Improving The Estimation Of Travel Demand For Traffic Simulation

Improving the Estimation of Travel Demand for Traffic Simulation

Many current traffic management schemes are tested and implemented using traffic simulation. An Origin- Destination (OD) matrix is an ideal input for such simulations. The underlying travel demand pattern produces observed link counts. One could use these counts to reconstruct the OD matrix. An offline approach to estimate a static OD matrix over the peak period for freeway sections using these counts is proposed in this research. Almost all the offline methods use linear models to approximate the relationship between the on-ramp and off-ramp counts. Previous work indicates that the use of a traffic flow model embedded in a search routine performs better than these linear models. In this research, that approach is enhanced using a microscopic traffic simulator, AIMSUN, and a gradient-based optimization routine, MINOS, interfaced to estimate an OD matrix. The problem is highly non-linear and non-smooth, and the optimization routine finds multiple local minima, but cannot guarantee a global minima. However, with a number of starting "seed" matrices, an OD matrix with a good fit in terms of reproducing traffic counts can be estimated. The dominance of the mainline counts in the OD estimation and an identifiability issue is indicated from the experiments. The quality of the estimates improves as the specification error, introduced due to the discrepancy between AIMSUN and the real-world process that generates the on-ramp and off-ramp counts, reduces.

Improving the Estimation of Travel Demand for Traffic Simulation

This report examined several methods for estimating Origin-Destination (OD) matrices for freeways using loop detector data. Least squares based methods were compared in terms of both off-line and on-line estimation. Simulated data and observed data were used for evaluating the static and recursive estimators. For off-line estimation, four fully constrained least squares methods were compared. The results showed that the variations of a constrained least squares approach produced more efficient estimates. For on-line estimation, two recursive least squares algorithms were examined. The first method extends Kalman Filtering to satisfy the natural constraints of the OD split parameters. The second was developed from sequential quadratic programming. These algorithms showed different capabilities to capture an abrupt change in the split parameters. Practical recommendations of the choice of different algorithms are given.

Annual Report