Particle swarm optimization with dual-level task allocation

Particle swarm optimization (PSO) is a well-known algorithm for global optimization over continuous search spaces. However, this algorithm is limited by the intense conflict between exploration and exploitation search processes. This improper adjustment of exploration and exploitation search processes can introduce an inappropriate level of diversity into the swarm, thereby either decelerating the convergence rate of the algorithm (caused by the excessive diversity) or inducing premature convergence (as a result of insufficient diversity). To address this issue, we propose a new PSO variant, namely, the PSO with dual-level task allocation (PSO–DLTA). Two task allocation modules, that is, the dimension-level task allocation (DTA) and the individual-level task allocation (ITA) modules, are developed in PSO–DLTA to balance the exploration and exploitation search processes. Unlike existing population-based and individual-based task allocation approaches, the DTA module assigns different search strategies to different dimensional components of a particle. Meanwhile, the ITA module serves as an alternative learning phase to enhance the PSO–DLTA particle if it fails to improve in terms of fitness in the DTA module. To demonstrate the effectiveness and efficiency of PSO–DLTA, we compare it with several recently developed optimization algorithms on 25 benchmark and 2 engineering design problems. Experimental results reveal that the proposed PSO–DLTA is more competitive than its contenders in terms of searching accuracy, reliability, and efficiency with respect to most of the tested functions.