Recent Posts



Powerful Origin-Destination Matrix Estimation (ODME) Package Released

The ODME techniques attempt to improve the accuracy of a “seed” trip matrix based on information from fragmentary data such as traffic counts, surveys, etc. They range from simple proportional fitting to sophisticated Maximum Likelihood and Linear Optimization techniques to estimate the best trip matrix consistent with the observations available and the criteria selected. The ODME process could be an integral component in establishing the baseline demand for a Dynamic Traffic Assignment (DTA) model.

Our newly released ODME package is based on rigorous academic research by the DynusT team. The solution algorithm solves a mathematical optimization problem in which the objective function is to minimize the difference between the simulated and actual volumes on selected links for three separate classes – single-occupancy-vehicle (SOV), high-occupancy-vehicles (HOV), and heavy vehicles (trucks).

The ODME process is summarized in Figure 1, with the bi-level mathematical optimization problem in which the upper level is to adjust the demand of individual OD pairs, and the lower level is reaching a DUO equilibrium. In other words, the ODME would solve for a new OD matrix, then feed the new OD matrix into DynusT to run through iterations to establish DUO, then feedback to the upper level to solve the ODME problem again. This inter-play of the ODME and DynusT continues until the stopping criteria are met. Solving the optimization problem ensures that the solution approach will improve the matching of the simulated and actual counts through iterations.

Figure 1: Workflow of DynusT ODME Software

Figure 2 illustrates a 45-degree scattergram reflecting the convergence of a sample HOV demand trip table before and after ODME, measured by the comparison of the DTA model runs to observed Average Daily Traffic (ADT). The DTA runs were performed in a large-scale network with (45,739 links, 22,955 nodes, 5,263 zones, 20+ millions vehicles). Running through the entire ODME step for all three vehicle types and up to 10 user classes takes about merely 4-5 hours for this mega-scale network, which is rather computationally efficient compared to other competing methods.

This package is a free add-on with the DynusT/DynuStudio Pro Version subscription license. For more details about the ODME package, please feel free to contact the DynusT team using the contact form on the DynusT website.

Figure 2: 45-degree line Scattergram showing improved matching with screen line counts

The technical detail of the methodology can be found in:

Hu, XB; Chiu, YC;, Villalobos, J., Nava, E. (2017) "A Sequential Decomposition Framework and Method for Calibrating Dynamic Origin—Destination Demand in a Congested Network", IEEE Transactions on Intelligent Transportation Systems, Vol 18(10), pp2790-2797.