Elite-ordinary synergistic particle swarm optimization

In particle swarm optimization (PSO), the canonical learning exemplars and information sharing mechanism are often criticized due to the loss of population diversity. Aiming at preserving population diversity and promoting search ability of PSO, this paper introduces an elite-ordinary synergistic particle swarm optimization (EOPSO). In EOPSO, particles are divided into elite and ordinary members based on their fitness performance. Each elite individual learns from itself to maintain population diversity and achieve high-level global exploration. The ordinary ones fly toward a unified target and carry out some assistant local exploitation. In addition, an information interaction based jump-out strategy is designed to overcome particles stagnation situations. The benchmark functions in CEC2017 are employed to compare the performance between the proposed EOPSO with 18 optimization methods (8 state-of-the-art PSO variants and 10 recently proposed non-PSO methods). Experimental comparisons demonstrate that, in EOPSO, the particles have the abilities to reasonably adjust population diversity, effectively avoid local optima, and accurately converge to the global optimum.