​Abstract: This article presents an extended study of a modular, cost-efficient Unmanned Aerial System (UAS) designed for fully autonomous disaster-response operations. Building upon prior conference work [12], this extension introduces a comprehensive Robot Operating System (ROS) and Gazebo-based simulation framework that validates the entire mission pipeline under realistic synthetic environments prior to field deployment. The system autonomously scans user-defined areas, detects disaster zones using a custom YOLOv8-based vision model, and executes payload drops without ground control intervention. An onboard sensor-fusion pipeline combines LiDAR range measurements, PX4 telemetry, and GPS data to reconstruct local terrain undulations, enabling low-altitude navigation with improved detection rates. Target coordinates are computed onboard via a heading-aware pixel-to-GPS conversion using a pinhole camera model and UAV yaw state. The ROS-Gazebo environment replicates sensor-plugin behavior, obstacle placement, and disaster-zone visual cues, enabling closed-loop validation of perception, navigation, and payload-delivery logic. All computations run on a Raspberry Pi 5, with pymavlink commanding the Pixhawk 4 flight controller. Extended system-level evaluations demonstrate a per-target detection rate of approximately 80 percent, successful payload delivery within a 5 m radius in 24 of 30 autonomous flights, 2 kg payload capacity, and 25-minute flight endurance. The ROS-Gazebo environment further demonstrates a measurable reduction in mission planning iteration time and an improvement in pipeline integration debugging speed compared to SITL-only workflows, based on developer observations across project iteration cycles.

Keywords: Autonomous UAS, Disaster response, ROS-Gazebo simulation, Terrain undulation mapping, YOLOv8 target detection, Payload delivery, LiDAR and telemetry fusion, Onboard navigation.

__________________________________________________________________________________________