Robust parameter design for system dynamics models: A formal approach based on goal-seeking behavior

Abstract

Parametric uncertainties in system dynamics models can cause undesirable behavior, and if unaccounted for in the design phase can impair the ability of the system to meet design requirements. Sensitivity analysis, while useful in the modeling process, does not by itself offer a systematic method to the problem of system design under uncertainties. In this work, we propose an optimization model-based approach with the aim of obtaining a parametric design for system dynamics models that is robust against uncertainties. Our research synthesizes recent developments in robust optimization technology, eigenvalue analysis and the goal-seeking behavior of decision agents. To this end, we develop a mathematical model that is computationally efficient to solve and effective in hedging against parametric uncertainties. Numerical case studies conducted using a hare and lynx model and an inventory-workforce model demonstrate significant improvements of the proposed designs in achieving system requirements under uncertainty. © 2012 System Dynamics Society.