Research objective
European Institutions have reached an agreement for a new General Safety Regulation (GSR) that mandates the implementation of certain ADAS systems into new vehicles from 2022. This includes advanced emergency braking (AEB) and emergency lane keeping system (ELK), among other systems. In order to assess the performance of such systems, it is required to analyse the pre-crash phase of accidents, which can be done by means of simulating crashes. One widespread database of simulation input data is the GIDAS-PCM, which contains around 10,000 accident reconstructions from two German investigation areas (Dresden, Hannover) since the year 1999. This paper focuses on developing a novel method for generating the pre-crash phase based on police recorded accidents. This allows generating simulations of a large number of accidents. The simulation results can later be used in combination with ADAS models to understand the effectiveness of those systems.

Methodology
The Fraunhofer IVI has access to all police reported accidents within Saxony (region in Germany) since the year 2010. The dataset for one accident contains up to 100 variables including injury severity, accident location and type of accident, among others. The available data contains up to 1 million accidents including fatal accidents, accidents with severe or slightly injured participants as well as accidents with material damage only. Using the information recorded by the police, the description of the accident and the aerial picture of the accident site as a basis, it is possible to create the trajectory for each participant. In order to obtain information on the speed profile of the participants involved in the accident, an analysis of in-depth accident data is carried out that defines the mean values and standard deviation of the speed profiles. The trajectory and speed profiles are used to reconstruct the accident scene and the pre-crash phase with the help of an accident reconstruction software. This information is later extracted and can be used to carry out an effectiveness assessment of ADAS systems which could have been beneficial to the participants.

Results
The newly developed method for creating simulation files by drawing on data from police recorded accidents delivers one main simulation case for each accident and up to 900 variations by using the mean values and standard deviations of initial speed, collision speed and additional parameters. Currently, the Fraunhofer IVI has ~10,000 simulation cases from 2018 and 2019, from which 1.2 million variations can be generated.

Limitation
This methodology has been developed based on the input data available at the Fraunhofer IVI, for the region of Saxony. An expansion of this methodology to other regions or countries would be subject to further studies on the availability of the required input data.

Key Words
accident analysis, simulation, effectiveness assessment, ADAS

Mr. Martin Urban, Fraunhofer Institute for Transportation and Infrastructure Systems, GERMANY; Dr.-Ing. Christian T. Erbsmehl, Fraunhofer Institute for Transport and Infrastructure IVI, GERMANY; Dipl.-Ing. Jorge Lorente Mallada, Toyota Motor Europe, BELGIUM; Dr.-Ing. Pablo Puente Guillen, Toyota Motor Europe, BELGIUM; Dipl.-Ing. Satoshi Taniguchi, Toyota Motor Corporation, JAPAN