(2022) ZED Mini ROV Localization & Dense Mapping Enclosure
Video Credit: University of Michigan
Over the past decade, significant strides have been made in underwater localization and mapping. However, challenges remain in accurately tracking unmanned underwater vehicles (UUVs) in low-texture environments—such as sandy bottoms or seagrass beds—where traditional vision-based mapping struggles due to the lack of distinct visual features.
Recognizing this gap, the University of Michigan Robotics Department set out to develop TURTLMap, a real-time localization and dense mapping system that leverages low-cost sensors and computational tools to improve underwater navigation. Their approach integrates advanced stereo vision, inertial tracking, and Doppler velocity logging (DVL) on a Blue Robotics BlueROV2 platform, enabling precise vehicle tracking even in challenging conditions, including wave-induced motion.
Engineering a Rugged Enclosure for High-Performance Mapping
To support this effort, Sexton collaborated with the University of Michigan team to design and manufacture a custom aluminum enclosure to house critical sensing and computing components, including:
- Stereolabs ZED Mini – A lightweight stereo camera with a 6.5 cm baseline and integrated 6-DoF IMU for motion tracking.
- NVIDIA Jetson Orin Nano – Providing onboard processing for real-time localization.
- Doppler Velocity Log (DVL) – Enabling accurate velocity estimation and navigation in featureless environments.
- IMU & Barometer – Enhancing motion compensation and depth estimation.
Precision Mounting & Sensor Integration
To achieve high-resolution bathymetric mapping, the ZED Mini was mounted downward facing on the front of the BlueROV2 sled, optimizing its field of view (102° x 57° x 118° FOV) to capture bottom features. Sexton’s solution included:
✅ Custom aluminum enclosure with internal mounting trays for secure sensor integration.
✅ Acrylic flat port to preserve optical clarity and wide-angle vision.
✅ Acrylic rear door for easy visual inspection of internal components, including the GPU and navigation sensors.
✅ Five SubCon connectors for robust power and data transmission, including:
- Two dedicated ports for power and data communication.
- Three auxiliary ports for future sensor expansion.
Successful Deployment & Future Applications
With this integrated system, the University of Michigan team successfully developed a real-time localization and dense mapping program, leveraging data from the sensors inside the Sexton-built enclosure. Their work represents a significant advancement in making underwater mapping more accessible, cost-effective, and scalable.
For more details, check out their published research, lecture, GitHub repository, and project webpage below!
Paper: Overleaf Example
Lecture: ROB572 Guest Lecture - Towards Marine Autonomy - Jingyu Song