A simple PID-based strategy for particle swarm optimization algorithm

Zhenglong Xiang, Daomin Ji, Heng Zhang, Hongrun Wu, Yuanxiang Li

Abstract: The particle swarm optimization (PSO) algorithm, which updates particles by considering their past momentum and current direction, has demonstrated its power in several optimization applications. However, the updating strategy followed by the standard PSO mainly aims to learn from the global optimum, which often leads to PSO suffering from premature convergence. Using the past momentum can result in the overshoot problem, which usually slows down convergence in complex optimization problems. Inspired by the massive success of the proportional-integral-derivative (PID) controller in automatic control, we first establish a connection between the PSO process and the PID controller-based control system. Thereafter, we propose a PID-based strategy for PSO (PBS-PSO) to accelerate convergence and adjust the search direction to get out of local optima. The proposed PBS-PSO utilizes the past, current, and change in global best together to update the search direction. We conduct experiments on the CEC2013 test suite benchmark. The experimental results demonstrate the effectiveness of our proposed PBS-PSO algorithm. Most importantly, we find that the proposed PID-controller-based strategy has good generalization ability because it can be combined with other PSO variants to improve convergence performance in most cases.