Safety Evaluation of Advanced Driver Assistance Systems as Human-machine Systems

Abstract

We propose an evaluation methodology to analyze the safety level of advanced driver assistance systems (ADAS) as a human–machine systems in terms of comparing the increase in the safety level during normal system operation and the decrease in the safety level during a system malfunction. We propose a concept of combined error for the human–machine system and quantify this combined error by driving simulator investigations and simulation studies. First, we investigated the drivers’ behavior when avoiding rear-end collisions with a preceding vehicle when equipped with ADASs-like adaptive cruise control systems (ACC) and lane keeping assistance systems (LKA). Then, we confirmed that the risk of collision induced by overdependence on the systems was not increased when the ACC and LKA were mounted on the vehicle using simulation studies based on the concept of combined error for the human–machine system. We also confirmed that the decrease in collisions when the ADASs operated appropriately was much larger than the increase in collisions during a system malfunction.