The Use of Markov Decision Processes for Australian Naval Aviation Training Schedules

Abstract

Optimization of pilot training schedules has unique features, including high costs, high and highly variable failure rates, small numbers of trainees, the need to maintain a number of available graduates in squadrons, and, particularly, the requirement to use graduates as course instructors. Currently in the Royal Australian Navy, a heuristic short planning horizon is in use to estimate recruitment numbers to retain operational capability into the future. In this paper, we explore how a Markov decision process (MDP) model can be applied and qualitatively compare it with shorter-term integer linear programming. Results demonstrate the robustness of the MDP formulation.