Grid-based dynamic robust multi-objective brain storm optimization algorithm

Rich works have been done on brain storm optimization algorithm solving static single- or multi-objective optimization problems, but less reports for dynamic multi-objective optimization problems. Based on this, a grid-based multi-objective brain storming algorithm with hybrid mutation operation is proposed to find the robust Pareto-optimal solution set over time. Grid-based clustering method partitions the objective space evenly along each objective and classifies the individuals located in the same grid into a cluster. Its computational complexity is less than k-means- and group-based clustering strategies. Traditional Gaussian-, Cauchy- and Chaotic-based mutation operators have different mutation steps and generate the new individuals with various diversity. In order to enhance the diversity and avoiding the premature convergence, a hybrid mutation strategy integrating above three mutation operators is presented. Experimental results for eight dynamic multi-objective benchmark functions show that the proposed algorithm can find robust Pareto-optimal solutions approximating the true Pareto front under more subsequent environments with the acceptable fitness threshold. The longer survival time also indicates that grid-based clustering method and hybrid mutation strategy are apt to better robustness.