μMOSM: A hybrid multi-objective micro evolutionary algorithm

In multi-objective optimization problems (MOPs), several mutually conflicting objectives are optimized simultaneously. In such scenarios, there is not a unique solution to the problem; instead, there is a set of solutions known as the Pareto front, representing the trade-off between objectives. Multi-objective evolutionary algorithms (MOEAs) can approximate these solutions in a single run. However, due to their resource-intensive nature, MOEAs are not suitable for solving real-time and engineering MOPs such as the optimization of manufacturing processes and energy consumption in wireless networks, where a fast convergence rate with less computational cost is required. Fortunately, micro versions of MOEAs can meet this requirement by utilizing a tiny population size. However, this can result in a rapid loss of diversity and the algorithm may easily fall into a local optimum. While some approaches such as the restart technique have been proposed to address this issue, hybrid techniques such as integrative, collaborative, and decomposition-based methods have not been effectively considered in the design of micro algorithms, despite hybridization being a widely accepted method for enhancing the diversity of evolutionary algorithms. In this study, we propose a hybrid micro MOEA called mu MOSM that can effectively tackle the diversity loss problem and accelerate the convergence rate in approximating Pareto front solutions. Experimental results on benchmark test suites and a real-world MOP demonstrate the advantages of our proposed algorithm and confirm that mu MOSM outperforms state-of-the-art MOEAs and micro MOEAs such as MOSM, ADE-MOIA, MMOPSO, NSGA-III, MOEA/D-FRRMAB, mu FAME, and ASMiGA.