Time series of travel speed on multilane freeways are considered complex and irregular particularly when addressing the variability across lanes. Literature shows evidence of interactions between speed variability and traffic mix and inclement weather, without extending these results to addressing speed predictability across lanes. We propose the development of a Bayesian system of equations in order to concurrently treat time series collected from each lane in an autoregressive methodological framework. Exogenous variables such as volume, percentage of trucks per lane, as well as precipitation levels are integrated into the model. The proposed approach improves on the predictability of travel speeds across lanes over the commonly used ARIMA models.
展开▼
