Location

Rockport, Maine

Date

28-6-2005

Session

SESSION 1 - Lectures Quantification of Driver Performance

Abstract

Drivers aim to maintain their vehicle within a number of individualsituated safety margins. Safety margin violations are characterized by rapid strongcorrective steering. Steering entropy was introduced to quantify drivers’ efforts tomaintain their lateral safety margins. In the original steering entropy, severalcomputational assumptions were made. The objective is to scrutinize andmotivate these choices and exemplify the effects of deviations from these choiceswith data from a driver distraction study. The new optimized algorithm is shownto yield significances where a number of classical metrics fail to find anysignificance. Its sensitivity is attributed to the fact that a number of observedchanges in steering behavior all manifest in a widened steering prediction errordistribution which the algorithm picks up sensitively with its log-based weightingof prediction error outliers and its use of a prediction filter that is maximallysensitive to the spectral characteristics of the baseline data.

Rights

Copyright © 2005 the author(s)

DC Citation

Proceedings of the Third International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design, June 27-30, 2005, Rockport, Maine. Iowa City, IA: Public Policy Center, University of Iowa, 2005: 25-32.

Share

COinS
 
Jun 28th, 12:00 AM

Steering Entropy Revisited

Rockport, Maine

Drivers aim to maintain their vehicle within a number of individualsituated safety margins. Safety margin violations are characterized by rapid strongcorrective steering. Steering entropy was introduced to quantify drivers’ efforts tomaintain their lateral safety margins. In the original steering entropy, severalcomputational assumptions were made. The objective is to scrutinize andmotivate these choices and exemplify the effects of deviations from these choiceswith data from a driver distraction study. The new optimized algorithm is shownto yield significances where a number of classical metrics fail to find anysignificance. Its sensitivity is attributed to the fact that a number of observedchanges in steering behavior all manifest in a widened steering prediction errordistribution which the algorithm picks up sensitively with its log-based weightingof prediction error outliers and its use of a prediction filter that is maximallysensitive to the spectral characteristics of the baseline data.