Real Time Estimation on Turning Rates of N-Leg Intersections

In urban traffic control, the estimation of time-varying turning rates in each entrance of intersection is important to improve system efficiency. This paper describes a stochastic modeling approach to estimate turning rates in entrances of N-leg intersection (N≥3) in real time. Based on the assumption that point detectors are set on exit lanes of intersections, this paper formulated a discrete-time stochastic model in which a compositive measurement, made up of exits flow measurement and the empirical turning rates, was used. To estimate turning rates, a recursive estimation algorithm was developed which primarily involved the application of a Kalman filter, trucation and normalization procedures. Preliminary model testing was used to explore the feasibility of estimating real-time turning rates using the proposed methdology. The result indicates that the estimation algorithm has a good convergency and the estimation errors are within the range of acceptability.