A Hybrid 3D-2D Point Cloud Fusion Framework for Adaptive Localization in Warehouse Robot Navigation Systems

Modern logistics environments require robust robot navigation, yet traditional two-dimensional localization methods face limitations in dynamic and feature-sparse settings. This work proposes a fusion framework integrating three-dimensional and two-dimensional point cloud data, combining Iterative Closest Point for geometric alignment with Monte Carlo localization for probabilistic position estimation. The method optimizes rotation and translation matrices via singular value decomposition and dynamically adapts computational resources to environmental complexity. Simulations demonstrate accurate localization in open spaces, irregular layouts, and dynamic obstacle scenarios. Real-world experiments on lidar-equipped omnidirectional robots validate reliable navigation and obstacle avoidance in static and dynamic environments. The framework balances precision and computational efficiency, enabling deployment on edge devices.