Intelligent Data for Incident Detection

Non-recurring incidents on our nation's freeways is estimated to cost over $60 billion a year. The challenge is to develop an effective methodology to electronically identify these incidents quickly in order to manage the freeway operation and restore the facility to its optimum operating level in the minimum time. Single vehicle loop detectors have traditionally been used to measure occupancy on the freeway segments as a means of determining when incidents have occurred. Most incident detection algorithms using the traditional occupancy parameter have either taken too long to report an incident or they have given an excessive number of false alarms. Trap travel time obtained from two loops in each lane is shown in this paper to have the promise of much faster detection of the incidents with a lower false alarm rate.