Robust optimisation of forest transportation networks: a case study

Abstract

Forest transportation costs are the major cost component for many forest product supply chains. In order to minimise these costs, many organisations have turned to optimisation models to guide decisions that are extremely complex in nature. These models generally assume that input parameters are known with certainty, but in reality they are often associated with a high degree of uncertainty. One way of dealing with uncertainty is through robust optimisation, a procedure which is capable of generating near-optimal solutions that are relatively unaffected by the surrounding uncertainty. We illustrate this by means of a case study that employs a robust optimisation procedure. Our procedure employs a two-phase approach. The first phase creates a more tractable problem by limiting the search to those solutions that are near optimal and feasible. The second phase simulates the affect of uncertainty on the solutions isolated in the previous phase. The simulation results are then evaluated for robustness by means of seven robustness performance measures. For our case study, the results show that (1) the deterministic solution is extremely unstable and highly reliant on a particular degree of uncertainty, that (2) the robust solution is dependent on the robust performance measure selected, and (3) that the true robust solution is different from the deterministic solution.