Observed-link-flow-based OD matrix estimation with stochastic factors

A new iterative procedure with the consideration of the integrating stochastic factors is established for estimating origin-destination (OD) matrices. The analytical form of the link choice proportions is developed. By minimizing the squared errors between estimated and observed link flows, the mathematical model balances all degrees of approximation for restricted equations. The recursive least squares method is conducted to evaluate OD matrices based on the estimation theory. The conducted method is more receptive to fluctuations in the observation due to integrating stochastic disturbances including the measurement errors and temporal variations of the observed link flows, and can make better use of traffic information from observation systems. The analysis and simulation study show that the method is feasible. Factors influencing the computed results of OD matrices are discussed and directions for future researches are proposed.