LUNA: Lightweight UAV Navigation Based on Airborne Vision for Disaster Management

Authors:
Publisher website
Google Scholar
Abstract
Natural disasters such as earthquakes, typhoons are bringing huge casualties and property losses every year. As an emerging research hotspot, UAV technology has demonstrated its potential in many areas, including the issues within disaster management. In this paper, we focus on the problem of autonomous UAV navigation based on the airborne vision and deep learning. Facing the poor conditions in disasters, we often cannot expect drones with limited power supply and processing capacity to complete heavy operations. For the task of autonomous navigation, existed methods mainly rely on sophisticated multi-device collaboration which may be difficult to quickly be deployed in an emergency. As a result, we propose a lightweight navigation strategy based on gimbal camera and onboard processing module to help achieve autonomous path-finding of UAVs in the post-disaster scenario. Our solutions can take effect on miniature UAV while being assisted by task offloading device such as mobile phones. The simulation results show that our proposed LUNA system can enable UAVs to automatically sense the disaster environment and take corresponding measures of in-flight adjustment.

This publication has 23 references indexed in Scilit: