Effective Stochastic Local Search Algorithms For Bi-Objective Permutation Flowshop Scheduling

In this report, we study stochastic local search (SLS) algorithms for biobjective optimization problems. In particular, we study multi-objective algorithms that belong to two different classes of approaches, those that use a scalarized acceptance criterion (SAC) and those that use a componentwise acceptance criterion (CWAC). As a representative example of a SAC algorithm, we study Two-Phase Local Search (TPLS). We propose an improved version of TPLS that has very good anytime properties: it achieves an as good approximation to the Pareto front as possible independent of the computation time, and thus, can be stopped at any moment while providing very good results, improving in this way over original TPLS versions. We believe that hybridization of algorithms belonging to the SAC and CWAC classes can result in more performing algorithms. We design such a hybrid algorithm that combines our improved version of TPLS and Pareto Local Search (PLS), a representative example of a CWAC algorithm. We use as test problems bi-objective Permutation Flowshop Scheduling Problems (PFSP), a widely studied class of scheduling problems. We carefully study each component of TPLS and PLS and their effect on the final solution quality. By comparing the hybrid algorithm with the previous state-of-theart, we show without ambiguity that our algorithm greatly improves upon the best ones previously known for bi-objective PFSPs.