Fast local search for fuzzy job shop scheduling

Abstract

In the sequel, we propose a new neighbourhood structure for local search for the fuzzy job shop scheduling problem. This is a variant of the well-known job shop problem, with uncertainty in task durations modelled using fuzzy numbers and where the goal is to minimise the expected makespan of the resulting schedule. The new neighbourhood structure is based in changing the relative order of subsequences of tasks within critical blocks. We study its theoretical properties and provide a makespan estimate which allows to select only feasible neighbours while covering a greater portion of the search space than a previous neighbourhood from the literature. Despite its larger search domain, experimental results show that this new structure notably reduces the computational load of local search with respect to the previous neighbourhood while maintaining or even improving solution quality.