UC Berkeley

Classical macroscopic traffic flow models, such as the Lighthill-Whitham-Richards (LWR) model, assume instantaneous vehicle velocity adjustment to local density and a unique fundamental diagram relationship. These simplifications limit their ability to capture real-world traffic phenomena observed in empirical datasets. This thesis develops and validates a novel macroscopic model that bridges theoretical concepts with empirical traffic observations through incorporation of non-local effects and degenerate diffusion. We establish calibration procedures linking model parameters to traffic flow data, demonstrating improved predictive accuracy particularly in congested regimes. We also develop extensions to two-dimensional networks, enabling simulation of urban transportation systems with arbitrary topology and present algorithms for generating unstructured meshes for 2D traffic flow simulation. This work establishes comprehensive methodology connecting theoretical advancement in macroscopic traffic flow modeling with computational efficiency and empirical validation, advancing the state-of-the-art in traffic simulation and providing tools for modern transportation management applications requiring forecasting and network-level performance prediction.