Rule-based modelling for adaptive complex infrastructure planning

Dealing with uncertainty in infrastructure planning is a challenge. Planning decisions need to be made in face of unknown future condition, and, in the meantime, it is essential that they are flexible enough to be adapted as new information unfolds. This indeed is important for multi-sector decision making where the complexity of the interconnected system and the uncertainty thereof hinders the modelling and analysis. Multistage stochastic optimisation provides a mechanism to incorporate these two attributes into planning decisions. However, its expensive computation as well as the appropriateness of its sequential decisions beyond the first few stages reduce its implementability. We introduce `Decision Rule' as a way to approximate the multistage problem, where the decisions at each stage are functions of the system complexity and the future uncertainty. We introduce a family of linear, polynomial, conditional if-then based rules and show how they approximate the multistage stochastic problem. We investigate their implications for urban water demand and supply network planning problem. Further we discuss some state-of-the-art and emerging tools for increasing the accuracy of the rules.