A multi-objective bat algorithm with a novel competitive mechanism and its application in controller tuning

With the complexity of multi-objective optimization scenarios, both convergence and diversity of optimization algorithms are put forward higher requirements, but the harmony between the two has not been completely resolved, especially in controller tuning issues. To address this challenge, a novel competitive mechanism-based multi-objective bat algorithm is proposed in this paper. Firstly, based on a pairwise competition strategy, a competitive bat algorithm is designed as a candidate evolution strategy that promotes the population to converge quickly without cumbersome external archives. Furthermore, to avoid premature convergence, the genetic algorithm is utilized to diversity the swarm. Secondly, a designed tribal competition mechanism achieves a dynamical adjusting of evolution strategies according to maturity. Thence, the proposed algorithm can effectively balance the convergence and diversity through the adaptive complementation of multiple evolution strategies. Finally, experiments on benchmark functions illustrate that the proposed algorithm statistically outperforms the compared 12 state-of-the-art algorithms on at least 15 out of 19 instances. Besides, the proposed algorithm is used to solve multi-objective tuning problems of two widely used controllers in a laboratory-developed permanent magnet synchronous motor system. Comparison with 4 representative algorithms verifies its effectiveness and practicality in real-life multi-objective optimization problems.