UC Davis

Real-time traffic information is crucial for traffic operations and management. Due to budget constraints, strategies for deploying sensors to effectively acquire the most critical information are desirable. While significant progress has been made on sensor location problem based on flow observability concept, most research on flow observability is based on static traffic data and only takes network topology into account. This thesis addresses the traffic network observability problem in a dynamic setting. Different from the observability concepts adopted for static settings, we consider all temporal and spatial relations of traffic flows to determine unobserved states. We first develop a state-space model to describe link density dynamics in virtue of a recently proposed link queue model. We then analyze the link queue model properties including phase transition, subspace boundary variation, and stability. Finally, based on full rank test or PBH test, we use examples to show critical network components where data need to be collected to ensure full network observability under different topology and congestion conditions.