Stochastic modeling and real-time prediction of incident effects on surface street traffic congestion

Abstract

Modeling and real-time prediction of incident-induced time-varying lane traffic states, e.g., mandatory lane-changing fractions, queue lengths, and delays are vital to investigate the time-varying incident effects on traffic congestion in both the spatial and temporal domains. This paper presents a discrete-time nonlinear stochastic model to characterize the time-varying relationships of specified lane traffic states under the condition of lane-blocking incidents on surface streets. The proposed stochastic model is composed of four types of equations: (1) recursive equations, (2) measurement equations, (3) delay-aggregation equations, and (4) boundary constraints. In addition, a recursive estimation algorithm is developed for real-time prediction of the specified time-varying lane traffic states. The proposed method is tested with simulated data generated using the Paramics traffic simulator. The preliminary tests indicate the capability of the proposed method to estimate incident effects on surface street traffic congestion in real time. We also expect that this study can provide real-time incident-related traffic information with benefits both for understanding the impact of incidents on non-recurrent traffic congestion of surface streets, and for developing advanced incident-responsive traffic control and management technologies.