Autonomous guided vehicle (AGV) navigation in high-density warehouses faces significant challenges due to narrow aisles and complex U-shaped traps. In such environments, traditional sampling-based path planning algorithms often converge slowly and produce suboptimal paths. To solve these issues, a novel hierarchical hybrid planning framework named MEG-RRT* (Multi-objective Evolutionary Guided RRT*) is proposed in this study. The proposed MEG-RRT* integrates an optimization engine based on NSGA-II into the sampling process. It guides exploration direction away from local minima by jointly optimizing convergence efficiency and safety-related objectives. Furthermore, a geometry-aware execution layer is introduced to improve motion through narrow passages and to refine the path structure. This layer includes radar-guided steering, adaptive step-size control, and ancestor shortcut operations. Comparative experiments were conducted in simulated scenarios of complex narrow passages and high-density warehouses to verify the superiority of the proposed MEG-RRT*. In complex narrow passages, the proposed algorithm achieves a 100% success rate; it also reduces convergence time by 43.5% compared to standard RRT* and by 44.9% compared to Informed-RRT*. In warehouse environments, it generates smooth, kinematically favorable paths that are 39% shorter than those produced by RRT-Connect. These results demonstrate that MEG-RRT* balances exploration efficiency and solution optimality, making it well suited for automated logistics applications.
Wu et al. (Fri,) studied this question.