MOSTA-II: An Efficient Multi-objective State Transition Algorithm via Modified RankOrdinal Non-dominated Sorting Approach

Optimization algorithm based on Pareto dominant strategy has been widely used in solving multi-objective optimization problems. However, the inappropriate search scope and dominance resistance solutions make it challenging to converge well, while the low efficiency of non-dominated sorting algorithms leads to an expensive computational cost. To alleviate those problems, MOSTA-II, an efficient multi-objective state transition algorithm via modified RankOrdinal non-dominated sorting approach, is proposed in this paper. First, a search scope adjustment scheme is designed to adaptively modify the population search range and accelerate population convergence; Second, a function value adjustment scheme is investigated to strengthen the selection pressure and timely eliminate dominance resistance solutions; Third, MRO, a modified RankOrdinal non-dominated sorting algorithm, is proposed to enhance the efficiency of candidate selection and reduce computational cost during the nondominated sorting process. Experiments on numerous standard test problems show that the proposed MOSTA-II enhances population convergence, effectively alleviates dominance resistance solutions, and improved the efficiency of candidate selection, achieving state-of-the-art results in terms of comparison counts and computational time costs.