AutoDrone 2025 - Autonomous Surface Drone Competition
Represented NTNU at the 2025 AutoDrone competition in Horten. Migrated the codebase from ROS to ROS2, upgraded hardware from Jetson Orin Nano 4GB to 8GB, and implemented autonomous navigation for competition missions.
Duration
Spring 2025
Role
Team Member
Institution
NTNU
Status
Completed
Technologies Used
Overview
This team project continued development of NTNU's autonomous sea drone for the 2025 AutoDrone competition in Horten, Norway. Building on a previous bachelor thesis, we migrated the entire codebase from ROS to ROS2, upgraded the onboard computer from Jetson Orin Nano 4GB to 8GB, and implemented mission logic for four competition tasks: Speed Gate, Obstacle Channel, Collision Avoidance, and Visual Docking. At the competition, three of four missions (obstacle avoidance, speed gate, and channel) were completed successfully, with only the docking test experiencing issues.
Problem Statement
Address issues from the previous year's competition entry (drone flipped during launch) by improving code structure, upgrading hardware, and implementing reliable autonomous navigation for buoy-based missions.
Challenges & Solutions
| Challenge | Solution | Outcome |
|---|---|---|
| ROS to ROS2 Migration | Rewrote all launch files, scripts, and communication nodes for ROS2 Humble | Modern, maintainable codebase with improved performance |
| YOLO Object Detection | Trained and deployed YOLOv8 for real-time buoy detection (red, green, yellow) | Reliable buoy detection for autonomous navigation |
| Hardware Upgrade | Upgraded from Jetson Orin Nano 4GB to 8GB for better model inference | Faster processing and more stable operation |
| Mission Logic Implementation | Developed Python scripts for each mission: speed-gate, obstacle-channel, collision-avoidance, docking | 3 of 4 missions completed successfully at competition |