A perennial goal of transport modellers has been to increase the efficiency of the demand model calibration process, this is because it requires the heavy use of normally scarce resources both in the data collection and parameter estimation stages. On the other hand, methods to re-estimate models in a continuous modelling framework in order to monitor plan performance and re-evaluate alternative options if necessary are also of great importance nowadays. This paper describes a general framework which allows the incorporation of different data sources in order to improve model estimation both in terms of accuracy and efficiency. The framework also allows re-estimation of models using new available data. The approach re-estimation of models using new available data. The approach has been developed to accommodate the most commonly used transport demand models, and the most typically available data sources. The main idea is to incorporate all the model functions involved in a joint likelihood function in order to estimate their parameters simultaneously using all the available data sets. Some alternatives to this method, such as a generalised least squares approach, are also discussed. The paper ends by discussing an example related to the updating of a joint destination-mode choice model, using prior model parameters and new information in the form of traffic counts and aggregate modal split information. Some illustrative simulation results are also shown.
Samenvatting