DATA-DRIVEN MODELS FOR SCHEDULING OPTIMIZATION UNDER UNCERTAINTY

Abstract

As a critical decision-making tool across a wide range of industries, scheduling aims to allocate limited resources to jobs over time optimally. During execution, however, the scheduling system is subject to considerable uncertainty which may cause infeasibilities and disturbances. This thesis proposes three different scheduling models which respectively cater for three well-known sources of uncertainties in practice. First, a data-driven scheduling optimization model using Renyi mean-entropy-skewness information criterion is developed to deal with resource cost uncertainties. Second, based on the concept of activity duration tolerance levels, a due-date achievement model using a proposed performance measure termed the activity exposure level is formulated for resource-constrained activity scheduling under uncertain activity durations. Third, a scheduling model is formulated to tackle uncertainties in resource disruptions by optimizing the threshold scenario, bounded by which the planned due-dates can be achieved.