The estimation of the time profile of an origin-destination matrix in the presence of unusual demand

Origin-Destination (OD) matrix is the fundamental information in the planning and management of a transportation system. When unusual traffic demands occur due to holidays, sport events, demonstration etc., it is even more important to have an updated OD matrix so that for short term period, traffic engineers can use it to design temporary traffic control while transportation service providers can plan and allocate their rolling stocks, bus fleets, and employees more efficiently so as to provide reliable and acceptable service to meet the public demand and at the same time maximize their profits. For medium to long term period, urban planners and transportation engineers can use the updated OD matrix to look for ways to improve the services and facilities of an area, and make appropriate decisions for the development of better infrastructure in a bid to release the stress on congested roads. Traditionally, the conventional 4-step transportation model is used to estimate the OD matrix. This approach is based on a large-scale household or roadside survey, and thus, the cost is very high and the survey can only be conducted once in every five to ten years. As a result, the OD matrix has long been treated as a piece of information that is supposed to be fixed for a long period of time, only to be updated infrequently. Therefore, for many years, transportation planners and traffic engineers are forced to work with such an out-dated OD matrix. The methodologies developed over the last twenty years using traffic counts on links to update the OD matrix has alleviated this problem somewhat but it still cannot solve the problem completely. It is because all these methods are based on the principle of optimizing the matching between the observed and the predicted link flows and they all ignore the changes that would have occurred in the urban forms of the OD pairs such as trip generation rates, social demographics and land use systems and hence with limited predictive power. Different from other approaches that use traffic counts to directly update the OD matrix, this paper proposes to express the OD flows in terms of demographic variables and urban forms first. Adopting from the energy industry the methodology of Conditional Demand Analysis (CDA) for forecasting electricity demand, this paper uses the CDA and Bayesian approach to estimate the trip generation rates and the time profile of the OD matrix. This approach not only allows the study of the effects of the exogenous variables (including factors due to unusual events) on the OD flows and hence updates the OD matrix accordingly, it also guarantees the existence of the estimates by reducing substantially the number of unknown parameters used in the models. The updated time profile of the OD matrix can also be used to forecast hour to hour traffic demand and link flows. Findings should be useful to traffic engineers, public transportation service providers, urban planners, and transportation engineers for transport management, especially in the presence of an unusual increase in traffic demand.