Abstract
The estimation of real-time Origin-Destination (OD) matrix for large network under data scarcity environment is presented. It focuses on four significant issues associated with data scarcity: 1) reduction of unknown variables by replacing OD flows with origin and destination flows; 2) consideration of noise in unknowns 3) development of a simulation algorithm to map unknowns and measurements precisely; 4) incorporation of speed data during feedback process. The approach is structured under unscented Kalman filter framework, featured by (a) a joint estimation model for estimating OD flows and route choices; (b) more sophisticated simulation model for mapping. It is empirically validated through a case study.
