Docking and Persistent Operations for a Resident Underwater Vehicle

Technical Summary: Docking and Persistent Operations for a Resident Underwater Vehicle

Problem Statement
Current ocean monitoring is constrained by the high logistical costs and sporadic nature of underwater observations. Existing methods rely either on broad-area, infrequent surveys using Autonomous Underwater Vehicles (AUVs) or long-term, fixed-point measurements via seafloor observatories. These approaches lack the spatial flexibility and continuous presence required for effective infrastructure inspection and environmental monitoring, particularly in the context of increasing threats to subsea assets. While resident underwater systems offer a solution, their deployment is hindered by challenges in autonomy, robotic resilience, and mechanical robustness in harsh, remote environments.

Methodology
This work presents the development, deployment, and operation of a resident infrastructure system at 90 meters depth, utilizing a mini-class Remotely Operated Vehicle (ROV) and a custom docking station.

• Hardware Architecture:

  • The ROV: A Blueye X3 mini-class ROV (approx. 10 kg) was modified with an external processor bottle (Jetson Nano) to handle autonomy and perception tasks. It is equipped with a forward-looking camera, IMU, depth sensor, Doppler Velocity Log (DVL), and an inductive charging puck.
  • The Docking Station: A 2m x 2m x 1m stainless steel structure deployed at 90m depth. It houses a 250 W DC power interface, a battery bank, and a 2D funnel-shaped guide with a circular array of permanent magnets for latching. The station supports multiple communication channels: acoustic (USBL), optical (LumaX modem), and inductive (power and data).
  • Deployment: The system was deployed at the NTNU Oceanlab using the R/V Gunnerus. The station was mounted on a decommissioned pig loop platform, and the ROV was initially tethered to the station to allow for human intervention and rapid parameter adjustments during trials.

• Software and Control:

  • Simulation: Development utilized a Gazebo-based simulator with a unified control architecture. Hydrodynamic models for the vehicle were derived from previous work.
  • Navigation and Fusion: An Extended Kalman Filter (EKF) fuses data from the USBL, DVL, IMU, and visual pose estimates.
  • Visual Guidance: ArUco markers were strategically placed on the docking station (ranging from 7 cm to 25 cm). A photogrammetric model was created to map tag positions accurately. The system uses a ROS 2 ArUco pose estimation pipeline to detect tags and estimate the robot's pose relative to the station.
  • Procedures:
    1. Acoustic Homing: The ROV uses a USBL system to locate the station from the surface. Upon receiving valid position fixes, it descends while maintaining a fixed heading.
    2. Docking: Upon visual contact, the ROV transitions to camera-based navigation, following U-shaped waypoint paths (Left/Right loops) to align with the docking funnel.
    3. Inspection: Post-docking, the ROV can undock and execute autonomous inspection routines around the station and the underlying platform.

Key Contributions

1. First Autonomous Deep-Water Results: This study presents the first autonomous operational results of a resident infrastructure system at both shallow water and 90-meter depth, moving beyond the manual operations of previous shallow-water trials.
2. Multi-Sensor Fusion for Docking: The implementation of a robust navigation system fusing USBL acoustic positioning with ArUco-based visual localization via an EKF, enabling autonomous homing and docking in real-world conditions.
3. Validated Inspection Routines: Demonstration of a complete autonomous workflow where the ROV docks, recharges, undocks, performs a full visual inspection of the station, and redocks.
4. Hardware Integration: The successful integration of a mini-class ROV with an external compute unit, inductive charging, and optical modem within a tethered resident configuration.

Results

• Docking Performance: The system achieved a 90% autonomous docking success rate in deep-water trials (9 out of 10 front approaches). Success rates for side approaches improved from 6/10 and 7/10 to 9/10 and 10/10 respectively after adjusting waypoint stand-off distances to account for visibility constraints.
• Mission Duration: Full inspection missions, including undocking, surveying, and redocking, were completed within four minutes.
• Acoustic Homing: Acoustic trials successfully guided the vehicle to the target location with a USBL update rate of 0.60 Hz and a measurement standard deviation of 0.089 m.
• Visual Detection: Simulation and field tests confirmed that a 5x5 ArUco dictionary provided the best balance of detection range and robustness. Field data showed that up to 14 markers could be detected simultaneously, though occlusion by marine life (fish) and equipment occasionally reduced detection counts.

Significance and Limitations
The paper claims that these results validate the feasibility of reliable, untethered operations at depth, highlighting the potential of resident ROV systems for scalable, cost-effective underwater monitoring. The work demonstrates that integrating acoustic and visual navigation allows for persistent presence without continuous surface support.

However, the authors remain modest regarding the current state of the system:

• Tethered Configuration: The current deployment relies on a tether for power and data during the resident phase to ensure operational safety and allow for human intervention. The system is not yet fully untethered for long-term deployment.
• Environmental Constraints: Performance was limited by reduced visibility at 90m depth due to a lack of external lighting and magnetic drift affecting compass-based heading estimates, which necessitated excluding compass data from the EKF.
• Logistical Footprint: The current docking station is large and requires heavy-lift vessels for deployment, limiting its flexibility compared to the small ROV itself.
• Future Work: The authors identify the development of robust mission planning, fault management, and a fully battery-powered, untethered resident vehicle as the next major milestones.