Multirecombinated evolutionary algorithms to solve multiobjective job shop scheduling

Abstract

Multiobjective optimization, also known as vector-valued criteria or multicriteria optimization, have long been used in many application areas where a problem involves multiple objectives, often conflicting, to be met or optimized. Scheduling problems is one of such application areas whose importance lays on its economical impact and its complexity. The present paper propases CPS-MCPC, a cooperative population search method with multiple crossovers per couple. The cooperati ve search CPS is implemented with in di viduals of a single population, which are selected for recombination using alternatively each criterion. MCPC a multirecombination approach is used to exploit good features of both selected parents. To test the potentials of the novel method for building the Pareto front regular and non-regular objectives functions were chosen: the makespan and the mean absolute deviation of job completion times from a common due date (an earliness/ tardiness related problem). The set of experiments conducted, used three basic representation schemes and contrasted results of the proposed approach against conventional methods of recombination. Details of implementation and results are discussed.