This project developed and demonstrated a technology for safe and efficient characterization and mitigation of underwater unexploded ordnance (UXO) in deep-sea intervention operations using a robotic manipulation framework called SHARC (SHared Autonomy for Remote Collaboration). Specifically, SHARC was adapted for underwater UXO remediation in shallow, kinetic coastal environments (i.e., surf zones) by augmenting it with acoustic and optical perception methods that are fused in real time. This enables SHARC to maintain worksite situational awareness even when conducting survey and remediation operations in turbid coastal waters. Operation in this type of dynamic environment presents a particularly challenging underwater mission scenario for UXO remediation because of typically poor visibility, and hydrodynamically driven disturbance of the workspace. 

The technical objective of this project was to demonstrate operation at a Technology Readiness Level 5 (TRL-5), wherein the system components are “integrated with reasonably realistic supporting elements so it can be tested in a simulated environment”. This research program was able to develop, test, and fully integrate the opti-acoustic methods into the SHARC framework.

The system integrates high-frequency acoustic imaging with optical sensing to generate real-time, fused three-dimensional (3D) reconstructions of underwater workspaces, enabling robotic operations even under extremely turbid conditions where optical visibility is less than 30cm. A wrist-mounted multibeam sonar and machine vision camera, combined with a high-pressure water jetting tool, were integrated onto a hydraulic manipulator arm and tested in a laboratory tank environment.

As part of this demonstration, SHARC was used to identify and excavate objects that were partially buried within gravel sediments and over washed in turbid water approximately 5ft (1.5m) deep and with visibility less than 1ft (30cm) within a test tank. Excavation was carried out using a high-pressure water jet traversing over user identified areas in SHARC’s 3D reconstruction of the workspace. Initial opti-acoustic survey of the worksite requires approximately 90 seconds of data collection to create a 3D reconstruction. This reconstruction technique is able to detect and localize objects made of wide range of physical materials (including metal, plastic, stone, wood, sediment, wire mesh, and netting) with positional accuracy to within ± 5cm. Excavation is carried out with the jetting tool operating along a user-specified path at a selectable constant standoff distance (typically between 10 and 20cm). Excavation was typically completed within 2-3 minutes, when excavating metal pipe sections that were buried in gravel to depths of approximately 10cm.

The opti-acoustic perception system demonstrated the ability to detect and localize objects composed of diverse materials with positional accuracy to within ±5 cm, and to support iterative excavation of partially buried objects within minutes. The system achieved TRL-5, with components integrated and tested in a simulated environment. These results establish the feasibility of automated opti-acoustic perception for underwater UXO remediation and provide a foundation for future integration onto robotic underwater vehicles for operational testing. It should be noted that the opti-acoustic fusion methods demonstrated here are intended for characterization and remediation of partially exposed or shallowly buried targets and may not be suitable for detection of fully buried munitions, which typically require electromagnetic induction or lower-frequency acoustic sensing methods. Having successfully completed these demonstration operations with this real-time perception method fully integrated into SHARC’s operational framework, this research program has satisfied the TRL-5 objective.

The demonstrations also incorporated several elements consistent with TRL-6 requirements, including testing of a fully integrated prototype system in a controlled laboratory environment with realistic environmental conditions, specifically: turbid water with visibility < 30cm, gravel sediment overburden, and partially buried heterogeneous objects. However, the current test environment, a 7ft diameter, 5ft tall static tank without wave or current generation, does not fully replicate the dynamic hydrodynamic conditions of an operational coastal environment. Additionally, the manipulator arm was mounted on a fixed baseplate, whereas an operational deployment would be on a remotely operated vehicle or bottom-crawling robot, which introduces additional degrees of freedom and platform stability challenges. Therefore, while the core opti-acoustic perception and jetting subsystems have been validated at a level approaching TRL-6, full TRL-6 demonstration will require testing in a higher fidelity facility that incorporates hydrodynamic forcing and vehicle-mounted operation. The system is now ready for advancement toward TRL-6 and TRL-7 objectives through integration onto a robotic underwater vehicle platform and testing in progressively more realistic operational environments. (Project Completion - 2026)

Phung, Amy and Camilli, Richard. 2026. Sonar-MASt3R: Real-Time Opti-Acoustic Fusion in Turbid, Unstructured Environments. 2026 IEEE International Conference on Robotics and Automation (ICRA). https://doi.org/10.48550/arXiv.2603.13585.