Systematic inference of online urban travel demand: decomposition, observability, and error correction

Origin-destination (OD) demand matrix plays an essential role in performance assessment and traffic management of road networks. While existing real-time models for time-varying OD estimation offer promising solutions, their applicability could be constrained by prior OD database development, insufficient system observability modelling, or suboptimal solving procedures. This study presents a systematic modelling framework to tackle these key challenges. Under this framework, a robust structural decomposition scheme is proposed, building upon which the modelling of dynamic traffic assignment and dynamic OD estimation is investigated. To ensure good convergence of the state space models, the system observability problem within the dynamic OD estimation context is properly addressed. Finally, two kinds of state space models with a solving procedure enriched by explicit error statistical analysis and adaptive error correction are developed. A real-world urban network in the downtown area of Kunshan, China was utilised to demonstrate the potential of the proposed framework.