01/2025/Robotics · Computer Vision
Submersible
Underwater Vehicle Control System
01 / Overview
The project
Co-architected the control software for a competition-class ROV. Built photogrammetry and 3D mapping algorithms that reconstruct the ocean floor from stereo imagery and feed real-time pose data back into the autonomy stack.
- Role
- Software Developer
- Year
- 2025
- Discipline
- Robotics · Computer Vision
02 / The why
Why.
Give a competition ROV the ability to model its surroundings underwater — specifically a makeshift coral reef — so operators (and eventually the vehicle itself) can navigate, inspect, and document the environment without relying on direct line of sight.
03 / Toolkit
Skills used
- 01SWIFT
- 02Python
- 03Photogrammetry
- 043D MAPPING
04 / In depth
How it came together
- 01
Designed the software architecture alongside the team lead, splitting responsibilities between low-level vehicle control and the perception/mapping stack.
- 02
Calibrated stereo cameras and tuned photogrammetry pipelines to reconstruct 3D point clouds of a synthetic reef under poor visibility.
- 03
Wrote C++ modules for real-time pose estimation and Python tooling for offline mesh reconstruction and visualization.
- 04
Iterated against pool tests, refining filtering and outlier rejection until reconstructions were stable enough to navigate from.
Outcome
MATE ROV Worlds
Stack
Swift/Python
Team
School Robotics
05 / Result
Outcome
The team qualified for and competed at the MATE ROV World Championship, with the mapping pipeline used during mission runs.
Next project
Cultivate
Solar-Powered Autonomous Irrigation