Traffic Congestion Analysis and Probability Estimation Based on Stochastic Characteristics of Traffic Arrival

Abstract

Urban traffic congestion has emerged as a global challenge constraining sustainable development. Estimating the traffic congestion probability is crucial since it presents valuable information for formulating congestion mitigation strategies and improving traffic management. The existing studies employ deterministic models to predict congestion; however, they do not consider the dynamic coupling between intrinsic traffic flow randomness (e.g. spatiotemporal heterogeneity in vehicle arrivals) and congestion formation mechanisms, causing prediction biases under high-uncertainty scenarios. In this study, we propose a probability-based congestion estimation framework that employs the stochastic traffic flow theory. The traffic arrival process is described using discrete probability distributions owing to the stochastic nature of traffic flows. To prevent the misclassification of transient traffic surges as congestion, we adopt a spatiotemporal persistence criterion with dual thresholds (vehicle accumulation exceeding a critical level and duration surpassing a minimum time) for congestion identification. Additionally, we perform empirical validation using traffic datasets from Portland, USA, which demonstrates that there is no statistically significant deviation from the measured data at the 95% confidence level in the calculated congestion probabilities. The proposed method facilitates the development of targeted congestion mitigation countermeasures and presents novel insights for future transportation planning.