Empirical analysis of dilemma zone using high-resolution event data

The dilemma zone (DZ) has been physically characterized based on two divergent definitions: Type I and Type II. However, treating DZ differently based on these definitions may lead to inaccurate results of DZ boundary and subsequent safety analyses. Moreover, an integrated empirical assessment of Type I and Type II definitions for consistency in boundary quantification is not yet well-addressed. To this end, we empirically analyzed the two DZ definitions by comparing their boundary dynamics with approach velocity and time of day. First, we proposed a rule-based matching methodology with 92% accuracy to match actuation events between the advance and stop-bar detectors. This methodology was then applied to process two months of high-resolution event data from an intersection approach, yielding 28,700 vehicle arrivals on yellow. Results showed that 13.2% of approaching vehicles fall into an indecision zone or make Type I-contrary stop/run decisions at the yellow onset. The Type I and Type II DZ boundaries were temporally segregated and did not significantly overlap. Our novel findings indicate a lack of consistency in quantifying DZ and emphasize a need for data-driven quantification of the DZ boundary and its dynamics.