A short-cut method for strategy optimisation using strategic transport models

This paper describes a methodology which permits optimal strategies for strategic transport models to be found by use of a limited number of model runs togther with regression modelling of the resulting response surface. Typically, it will be the case that the number of policy variables is sufficiently large that the strategic model cannot be run for all possible combinations of their levels. Furthermore it can be very difficult to interpret the results from a large number of model runs where there are a lot of policy variables changing levels between runs. The proposed methodology models the response surface specifically in the locality of the optimum, thereby greatly clarifying what policy combinations should be further tested with the strategic model. A case study, for the city of Edinburgh, indicates that this methodology can identify improved strategies compared to conventional methods, even when the number of model runs used are far fewer than with the conventional methods.