Autonomous BlueBoat Navigation Research

I am working on research focused on building an autonomous navigation and perception system for a BlueBoat surface vessel. The project explores how computer vision, depth estimation, object detection, and sensor fusion can help the boat better understand its environment and navigate safely on water.

My work focuses mainly on the perception side of the system. I have tested and compared different monocular depth estimation models, including Apple Depth Pro, MiDaS, ZoeDepth, and ORB-SLAM-based approaches, to evaluate how well they estimate distance from a single camera feed. A key part of this research is determining which models are accurate enough for navigation while still being lightweight enough to run on a Raspberry Pi 5 onboard the boat.

I am also contributing to the broader navigation pipeline through YOLOv10-based object detection research, sonar-based position estimation using an Extended Kalman Filter, and high-level navigation logic for autonomous movement. Together, these components help the boat detect obstacles, estimate distance, track its position, and make safer navigation decisions.

Technical Focus

• Tested monocular depth estimation models for camera-based obstacle awareness
• Compared Apple Depth Pro, MiDaS, ZoeDepth, and ORB-SLAM-style approaches
• Evaluated model performance based on accuracy, depth quality, speed, and Raspberry Pi 5 feasibility
• Supported YOLOv10 object detection research for identifying obstacles in the boat's surroundings
• Worked with sonar-based position estimation using an Extended Kalman Filter
• Helped design the perception and navigation pipeline for autonomous BlueBoat movement