Development and Tech Transfer of an Integrated Robust Traffic State and Parameter Estimation and Adaptive Ramp Metering Control System

Timely updated knowledge of traffic state and traffic flow parameters is essential for effective traffic control. Because of its simplicity, the Cell Transmission Model (CTM) has been widely used to serve as the underlying traffic flow model based on which a traffic state estimation scheme is built. A prominent feature of CTM is that for a given road cell, CTM switches between two modes: free-flow mode and congestion mode. For each mode, the traffic flow dynamics are linear in traffic density. As a result, most existing traffic state estimators that were derived from CTM switch among two linear working modes, and hence Kalman filter (KF) or ensemble Kalman filter (EnKF) can be applied. These existing estimators have assumed that traffic flow parameters including the free-flow speed and the critical density are known and time-invariant. However, in reality, these traffic flow parameters can be time-varying rather than fixed-valued.

Dr. Zhou and Dr. Ozbay found that, if the traffic flow parameters are time-varying and/or the knowledge of these parameters are biased, the performances of a traffic state estimator that has assumed them to be known and fixed-valued can be significantly downgraded. Moreover, only augmenting these parameters into the state vector and then resorting to nonlinear recursive estimation techniques such as extended Kalman filter (EKF) cannot solve the issue. This is because, under a CTM-based traffic estimator, the critical density is unobservable under free-flow conditions, and hence biased initial knowledge of the critical density can cause false switching of the working model of the estimator and distort the estimation afterward.