Salp Swarm Algorithm Refinement: Three Core Strategies for Optimization Performance Boost

To tackle the deficiencies such as frequent premature convergence and relatively slow convergence rate of the Salp Swarm Algorithm (SSA) incurred by its fitness-prioritized leader selection mechanism and inefficient utilization of population information, this paper proposes three core refinement strategies to boost the performance of SSA, and the refined algorithm named PDMSSA. Firstly, we propose the Pareto-guided leader selection strategy to conduct fast non-dominated sorting based on both the solutions’ fitness values and their spatial distribution to ensure leaders are both high-quality and widely distributed in the search space. The scale-free network-based multi-source information fusion mechanism is the second core strategy adopted to refine position updating of followers with the joint guidance of their network-selected neighbor and elite reference solutions. At last, a state-aware directional tabu search mechanism is introduced to suppress redundant search behaviors and facilitating targeted escape from stagnation. The results of Friedman and Wilcoxon rank-sum tests on the experiment data of the refined SSA against with the vanilla SSA, four advanced SSA variants, and five state-of-the-art metaheuristic algorithms on the public CEC2017 and CEC2022 benchmark suites, illustrate that the three core strategies are able to boost up the performance of the SSA efficiently, and the refined SSA can achieve competitive convergence accuracy, stability, and robustness among all co-evaluated algorithms.