Archived Simulated Annealing for Unrelated Parallel Machine Scheduling to Minimize Fuzzy Makespan and Tardiness

The research proposes a two-phase matching-based decoding scheme that will be incorporated into bi-objective simulated annealing (SA) to solve unrelated parallel machine scheduling problems (UPMSP). The two objectives are to maximize the satisfactions of makespan and average tardiness, in terms of fuzzy measure. The effectiveness of the decoding scheme is investigated using four simulated annealing algorithms: acceptance probability rules based on fitness value difference and Pareto dominance, as well as solution search based on fixed weighted and random weighted directions. An experiment was conducted to evaluate the performance of the proposed eight SAs via instances generated by a method in the literature. The experiment results indicate that (1) two-phase decoding method that uses max-min matching first and Hungarian method second significantly improves proximity quality; (2) Pareto dominance rule probability and random weighted direction search produce solutions with better diversity.