and hesitant. A key challenge to introducing autonomous systems onto the market is thus to ensure safety without limiting speed and comfort to the extent that it negates acceptance of the technology.

In the Layers of Protection Architecture for Autonomous Systems (LOPAAS) project, Fraunhofer IESE, Fraunhofer IKS, and the University of York are pursuing the goal of enabling autonomous vehicles to drive faster and more safely. The results of the project will then be incorporated into standards for technology transfer purposes.

IESE is contributing its expertise in dynamic risk management, which enables autonomous systems to assess and control the risks of their options for action in a situation-specific manner, while IKS is focusing on trustworthy AI-based situation detection as well as runtime monitoring of the associated uncertainties. The University in York is contributing its expertise in systematically generating comprehensive and traceable safety reasoning.

“Current approaches assume worst-case scenarios to ensure optimal safety. Among other things, they are based on calculations of physical laws governing how objects move. However, this leads to reduced speed of the vehicle,” says Dr Rasmus Adler, programme manager Autonomous Systems at Fraunhofer IESE and project manager of LOPAAS. “It is also difficult to correctly assess multiple risks that can occur simultaneously, such as a pedestrian suddenly appearing on the left of the vehicle and a cyclist on the right side of the vehicle. The aim is to implement an understanding of risk in vehicles that does not calculate the worst case and thus does not overestimate all risks.”

The researchers’ new methodology is already being applied in the field of intralogistics. A project with Hitachi, for example, focuses on safe and efficient collaboration between autonomous mobile robots and human workers in industrial warehouses. The underlying solution principle is to replace static worst-case assumptions commonly used for safety design with dynamic safety mechanisms that utilise knowledge about the specific current situation of a driverless transport system.

For example, the assumption of how likely a person is to move in the intended direction of travel of a machine can be more accurately estimated based on the current work task or previous movement of people at that location. This also allows the system to better estimate whether or not proactive braking is actually necessary. Systems should monitor the relevant characteristics of themselves and their context, project these properties into the future, and draw conclusions about their effect on risk.

“In simple environments like warehouses, our approach to dynamic risk management works very well. Hitachi plans to equip its driverless forklifts with this. We will be optimising our methodology for complex traffic situations with robotaxis and autopilots until the project ends in June 2024,” says Adler. “For this purpose, we are also using AI and data-driven models, which are essential for environment recognition and object classification.”