Flexible Job-Shop Scheduling Problems

Planning and scheduling problems in various industrial environments are combinatorial and very difficult. Generally, it is extremely hard to solve these types of problems in their general form. Scheduling can be formulated as a problem of determining the best sequence to execute a set of tasks on a set of resources, respecting specific constraints like precedence or disjunctive constraints (Carlier & Chrétienne, 1988). They consist generally in a simultaneous optimization of a set of non-homogeneous and conflicting goals. Therefore, the exact algorithms such as branch and bound, dynamic programming, and linear programming are not suitable for such problems and need a lot of time and memory to converge. Because of this difficulty, experts prefer to find not necessary the optimal solution, but a good one to solve the problem. To this end, new search techniques such as genetic algorithms (Dasgupta & Michalewicz, 1997; Sarker, Abbas & Newton, 2001), simulated annealing (Kirkpatrick, Gelatt & Vecchi, 1983), and tabu search (Golver, Taillard & De Werra, 1993) are proposed to reach this aim: construct an approximated solution for a large set of hard optimization problems. In this article, we are interested in the evolutionary techniques and their application to an important branch of scheduling problems. We aim in particular to present an overview of the recent models proposed to solve flexible job shop scheduling problems using genetic algorithms.