Abstract: A dynamic Origin-Destination (O-D) matrix model based on least square approach is proposed for urban rail transit using moving-average strategy. Statistic analysis of historic automatic fare collection records shows that, in continuous periods, there is a strong average O-D flow relevance between priori passenger flow and current passenger flow. Therefore, a dynamic O-D matrix estimation model is established based on moving-average strategy. This model can utilize the priori flow information more efficiently. A passenger flow arrival coefficient based on the distribution of O-D travel time is defined to characterize the dynamics of traffic flow, thus establishing the relationship between O-D flow and in-and-out flow at stations. Finally, a case analysis of Beijing rail transit network is carried out. Results show that, comparing with existing model, the accuracy of the proposed model is improved about 15%~20% from the viewpoint of average relative deviation in a day. The proposed model can improve the accuracy of dynamic O-D matrix estimation in short time range for urban rail transit greatly.

Key words: engineering of communications and transportation system, dynamic O-D matrix estimation, the least square approach, urban rail transit, moving-average strategy