A Kalman Filter for Quasi-Dynamic o-d Flow Estimation/Updating

This paper proposes an extended Kalman filter for quasi-dynamic estimation/updating of o-d flows from traffic counts. The quasi-dynamic assumption—that is considering constant o-d shares across a reference period, whilst total flows leaving each origin may vary for each sub-period within the reference period—has been proven already realistic and effective in off-line o-d flows estimation using generalized least squares estimators. The specification of the state variables and of the corresponding transition and measurement equations of a quasi-dynamic extended Kalman filter are illustrated, and a closed-form linearization is presented under the assumption of an uncongested network and error-free assignment matrix. Results show satisfactory performance and parsimonious computational burden on real-size networks