This paper introduces two Kalman filter (KF) implementations for the estimation of turning rate at signalized intersections, assuming the availability of Floating Car Data (FCD). The developed filters are tested in microscopic simulation under various scenarios. The first filter is a typical (linear) KF that utilizes only real-time FCD. The second is an Extended (non-linear) Kalman filter that utilizes both real-time and offline (historical profiles) FCD. The aim of this paper is to demonstrate how Kalman filter can be used to estimate in real-time the total turning rate at an intersection, based on limited (and error-prone) measurements coming from FCD. Moreover, it is shown how to tune the filters to account for different errors in real-time measurements and in historical profiles, in order to use it in real-world applications.     «

This paper introduces two Kalman filter (KF) implementations for the estimation of turning rate at signalized intersections, assuming the availability of Floating Car Data (FCD). The developed filters are tested in microscopic simulation under various scenarios. The first filter is a typical (linear) KF that utilizes only real-time FCD. The second is an Extended (non-linear) Kalman filter that utilizes both real-time and offline (historical profiles) FCD. The aim of this paper is to demonstrate...     »