A sequential detection approach to real-time freeway incident detection and characterization

In this paper, a new methodology is presented for real-time detection and characterization of freeway incidents. The proposed technology is capable of detecting freeway incidents in real time as well as characterizing incidents of terms of time-varying lane-changing fractions and queue lengths in blocked lanes, the lanes blocked due to incidents, and duration of incident, etc. The architecture of the proposed incident detection approach consists of three sequential procedures: (1) symptom identification for identification of anomalous changes in traffic characteristics probably caused by incidents, (2) signal processing for stochastic estimation of incident-related lane traffic characteristics, and (3) pattern recognition for incident detection. Lane traffic count and occupancy are two major types of input data, which can be readily collected from point detectors. The primary techniques utilized to develop the proposed method include: (1) discrete-time, nonlinear, stochastic system modeling used in the signal processing procedure, and (2) modified sequential probability ratio tests employed in the pattern recognition procedure. Off-line tests were conducted to substantiate the performance of the proposed incident detection algorithm based on simulated data generated employing the calibrated INTRAS simulation model and on real incident data collected on the I-880 freeway in Oakland, California. The test results indicate the feasibility of achieving real-time incident detection and characterization utilizing the proposed method.