Performance Investigation on Binary Particle Swarm Optimization for Global Optimization

Binary particle swarm optimization (BinPSO) is introduced as a population-based random search algorithm for discrete binary optimization problems. A number of BinPSO variants have been introduced in the literature and showed performance improvements over the original BinPSO algorithm. However, no detailed performance comparison between these BinPSO variants has been found in the current literature. In this paper, a more thorough performance comparison study on the BinPSO variants in terms of convergence speed, solution quality and performance stability is presented. The BinPSO variants are further compared with a newly adopted cooperative BinPSO variant. The performance evaluation is conducted using De Jong’s test functions, several complex multimodal functions, and a real-world engineering problem, namely optimization of the detection performance of cooperative spectrum sensing in cognitive radio networks. Results show that most of the BinPSO variants exhibit excellent performance on solving De Jong’s test functions while the cooperative BinPSO variant performs better on the complex multimodal problems and the real-world engineering problem. Overall, the cooperative BinPSO variant shows the most promising performance, especially in terms of solution quality and performance stability.