A Cloud-Based Visual Map Reconstruction for UAV Navigation Using Wireless Streaming

Obstacle maps are essential for autonomous UAVs to achieve navigation. Different kinds of sensors are used for 3D map reconstruction, such as: LIDAR, radar and cameras. However, 3D obstacle map reconstruction from UAVs has been still a challenging task due to the limitation of on-board computational resources. With the development of wireless streaming technologies, a cloud-based solution for visual map reconstruction would solve the issue. A collaborative aerial-ground system for visual map reconstruction using wireless streaming includes several drones and a fixed or mobile ground station. Images are taken from equipped cameras on the UAV while being streamed to a ground station through a high-speed wireless link, such as: WiFi or the mobile network (4G, 5G). The map is then quickly reconstructed from the images by means of a very strong computer, and the output obstacles map is then sent back to the UAV for navigation tasks. In this paper, we propose and implement a cloud-based visual map reconstruction for UAV navigation with a WiFi link. The system worked well for both indoor and outdoor experiments with achieved transfer times per image of less than 0.29 s and a total 3D reconstruction time of less than 205 s for 87 images with a resolution of 1920x1080 px.