Using GPS data to detect critical events in motorcycle rider behaviour

A relevant research issue in naturalistic studies is related to analysis and evaluation of the huge data recorded during the experiment. A procedure to identify critical events of PTW is presented and tested using parameters measured or derived by GPS. Raw measures of speed, position and time are copied directly from the data string stored by the GPS in National Marine Electronics Association (NMEA) protocol. From these other measures useful in explaining the rider behaviour were derived. Smoothing cubic spline and Butterworth filters were used to improve the estimation of measures effected by GPS inaccuracy. The Mahalanobis distance was adopted to identify multivariable outliers for each combination of raw and derived measures. The events classified by Mahalanobis distance as outliers from the ‘normal’ rider behaviour were compared with the critical traffic conflicts self-reported by the rider during the experimental test in order to evaluate recall and precision of the method.