Principal component analysis enhanced pigeon-inspired optimization for function optimization and multi-scenario UAV path planning

Abstract Standard Pigeon-Inspired Optimization (PIO) suffers from premature convergence, low accuracy, and blind search direction on ill-conditioned and multimodal problems. This paper proposes a Principal Component Analysis enhanced PIO algorithm (PCA-PIO) that introduces PCA into the map-compass operator to extract principal directions from the advantageous population for anisotropic directional search, combined with an adaptive stagnation detection and restart mechanism. Experiments on 12 benchmark functions, the modern CEC2017 and CEC2022 test suites, and $$D=100$$ high-dimensional tests show that PCA-PIO significantly outperforms PIO and CPIO in convergence accuracy and robustness across all tested functions, and outperforms PSO on 10 out of 12 classical benchmark functions (with comparable performance on the remaining 2) (PSO attains lower average ranks on the CEC hybrid/composition suites). Although it still lags behind GWO on ill-conditioned and multimodal functions, it achieves up to three orders of magnitude improvement over PIO and CPIO on the Rosenbrock function, consistent with the No Free Lunch theorem. The improvements are statistically confirmed by Bonferroni-corrected Wilcoxon tests. UAV 3D path planning in mountainous, urban, and maritime scenarios under explicit kinematic feasibility constraints further demonstrates that PCA-PIO achieves optimal or near-optimal path length, minimal turning angles, and the highest success rates, showing excellent engineering practicality.