Traffic state estimation (TSE) is used for real-time estimation of the traffic characteristics (such as flow rate, flow speed, and flow density) of each link in a transportation network, provided with sparse observations. The complex urban road dynamics and flow entry and exit on urban roads challenge the application of TSE on large-scale urban road networks. Because of increasingly available data from various sources, such as cell phones, GPS, probe vehicles, and inductive loops, a theoretical framework is needed to fuse all data to best estimate traffic states in large-scale urban networks. In this context, a Bayesian probabilistic model to estimate traffic states is proposed, along with an expectation–maximization extended Kalman filter (EM-EKF) algorithm. The model incorporates a mesoscopic traffic flow propagation model (the link queue model) that can be computationally efficient for large-scale networks. The Bayesian framework can seamlessly integrate multiple data sources for best inferring flow propagation and flow entry and exit along roads. A synthetic test bed was created. The experiments show that the EM-EKF algorithm can promptly estimate traffic states. Another advantage is that the EM-EKF can update its model parameters in real time to adapt to unknown traffic incidents, such as lane closures. Finally, the proposed methodology was applied to estimating travel speed for an urban network in the Washington, D.C., area and resulted in satisfactory estimation results with an 8.5% error rate.
Effect of Enhanced ADAS Camera Capability on Traffic State Estimation
