Estimating non-linear utility functions of time use in the context of an activity schedule adaptation model

Progress in activity-based modeling has recently focused on scheduling and rescheduling decisions (e.g. Gärling et al., 1999). In contributing to this line of research, the authors suggested a comprehensive model, called Aurora, of activity rescheduling decisions as a function of time pressure (Timmermans, et al., 2001). While the original paper focused on duration adjustment and schedule composition, later the proposed theory was elaborated and extended to include many different facets of activity rescheduling behavior (Joh et al., 2001, 2002). Numerical simulations supported the face validity of the model. Given the potential of the model, the next phase in the research project is concerned with the estimation of this complex, nonlinear model. This paper develops and tests an appropriate estimation method for the model, using a combination of theory and dedicated genetic algorithms. Results of numerical experiments are discussed. The paper first summarizes the theory underlying the model. Next, a method to estimate the model is proposed. The properties of this method are explored numerically using simulated data. The estimation method is tested using both perfect and noisy data. Finally, some conclusions are drawn and avenues for future research are suggested.