GPS-Denied VIO SLAM Mapping

GPS-denied environments such as forests, indoor corridors, urban canyons, tunnels, or disaster zones, autonomous drones cannot rely on satellite positioning for navigation. To solve this problem, the proposed GPS-denied VIO SLAM mapping system combines Visual–Inertial Odometry (VIO), onboard camera data, IMU measurements, and CNN-based scene interpretation to estimate the drone’s motion and build a local map in real time. The drone uses its onboard camera to observe visual features in the environment, while the IMU provides acceleration and rotation measurements. By fusing these data sources, the system estimates the drone trajectory even when GPS is unavailable. The VIO module tracks visual features across image frames, estimates relative motion, and continuously updates the drone's pose. At the same time, the CNN-based perception layer helps identify free space, vegetation, obstacles, and navigable regions. The mapping interface shows the drone’s live forward navigation view, detected visual features, free-space score, vegetation percentage, and the estimated trajectory on a 2D local map. The blue path represents the reconstructed drone trajectory, while the surrounding point cloud indicates detected environmental structure. This allows the UAV to maintain situational awareness and continue navigation in complex terrain where GPS signals are blocked or unreliable.

https://video.wixstatic.com/video/300297_438a8d6caad0478cb35ba6c583cdc639/720p/mp4/file.mp4

During flight, the interface displays the live onboard camera view, navigation direction, feature points, free-space score, vegetation percentage, and the estimated 2D trajectory. The blue line represents the drone’s reconstructed path, while the surrounding point cloud shows detected environmental structure. This enables the UAV to maintain stable navigation, avoid dense vegetation, and continue mapping even when GPS signals are unavailable.

This approach is especially useful for search-and-rescue missions, forest inspection, disaster response, and autonomous exploration, where drones must operate safely without external positioning infrastructure. By integrating VIO, CNN perception, and SLAM-based mapping, the system provides a resilient navigation framework for autonomous UAV operation in GPS-denied scenarios.