Sea++
Our inaugural AUV, built by a first-year team of 10 students. Used a BlueROV2 R2 frame with off-the-shelf components, YOLO v4 object detection, ROS with PID control, dual lowlight cameras, and an NVIDIA Jetson Nano. Designed as a reliable foundation for future years.
Competition Video
At a glance
Specifications
- Onboard computer
- NVIDIA Jetson Nano
- 4 GB RAM
- Flight controller
- Pixhawk PX4
- Via MAVROS / MAVLink
- Frame
- Blue Robotics BlueROV2 R2
- Cameras
- 2x Low-Light HD USB
- One forward, one downward
- Backup sensing
- 2x Ping sonar altimeters
- Mounted beside the front camera
- Claw
- Modified Newton Subsea Gripper
- Custom 3D-printed jaws and 4 mm carbon-fiber forks
- Vision
- YOLO v4 + Sea-Thru
- Color correction for underwater glare and tint
- Power
- Turnigy 5000 mAh 4S LiPo
- Team
- 10 students
- Our first season
Design
Highlights
Off-the-shelf foundation
With little manufacturing equipment, we used a BlueROV2 R2 frame for its durability and support for off-the-shelf parts.
Two-camera vision
A forward camera localized the vehicle relative to the gate and buoys; a downward camera detected symbols and centered the vehicle over them.
Sea-Thru color correction
We applied Derya Akkaynak’s Sea-Thru algorithm to remove underwater tint, which also let us train on images taken on land and in simulation.
Behavior tree from day one
Chose a behavior tree over a finite state machine for easier scaling, runtime reconfiguration, and ROS integration.