Scheduling Algorithms for Non-overlapping Setups Problem- Acronym

This thesis examines the problem of task scheduling on parallel identical machines with setup operation for every pair of scheduled consecutive tasks. The length of the setup is sequence-dependent, and the setup must be executed in-between the relevant tasks by a machine setter. On any machine at any given time, there can be only one task or setup processed. No machine setter can work multiple setups at the same time. The goal is to minimize the makespan, which is the time when the last machine ends. We propose Constraint programming (CP) models to find the optimal solution and heuristic algorithms that are suitable for large instances of the problem. We propose warm start techniques that greatly increase the CP effectivity. We propose a hybrid algorithm called SMETI, which combines the mentioned methods with sub-problem CP modeling to provide fast, good results while being able to reach optimum. Experimental comparisons to the state-of-the-art solutions of similar problems show the proposed approaches to be more effective, finding the same quality solution faster or providing a better solution at the same time limit.