Incident detection on urban freeways

An automatic incident-detection model using the standard normal deviate (SNA) of the control variable (energy or lane occupancy) was proposed, developed, and evaluated. Two strategies were tested using a 3-and 5- minute data base for each control variable. Strategy A required one SND value to be critical; strategy B required two successive SND values to be critical. Strategy B, using lane occupancy with a 5-minute time base, produced the best results. It detected 92 percent of the 35 incidents studied during moderate and heavy flow, with a computer response time of 1.1 minutes and a 1 percent false-alarm rate during the peak perod. Based on a limited sample size, the study indicated that the SND model was as effective as the composite model, which was considered to be the best existing model. Because the SND model does not require separate distribution curves for various traffic conditions, it may be a more attractive model for an operational system. Relationships were developed and presented that identify sensor spacing requirements for an incident-detection system using a station model.