Dynamic Estimation of Traffic State on Surface Street Network

Reliable traffic information is essential for the development of efficient traffic control and management strategies for traffic networks in advanced traffic management systems (ATMS) and advanced traveler information systems (ATIS). Real-time traffic information is utilized for various purposes such as dynamic route guidance, incident detection, and control of variable message signs. Traffic information can be obtained from several vehicle detectors installed on the road network. However, using only detectors cannot give us a complete view of current traffic state on the whole network since detector can measure traffic state only at a point where it is installed. The mathematical model of traffic flow such as the macroscopic model (e.g., the LWR model proposed by Lighthill and Whitham and Richards) can then be used to connect the local view of measured traffic state from all detectors to the estimated complete view for the whole network. Nevertheless, when the network is complicated and has some uncertainties as in the case of surface street, the traffic flow model alone may sometimes fail to replicate the real traffic situation. Moreover, detector data may also include some noises due to the measurement error. To compensate for these problems, several adaptive filtering techniques such as the Kalman filter (KF) and its extension (extended Kalman filter: EKF) were introduced and integrated into the macroscopic model for real-time estimation of traffic state. The objectives of this paper are twofold: 1) to model the surface street network using discretized macroscopic model and 2) to evaluate the effectiveness of integrating the Kalman filter with macroscopic model. In chapter two, the macroscopic model and its numerical method for the case of surface street network are discussed. Third, the brief outline of how to integrate the EKF with the macroscopic model is given. In chapter four, the method is tested on the sample network and compared with the results obtained from simulation software. Finally, the conclusion is drawn and the area for future research is discussed.