Autonomous Drone for the Point-to-Point Navigation

Unmanned Aerial Vehicles (UAVs) are transforming modern logistics by enabling autonomous systems. However, safe navigation in urban and rural environments is still a challenge due to obstacles such as trees, buildings, houses and transmission wires. This research presents the development of an autonomous drone from point-to-point navigation, integrating real time obstacle avoidance. The system integrates a Pixhawk 6x as the flight controller and a Jetson Nano as the onboard processing unit. A camera and a Lidar sensor take the input from the real environment, enabling obstacle detection using a deep learning model (Yolov8) and Convolutional Neural Network (CNN). The Jetson Nano processes real time video streams, detects obstacles and sends corrective navigation commands to the Pixhawk flight controller using the MAV Link protocol. Based on the detected obstacles, the drone adjusts its trajectory. It ensures a safe and efficient flight path between pre-defined waypoints. The proposed system is tested in various environments. It shows good obstacle detection and real time flight path correction. The system is tested experimentally and the result shows an improvement in (1) real-time object detection accuracy, (2) obstacle avoidance efficiency and (3) navigation stability. Along with avoiding the obstacles, path correction algorithms reroute around the obstacles with minimal deviation from its original course. This research contributes to advancing autonomous UAV navigation systems, particularly in obstacle avoidance. Our findings show the potential for UAVs to work safely and autonomously in real world conditions. This work can be expanded to integrate LiDAR for enhanced depth sensitivity. It can expand the system’s capabilities for multi-object tracking and decision making in complex environments.