A spring oscillator model used for particle swarm optimizer

Specific to the difficulty of optimization on complex multimodal problems, this paper proposes a spring oscillator model used for particle swarm optimizer algorithm (SOMPSO). In SOMPSO, the particles that trapped in the local optima in some dimensions and certain individual extreme points whose corresponding dimensions' positions are the farthest from them, will constitute the vibrators and the equilibrium points of several spring oscillator models (SOM) respectively. Velocities and positions of particles will be updated dynamically referred to the physical principle of SOM. This SOM enlarges the search space of particles to increase the diversity of the swarm. The experiment results show that, SOMPSO algorithm has good performance when compared with other four variants of the particle swarm optimizer (PSO) on the optimization of the multimodal composition functions.