The estimation of future freight needs requires the use of network and freight demand models. When characterizing freight demand, basic data are sought to appropriately model the decision processes associated with freight generation, distribution, and consumption. In this context, freight origin-destination (OD) matrices are one of the most important data a planner could have, which is why a significant amount of effort, time and money is spent on their estimation. The estimation of OD matrices can be done by: (a) direct sampling methods; and, (b) using secondary data sources such as traffic counts. The latter techniques are referred here as OD synthesis (ODS).

Direct sample estimation includes all methodologies in which the OD data are obtained by interviewing the users. These approaches have some well known limitations: roadside interviews tend to double count trips; on board interviews may lead to bias in the parameters of random utility models; mail interviews are often biased because the rate of response varies across the population; and home interviews, though able to provide statistically sound estimates of OD, require a great deal of planning, time, effort and money (Ortuzar and Willumsen, 2001).

ODS overcomes these limitations by bypassing the need for surveys. In ODS, the traffic counts "which are a function of the OD flows" are used to estimate the OD matrices. Although there are hundreds of papers on passenger ODS, only a handful deal with freight (Tamin and Willumsen, 1988; Gedeon et al., 1993; List and Turnquist, 1994; Tavasszy et al., 1994; Al-Battaineh and Kaysi, 2005; Holgu??n-Veras and Patil, 2007; Holgu??n-Veras and Patil, 2008). This highlights the importance of funding research in this important subject as the increasing availability of GPS and ITS data provide unique opportunities for transformative and innovative contributions to practice.

As part of a National Science Foundation project, team members developed new formulations to conduct freight ODS, and a novel tour based freight demand model. The freight ODS formulations use a gravity model to estimate commodity OD flows, and an empty trip model based on a simplified trip chain model to approximate tour behavior. The first formulation considered a single generic commodity (Holguin-Veras and Patil, 2007), while the second considered multiple commodities. The performance of these formulations was assessed using actual OD data. The numerical tests showed a reduction in the total summation of errors of 29% to 40% with respect to alternative models. A key limitation of these models is that they require, as an input, the amount of cargo produced and attracted by each zone.

The objective of the proposed work is to enhance this line of work by expanding the formulations to consider the case in which the amount of cargo produced and attracted by each zone is unknown, or there is uncertainty in their estimation. Among other things, this will enable MPOs to produce quick estimates of freight OD matrices on the basis of traffic counts. The proposed project would focus on the development of the required mathematical algorithms to produce such estimation. Among other things, the availability of such methodologies will facilitate integration of freight into the metropolitan transportation planning process.