Abstract
It is important to understand what kind of performance index should be used for performance-oriented highway planning and design. Trip length is one of the important indices to reflect highway function. The road design for traffic function should not be used by the drivers with purpose of short trip since those drivers may travel longer distance due to lack of entry/exit. On the other hand, road design for access function should not be used for drivers of long trip since vehicles have to slow down when entry in or exit from facilities. Short trips and long trips are mixed on the road due to the current urban planning along the road. In this paper, we approximate trip length distributions by applying Gaussian Mixture Model. The trip length distributions at five intersections along National Road 16 between Kasukabe City, Saitama and Shiroi City, Chiba are obtained from the probe data of October, 2015. The result shows that every distributions can be composed by six or seven normal distributions. We achieved high correlation, more than 0.9, between actual and estimated trip length distributions.
