Adaptive integration of Lévy flight strategies in Siberian tiger optimization for high-dimensional global search

Lévy flight (LF) offers a stochastic mechanism characterized by sporadic long-range moves, which can significantly improve the exploration capability of metaheuristic algorithms. In this study, we introduce and evaluate four distinct strategies for embedding LF into the Siberian tiger optimization (STO) algorithm, a swarm intelligence method inspired by tiger hunting behavior. While STO has proven effective in global optimization, its performance often deteriorates due to premature convergence and limited diversity in complex search spaces. To counter these limitations, we propose four LF-enhanced STO variants – LFSTO-direct, LFSTO-randprob, LFSTO-trial, and LFSTO-trap – each integrating LF in a unique adaptive manner. Their performance is benchmarked using the CEC-2017 test suite, focusing on convergence accuracy and robustness. Results show that the LFSTO-randprob approach outperforms the standard STO and other variants in both solution quality and consistency. This work presents a systematic framework for leveraging Lévy-based dynamics to enhance the balance between exploration and exploitation in nature-inspired optimizers.