Empirical evaluation of a unified Bayesian object and situation assessment approach for lane change assistance

Abstract

Lane change and merging maneuvers are accounting for a significant share of critical accidents in road traffic. Thus, advanced driver assistance systems giving support during such maneuvers may be expected to significantly contribute to the goal of road safety. In this paper, a system for actively determining lateral maneuver recommendations is evaluated during an exemplary highway scene. The system is utilizing a unified Bayesian data fusion approach which covers the complete data processing chain including object, situation, and impact assessment. The results show how this method facilitates a holistic handling of uncertainties from sensor to decision level.