Abstract
Inspired by the intricate group dynamics of wild gorilla populations, the Artificial Gorilla Troops Optimizer (GTO) represents a novel approach in swarm intelligence. Despite its effectiveness in performing global exploration, GTO is prone to early convergence and can easily become stuck in local optima, especially when addressing optimization problems with intricate constraints and rugged search spaces. To overcome these limitations, this paper introduces the Multi-Strategy Integrated Gorilla Troops Optimizer (MSIGTO), which integrates Latin Hypercube Sampling (LHS), Lévy Flight (LF), and the Cauchy Inverse Cumulative Distribution Operator (CICDO). The diversity of the initial population is enhanced through LHS, and the exploration and convergence characteristics of the algorithm are further improved by LF and CICDO. To validate its effectiveness, MSIGTO is compared with 8 representative population-based optimization algorithms. Experimental evaluations on the 2017 IEEE Congress on Evolutionary Computation (CEC2017) and 2022 IEEE Congress on Evolutionary Computation (CEC2022) benchmark suites demonstrate that MSIGTO achieves a Friedman mean rank of 1.48 on 100 dimensional problems and 1.75 on 20 dimensional problems, respectively. These results indicate superior global exploration capability, convergence efficiency, and solution robustness compared with 8 population-based optimization algorithms. The algorithm's practicality was further verified on four constrained real-world engineering problems, including the speed reducer design problem, the gear train design problem, the multiple disk clutch brake design problem, and the selective harmonic elimination pulse-width modulation problem for three-level inverters. Overall, the results confirm that MSIGTO is an effective optimizer with broad potential for engineering optimization applications.