Abstract:
Classical traffic flow models cannot be readily applied in heterogeneous traffic systems owing to the complex nature of their traffic dynamics. This paper develops a stochastic macroscopic model for traffic state estimation and short-term prediction in such systems. The proposed model takes into account the wide variation in the operating and performance characteristics of vehicles in heterogeneous condition through the use of variable fundamental diagrams (FDs) for different links. The model also allows for the underestimation of flow and speed due to the effect of vehicular influence area in the stated traffic condition. For this, normally distributed stochastic state influencing terms are used with the basic state estimation equations. In addition, an empirical parameter is introduced in the speed dynamics of the model to capture the sensitivity of traffic speed to the speeds of multiple leaders in a heterogeneous mix. To confirm the structure of the FD, initially the speed-density plots of the field data for different links are fitted with four general structures: namely, the linear, logarithmic, exponential and polynomial forms. It is revealed that the 3rd degree polynomial structure is best suited for prevailing traffic condition. The optimized link-specific parameters of the model comply with those obtained from the regression analysis. Field validation with high-resolution traffic data proved that the proposed model can capture traffic dynamics quite accurately. To determine the individual contributions of the proposed model features, different structural variations of the final model are also investigated. It is revealed that the link-specific FD parameters and the stochastic traffic state influencing terms improve the model performance the most, followed by the empirical car-following parameter. Finally, compatibility analysis is performed on the proposed macroscopic model and a microscopic simulation model, VISSIM, to evaluate the performance of the macro model under varying traffic demand levels. Based on the performance of the two models, it is found that the prediction of traffic states from the macroscopic model is generally consistent with that from VISSIM simulation.
