A Least Squares Method for Origin-destination Estimation Incorporating Variability of Day-to-day Travel Demand

The estimation of an origin-destination (O-D) trip matrix from observed link flow is an essential part of the 4-step transport planning process. Conventional methods for estimating O-D trip matrices usually consider the fluctuation of link flows observed from loop counts as an error term that can be ignored, resulting in a significant loss of information. This work, instead, explicitly considers the fluctuation of link flows as a result of the volatility of demand. A strategic user equilibrium (StrUE) model is implemented, which aims to minimize the expected travel time, assuming users have learned the actual demand distribution. Expressions for the analytical link travel times and corresponding variability are derived from the StrUE model. This information is utilized in the least squares method proposed in this paper. The objective of the proposed model is therefore to estimate a demand distribution that minimizes the distances between the two terms: 1) the mean of the observed link flows and the estimated mean of the link flows; 2) standard deviation of observed link flows and estimated standard deviation of the link flows. A solution algorithm is proposed, and results are presented from numerical experiments are implemented on a virtual network.