| dc.description.abstract | This thesis studies the optimization of the Flexible Job Shop Scheduling Problem
(FJSSP) in semiconductor manufacturing through the application of Integer
Programming (IP), Constraint Programming (CP), Hybrid Genetic Algorithm Simulated Annealing (GA-SA), and Deep Reinforcement Learning (DRL). The paper
evaluates the efficacy of these techniques in improving scheduling efficiency through
a comparative analysis among numerical, medium, and large case sets. While DRL
shows promise in managing complex events despite sporadic poor results, CP is rather
adept at providing quick answers. GA-SA offers a harmonic mix between the efficiency
and quality of the solutions. The results highlight in this industry the significant social,
environmental, and financial benefits of implementing advanced scheduling
techniques. This work suggests that future research should prioritize hybrid models that
combine DRL and CP to improve flexibility and efficiency. The practical consequences
of implementing these models strategically within semiconductor manufacturing are to
maximize output and resource use, hence supporting a sustainable industr | en_US |
