Work

ESLS: A Vision-Based Emergency Safe Landing System for UAVs

Year: 2025

Type: conference-paper

Abstract: Uncrewed Aerial Vehicles (UAVs) are increasingly used across a wide range of missions, including healthcare-related operations such as medical supply delivery and search and rescue (SAR). However, performing safe emergency landings remains a critical challenge, especially in GPS-denied or cluttered environments such as forests or disaster zones. This paper presents an Emergency Safe Landing System (ESLS) designed to support emergency descent in any mission context. ESLS integrates RTAB-MAP SLAM with visual-inertial odometry (VIO), combining data from an onboard IMU and RGB-D camera to enable real-time 3D mapping and localization. The system uses YOLO-v5 object detection fused with a binary occupancy map. This allows robust identification of unobstructed areas in dynamic and unstructured environments. ESLS supports two operational modes: (1) Emergency Safe Landing Zone Detection (ESLZD), which selects the safest available landing zone; and (2) Search-and-Rescue Mode (ESLZD-SAR), which prioritizes landing safely near a detected survivor. Simulation results in ArduPilot Gazebo show landing zone detection success rates of up to 98% and landing success rates of up to 96%, highlighting the system’s potential for reliable deployment in both standard and SAR-specific UAV operations.

Source: Procedia Computer Science

DOI: 10.1016/j.procs.2025.10.206

Authors: Ahmad Merei, Hamid Mcheick, Alia Ghaddar, Giovanni Beltrame

Institutions: Université du Québec à Chicoutimi, Polytechnique Montréal, Université du Québec, International University of Beirut, Lebanese International University

Language: English

Topic: Robotics and Sensor-Based Localization

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

SDG: Sustainable cities and communities

Open Access status: diamond

Funder: National Research Council Canada