A Combined Meta-Heuristic with Hyper-Heuristic Approach to Single Machine Production Scheduling Problem

HIS work is motivated by the scheduling problem of a real world single machine production encountered in a metal industry. The objective function to consider is the minimization of mean tardiness and flow time. This problem belongs to the class of difficult problems (NP-complete). Due to the dynamic and the difficulty for searching the solution, deterministic searching methods do not work effectively when the problem size is getting bigger. Most techniques are domain-specific, which means that their applications are fit rather to specific than to general problems. The performance of the algorithm can be drastically reduced if there is a change in the problem being modeled. Unfortunately, real problems change dynamically and rapidly by nature. This lead to the need for a technique that is easily adapted to a variety of changes. Hyper-heuristic is a methodology that has multi-level heuristics, in which a high level heuristic coordinates lower level ones [1]. This algorithm provides searching framework that more general and non domain-specific. Hyper-heuristic methodology is more flexible in the search process and can be easily applied to a larger scope of issues [2]. This construction of this method is motivated by the need for flexible search techniques that can be easily adapted to respond to changes and free of domain-specific problems. This technique does not directly conduct a search on the solution space, but prior to the heuristic space. In this work, we compare two variants of genetic algorithm as meta-heuristic that are combined with hyper-heuristic approach to solve a real single machine scheduling problem. In the first variant, Genetic Algorithm is used as the high level heuristic to choose some low level heuristic (MRT, SPT, LPT,