Models for surveillance of variability in motorway traffic flow

The accurate description of the fluctuations of the traffic flow dynamics has been recognized as a crucial element for monitoring and controlling the operational performance of road networks with uninterrupted (motorway/freeway) and/or interrupted (arterial) flow conditions. Nonetheless, analytical frameworks for modeling and predicting the higher statistical moments of traffic flow have only recently illustrated in the relevant literature. This paper describes a discrete-time parametric stochastic model, referred to as Stochastic Volatility (SV) model, which enables the online adaptive forecasting of traffic variability, as it is expressed by the measure of volatility, i.e. the conditional variance of the traffic flow level. The SV methodology provides a theoretically sound and parsimonious modeling framework, which describes the traffic volatility dynamics as a latent stochastic (low-order Markov) process. Empirical results based on real speed data from a U.S. freeway illustrates that the SV model can provide considerable accuracy gains in the short-term prediction of traffic variability, in comparison to other models hitherto proposed in the literature.