Driver Assistance Systems with surrounding perception systems, often called Advanced Driver Assistance Systems (ADAS), are attributed to have a high potential to reduce accidents. Therefore they are growingly interconnected and have increasing options to intervene in the driving dynamics of the vehicle. The expansion of intervention options are generating new risks, which needs to be evaluated to get approval for public traffic. Specific for these systems is that they derive the decision about the intervening action from a situation representation. In addition to the failure of components, which is addressed for example by the ISO 26262 standard, this adds the risk of an unintended reaction of the system. Causes of the unintended reaction are an incorrect interpretation of the situation, for example as a result of situations that have not been considered in the specification of the system. To identify the requirements for an approval method of ADAS, existing approaches are summed up and the effort with these methods, for example according to the requirements of ISO 26262, is quantified. The analysis reveals that existing methods have gaps concerning the verifiability of the completeness of possible situations and the scope of the requirements in connection with their approvability in testing. Due to the expected high effort for the approval of safe behavior of ADAS, prioritization and transferability of test results are necessary. To address the problem of completeness, an approach for an objective evaluation and comparison of driving situations is described. The overall findings are concluded and the necessary steps for the development of a unified approval strategy for Advanced Driving Assistance Systems are summarized. (Author/publisher)
Abstract