A Unified Bayesian Approach for Object and Situation Assessment

Object and situation assessment are crucial components of many advanced driver assistance systems (ADASs). Current object assessment algorithms usually provide a probabilistic representation of relevant entities in the vehicle's environment. Similarly, probabilistic approaches for situation assessment have been proposed. However, the interface between both stages is still an unsolved issue. In this paper, a direct link between Bayes filters and Bayesian networks is proposed which is based on adaptive likelihood nodes. The method is illustrated on the example of a system which automatically determines lane change maneuver recommendations. The presented results using both simulated and real data show that the proposed approach provides a unified approach for handling uncertainties in ADAS applications.