Identifying and testing the Operational Design Domain factors of Lane Keeping System at horizontal curves using PreScan

Abstract

The European Commission has mandated the requirement of advanced safety features in all vehicles sold in the EU from 2022 onwards. This will help drivers get gradually used to ADS, eventually transitioning to fully autonomous driving in the distant future. There are more vehicles on road now equipped with lane-keeping systems and other driver assistance systems due to the added benefit of safety. One such Advanced Driver Assistance System (ADAS) feature is the lane-keeping system which has plenty of performance evaluation research being done on it but lacks research from an Operational Design Domain (ODD) perspective in simulation. ODD is the set of conditions for which the vehicle is designed to function. There are many ODD taxonomies present in academic literature for the classification of ODD attributes in the ODD definition. These are recommended to be used by Original Equipment Manufacturers (OEM) to classify their ODD definition of the system for effective comparison between the same ADAS feature offered by two different OEMs. However, this is not seen to be done by the OEMs and causes a disparity between the ODD definitions, leading to misunderstanding or misinterpretation of the defined ODD. The OEMs do not exactly specify the ODD boundaries in publicly available instruction manuals of the vehicles. The conditions where the system cannot function properly are very vaguely mentioned. Furthermore, the different OEMs provides the same ADAS feature with different names with differently defined ODD boundaries. Bringing these ODD definitions of all these OEMs under one umbrella is currently being attempted by policymakers and regulatory organisations. A standard way of testing the impact of the different ODD attributes on the performance of a lane-keeping system is still absent. Therefore, this research aims to provide a method to assess the ODD boundaries of a lane-keeping system using the simulation software Prescan by Siemens and is intended to be used by researchers, policymakers and OEMs to refine the ODD and to test the effect of the ODD attributes on the performance of the system. It was found in this research that there is a strong interdependency between the tested ODD attributes. This can lead to a possibility that the ODD definition may not exactly be a straightforward set of specific values in the form of a table, but instead can be dynamic considering this interdependency. It was also observed that some test cases classified inside ODD from the initial ODD assessment had poor lane-keeping performance and some test cases classified outside ODD had good lane-keeping performance. Another observation from the test cases was the ability of the test vehicle to move at higher speeds in narrow lane widths with good lane-keeping performance compared to wider lane widths. This falls in line with findings from previous literature and will require validation. The research approach of this study is used to develop a conceptual framework to assess the ODD boundaries of an ADAS feature in a simulation environment.