Taking deep uncertainty into account in traffic models

Abstract

Covid-19 has proven how uncertain the future can be. However, traffic models fail to fully acknowledge this uncertainty. Therefore, decisions made based on these models are not very robust. In this research, Robust Decision Making (RDM) method is applied on a macro-level traffic model made for the municipality of Groningen. This method, consisting of 5 steps, can make the decision making process more robust. First RDM frames the system in which the decision is taken. Then, within the boundaries of the system, the model in run on a large set of possible futures. These futures are picked based on Latin hypercube sampling. Finally, with the results obtained, scenario discovery and global sensitivity tools can be applied. These tools help to find regions of interest within the sample space and to analyse how different variables influence each other. Thereby, based on robustness metrics the most robust policy can be chosen. With the application of RDM in traffic models good policies and interesting scenarios are found. However, due to the fact that traffic models are often over fitted, the run time of the model is quite long and the level of interaction quite low. This makes the application of RDM both difficult and hard to fully exploit. The discussion of this research hands some possible tools to manage the run time. Multi resolution modelling and the use of providers like Amazon web services are two of the most prominent options. Thereby, more in-depth deep uncertainty, like Multi Objective RDM (MORDM) methods are possible when the run time can be better managed.