APSO-SL: An Adaptive Particle Swarm Optimization with State-Based Learning Strategy

Abstract

Particle swarm optimization (PSO) has been extensively used to solve practical engineering problems, due to its efficient performance. Although PSO is simple and efficient, it still has the problem of premature convergence. In order to address this shortcoming, an adaptive particle swarm optimization with state-based learning strategy (APSO-SL) is put forward. In APSO-SL, the population distribution evaluation mechanism (PDEM) is used to evaluate the state of the whole population. In contrast to using iterations to just the population state, using the population spatial distribution is more intuitive and accurate. In PDEM, the population center position and best position for calculation are used for calculation, greatly reducing the algorithm’s computational complexity. In addition, an adaptive learning strategy (ALS) has been proposed to avoid the whole population’s premature convergence. In ALS, different learning strategies are adopted according to the population state to ensure the population diversity. The performance of APSO-SL is evaluated on the CEC2013 and CEC2017 test suites, and one engineering problem. Experimental results show that APSO-SL has the best performance compared with other competitive PSO variants.