Abstract
This research presents a multi-objective policy design based on simulating system dynamics, a simulation technique capable of explicitly modelling the feedback loops of decision rules and evaluating the dynamics of complex processes and systems. The novel feature of our approach is that performance is not measured by a single value, but rather performance measures are optimised based on their trajectories, such as the degree of inventory oscillation and the amplification ratio between the order rates of two parties over time (e.g. the bullwhip effect). A multi-objective genetic algorithm termed NSGA-II is employed to generate a set of nondominated solutions. In order to demonstrate the performance of our approach, we adapt and evaluate the dynamic method for a well-known case study: the beer game model of a two-stage supply chain.