From Data to Information: Probabilistic Methods for Dealing with Sensor Inaccuracy

Abstract

As the number of assistance systems in the automobile increase, we are increasingly supplied with data from various sensors about the state of the vehicle, the driver and the driving environment. A fusion of the available data from these sensors can provide the assistance systems with a valuable overview of the driving situation, which can be used for situation-based tuning of the assistance offered. The raw data returned by several of the sensors is, however, intrinsically uncertain, making a simple, rule-based evaluation of the data impractical. Computations based on probability theory are therefore used to fuse the available data and draw inferences about the driving situation from the uncertain data available. A probability value assigned to the inference provides an indication of the level of confidence that may be associated with the inferred information.