Abstract

Reliable autonomous navigation of Unmanned Aerial Vehicles (UAVs) in GPS-denied environments remains a significant challenge due to signal attenuation, spoofing, and jamming. Indoor environments such as warehouses and industrial facilities suffer from GPS degradation caused by structural occlusions, while outdoor scenarios including forests and flat terrains lack reliable localization cues. To address these challenges, this paper presents a multi-sensor fusion framework for robust UAV navigation in GPS-compromised environments. The proposed approach employs a dual-strategy navigation framework: (i) Visual Simultaneous Localization and Mapping (VSLAM) with 3D point cloud generation for structured indoor environments, and (ii) a VSLAM-based 2D digital twin representation for outdoor environments, integrating visual data with inertial and altitude measurements. Software-in-the-loop (SIL) simulations using ROS and RTAB-Map validate autonomous navigation with obstacle avoidance in warehouse scenarios, while Hardware-in-the-loop (HIL) experiments on an in-house developed UAV platform demonstrate accurate indoor mapping and localization. The results demonstrate that the proposed vision-centric framework enables reliable navigation without dependence on GPS, offering a scalable and resilient solution for UAV operations in GPS-denied environments, with applications in inventory management, surveillance, and search-and-rescue missions.