Drivers aim to maintain their vehicle within a number of individual situated safety margins. Safety margin violations are characterized by rapid strong corrective steering. Steering entropy was introduced to quantify drivers' efforts to maintain their lateral safety margins. In the original steering entropy, several computational assumptions were made. The objective is to scrutinize and motivate these choices and exemplify the effects of deviations from these choices with data from a driver distraction study. The new optimized algorithm is shown to yield significances where a number of classical metrics fail to find any significance. Its sensitivity is attributed to the fact that a number of observed changes in steering behavior all manifest in a widened steering prediction error distribution which thealgorithm picks up sensitively with its log-based weighting of predictionerror outliers and its use of a prediction filter that is maximally sensitive to the spectral characteristics of the baseline data.
Samenvatting