Locally Informed Competitive Swarm Optimizer with an External Archive for Multimodal Optimization

Multimodal optimization problems are challenging problems commonly encountered in diverse domains such as logistics, engineering design, and scientific research. Swarm optimizers are promising candidates for solving these problems. However, compared to other evolutionary computation paradigms, the performance of swarm optimizers in multimodal optimization is less than satisfactory and has much room for improvement. Competitive swarm optimizer (CSO) is a relatively new swarm optimizer whose effectiveness in real-parameter optimization has been demonstrated theoretically and experimentally. To harness the potential of CSO, this paper combines the locally informed mechanism of particle swarm optimization (PSO) with the pairwise competition mechanism of CSO, resulting in a Locally Informed CSO (LICSO). LICSO enhances the competition mechanism by refining the selection of competitors. Additionally, recognizing the absence of a memory component in CSO to record historical best positions, an external archive is incorporated to store potential optima. A corresponding archive management strategy is proposed to prevent the loss of identified optima. Experimental evaluations on a set of benchmark problems are conducted to assess the performance of LICSO. The results demonstrate that LICSO compares favorably with state-of-the-art swarm optimizers for multimodal optimization.