An adaptive robust optimization model for parallel machine scheduling

Real-life parallel machine scheduling problems can be characterized by: (i) limited information about the exact task duration at time, and (ii) an opportunity to reschedule remaining tasks each time a processing is completed becomes idle. Robust optimization natural methodology cope with first characteristic of uncertainty, yet existing literature on robust does not explicitly consider second possibility adjust decisions as more available, despite that re-optimizing schedule every new emerges standard practice. In this paper, we develop adaptive approach takes into account, beginning planning horizon, adjusted. We demonstrate suggested lead better here-and-now makespan guarantees. To end, mixed integer linear programming model for scheduling, two-stage approximation heuristic, where minimize worst-case makespan. Using model, show via numerical study leads solutions stable realizations compared static approaches.