A robust multi-response VNS-aiNet approach for solving scheduling problems under unrelated parallel machines environments

Abstract

In the last few years, a lot of researchers addresses scheduling problems considering unrelated parallel machines to bridge theoretical problems and real-world manufacture environments. Most of these works consider objective function’s characteristics to assist the algorithms in exploiting the search space. This assistance usually leads to a better convergence of the algorithms, but it tends to leave these algorithms restricted only to the own optimization criterion studied. This feature hinders to extend these works to optimization criteria with less theoretical visibility. This work proposes an immuno-inspired algorithm for the resolution of scheduling problems under unrelated parallel environments, evaluating the hypothesis that it is possible to construct a robust approach that is extensible to different optimization criteria, achieving results similar to the state-of-art approaches in the literature for specific criteria. Four case studies are proposed, each one addressing an optimization criterion, which is a benchmark for sequencing problems. The criteria are Total Weighted Tardiness, Makespan, Total Tardiness, and Total Weighted Completion Time. The experiments indicate that the results achieved by the proposed approach are compatible and, in many cases, superior to the state-of-art approaches for the optimization criteria studied, using computational resources compatible with those presented by these approaches.