Segmented regression analysis for estimation of traffic characteristics - application to local data, section data and information derived from position reports

Abstract

Infrastructure-based traffic data are continually available, but their spatial explanatory power is limited. Positioning data delivered from a vehicle fleet may be used to derive link-related speeds for a complete road network, but they are usually only sporadically available. This paper describes a new regression-based approach using historically observed interdependencies between various traffic characteristics and currently available traffic data for a network-wide traffic state estimation. Hence, the method combines the complementary advantages of road-based and vehicle-based data detection. The approach enables the integration of several types of relevant traffic data, and the handling of incomplete data with variable accuracy. It has been successfully tested in a section of the road network in Munich.