ASSIGNMENT-MATRIX-FREE DYNAMIC ESTIMATION OF ORIGIN-DESTINATION MATRICES

Abstract

 This paper presents a new approach for dynamic es timation of origin-destina tion (OD) matrices based on Unscented Kalman Filter (UKF). The new approach, di ffers from most of the conventional approaches, does not assume linearity of the original nonlinear relationship between OD parameters and measurements. Instead, the nonlinearity is retained by representing th is relationship using dynamic traffic simulation. This eliminates the need to compute the assignment matrix an d offers some degrees of flexibility in selecting the traffic simulation model and incorporating speed as additional measurement variables. Preliminary results demonstrate the potential of the proposed approach and the contribution from using speed data.