CONTROL: Manageable Risk for Autonomous Vehicles on Road and Rail

What happens when an autonomous vehicle receives conflicting signals from its sensors, for example in dense fog or when encountering an unexpected obstacle? How can maximum safety be guaranteed in such a situation?

This is one of the questions addressed by the research project "CONTROL". The project's goal is to make autonomous systems for road and rail more trustworthy.

To achieve this, 24 project partners are developing a cross-industry, system-level approach for handling unexpected situations in open, real-world environments. In this way, uncertainties can be identified, assessed and managed.

CONTROL thereby lays the foundation for safe autonomous driving in complex open environments, whether in the form of autonomous cars, trucks, buses or trains.

To address the complex reality of open environments, CONTROL treats the autonomous vehicle as an integrated system. This system continuously assesses whether it can operate safely. When faced with uncertainty, it responds with adaptive measures such as reducing speed or avoiding obstacles. To enable this, the project partners are developing a cross-industry safety argumentation framework that systematically captures and assesses uncertainties and integrates them into the architecture of autonomous systems.

CONTROL provides a key building block for the software-defined vehicle (SDV).

CONTROL addresses the question of how autonomous road vehicles and trains can safely handle the so-called "long-tail" problem.

The "long-tail" problem describes the large number of rare but safety-critical situations that arise in complex and not fully describable environments, which autonomous systems must handle reliably. These situations are too rare to be tested cost-effectively, yet too frequently to be ignored.

Within the CONTROL project, Fraunhofer IKS focuses in particular on conflicting sensor readings from lidar, camera and radar systems. This complicates decision-making in autonomous systems, making it difficult to determine which data can be trusted.

To address these challenges, Fraunhofer IKS is developing a residual risk analysis tool for autonomous systems. It analyses situations with inconsistent sensor data and determines how safely and reliably the vehicle operates even in unusual circumstances. This form of residual risk assessment is enabled by an innovative sensor fusion technology that quantifies inconsistencies in sensor measurements in real time and adapts control decisions accordingly. This dynamic awareness of uncertainty allows the performance of autonomous systems to be increased while providing measurable safety assurances.

These detailed safety analyses give autonomous systems greater decision-making flexibility, increasing both the utility and economic viability of autonomous vehicles.

Existing analysis methods only consider clearly identifiable faults. The Fraunhofer IKS tool instead offers a more differentiated analysis of diverging measurement results in environment perception and sensor fusion. With this tool, Fraunhofer IKS aims to deliver first safety analysis for modern sensor fusion concepts, improving safety even when sensor signals conflict.

The tool and associated methods developed within the project can also be applied in other organizations. Contact us directly to discuss your requirements.

Contact us directly to discuss your requirements

CONTROL is a project of the VDA Lead Initiative for Autonomous and Connected Driving. The project is funded by the Federal Ministry for Economic Affairs and Energy.