Robust scheduling in forest operations planning

Abstract

Forest operations planning is a complex decision process which considers multiple objectives on the strategic, tactical and operational horizons. Decisions such as where to harvest and in what order over different time periods are just some of the many diverse and complex decisions that are needed to be made. An important issue in real-world optimization of forest harvesting planning is how to treat uncertainty of a biological nature, namely the uncertainty due to different growth rates of trees which affects their respective yields. Another important issue is in the effective use of high capital intensive forest harvesting machinery by suitable routing and scheduling assignments. The focus of this thesis is to investigate the effects of incorporating the robust formulation and a machinery assignment problem collectively to a forest harvesting model. The amount of variability in the harvest yield can be measured by sampling from historical data and suitable protection against uncertainty can be set after incorporating the use of a suitable robust formulation. A trade off between robustness to uncertainty with the deterioration in the objective value ensues. Using models based on industrial and slightly modified data, both the robust and routing formulations have been shown to affect the solution and its underlying structure thus making them necessary considerations. A study of feasibility using Monte Carlo simulation is then undertaken to evaluate the difference in average performances of the formulations as well as to obtain a method of setting the required protections with an acceptable probability of infeasibility under a given set of scenarios.