Abstract
Traffic state estimation is an essential component for Dynamic Traffic Management (DTM). This thesis develops a Lagrangian multi-class traffic state estimation method, which offers both computational and theoretical advantages over the conventional Eulerian method, as well as providing timely, accurate and reliable class-specific traffic information for DTM at a network scale. The data pre-processing methods that were developed improve both model and observation inputs, and thus can additionally benefit real-world traffic state estimation. (Author/publisher)