Fraunhofer Institute for High-Speed Dynamics, Ernst-Mach-Institut, EMIThe KIsSME project shows how innovative algorithms can filter out the critical moments in traffic from huge mountains of data, advancing research and technology while conserving resources.
The aim of the KIsSME project was to develop algorithms for efficient data acquisition in test vehicles. To this end, Fraunhofer EMI addressed the question of how safety-relevant situations can be identified from a vehicle’s sensor data. A modular evaluation framework was developed as a core component, which enables the calculation of an overall criticality for the respective driving situation. Input variables are evaluation measures (metrics) that can be derived from the driving dynamics of individual road users and the relative movement of several road users to each other.
Safety-critical value ranges are identified for all metrics and scaled on this basis in order to obtain dimensionless and therefore comparable variables. The combination of the scaled metrics ultimately leads to the desired overall criticality. The main advantages of the system over previous approaches to criticality assessment are the modularity of the system when selecting the metrics and the ability to evaluate very complex driving scenarios by calculating an accumulated criticality value.
EMI has also implemented an AI method for generating new scenarios and predicting vehicle trajectories. These predictions can in turn be used to calculate more detailed metrics for criticality assessment.
The KIsSME project was completed in 2023. EMI employees are now concentrating on applying the knowledge gained in follow-up projects. In the AVEAS project, for example, methods for the data-based optimization of traffic simulations are being investigated, with the identification and simulation of safety-critical situations playing a key role.
