| In this thesis, a modified second-order continuum model is used to describe the traffic behaviour along highways. The model is identified and verified using several sets of traffic measurements collected from a major highway in metropolitan Toronto, Canada.; A robust nonlinear sliding mode observer is developed to generate estimates of average velocity and density for a segment of a highway within a corridor, given loop detector measurements at the end-points of the segment. The sliding mode approach has several advantages over other estimation techniques such as the Kalman Filtering including proof of estimate convergence and simplified computations. However, the primary advantage is the robustness of the observer with respect to unmodelled dynamics and disturbances. Unmodelled dynamics are associated with the traffic factors whose effects cannot be captured (properly) in the traffic flow models, e.g., road geometry and weather conditions. On the other hand, model disturbances such as unavailable (not measured) traffic flow at a ramp or measurements provided by a faulty detector can also create unpredictable traffic states. Based on the presented traffic model, a systematic design procedure is developed to make the observer robust with respect to the modelling uncertainties and unavailable traffic states. Simulation and experimental results show the effectiveness of the proposed observer in estimating the states of a highway traffic system.; Moreover, a new decentralized state feedback linearizing controller for ramp metering using variable structure control is presented. The main aim is to develop a robust controller to locally stabilize freeway traffic despite the presence of disturbances and modelling errors. Simulation results show that the proposed controller provides improved performance in achieving the design objectives over other existing ramp control strategies such as neural network and linear feedback controllers. |
