Analyzing Speed on Mountainous Freeways with Longitudinal Data – A Driving Simulator Study

ABSTRACT:Operating speed profiles are widely used to quantitatively measure roadway design consistency. The majority of early studies modeled the speed profiles based on a sequence of spot measurements. Such an approach may produce oversimplified speed profile estimates if the road alignment is complex; mountainous freeways are a good example. Speed measurements with instrumented vehicles, whether with dedicated or regular drivers, was a considerable improvement. Longitudinal speed measurements are typically frequent and they require an analysis that accounts for considerably correlated measurements. This paper proposes such a method of analysis of data obtained in a driving simulator with longitudinal speed measurements. A 24 km long mountainous freeway designed in Hunan, China, was modeled as simulation scenario and driving simulation experiments were carried out to collect speeds of individual vehicles on this road prior to construction. Grade and curvature related variables extracted from adjacent upstream and downstream segments were used to predict speeds using linear mixed models with spatial error covariance structure. The optimal range to extract geometric parameters (400m) and the best covariance structure (spherical) were determined. The results showed that larger average grades, larger average curvature, steeper curves on the downstream segment and existence of reverse curves would reduce drivers’ speed. An out-of-sample test of the speed model showed that the overall prediction MAE (Mean Absolute Error) is 3.19 km/h and RMSE (Root Mean Square Error) is 4.30 km/h, which is relatively lower compared to previous research. It was concluded that the linear mixed speed models corrected for spatial autocorrelation can be used to analysis effect of geometric parameters on the operating speed for mountainous freeways with complex alignments and, and it may become a practical model to predict speed profiles and further to evaluate design consistency.

Xuesong Wang*, Qiming Guo, Andrew Tarko. Analyzing Speed on Mountainous Freeways with Longitudinal Data – A Driving Simulator Study. Transportation Research Board 97th Annual Meeting, Washington D.C., USA, 2018. 1.7-11.