Marine unexploded ordnance (UXO) is a significant environmental contaminant, particularly in shallow waters where it may intersect with the public. Effective remediation requires assessment at a minimum, and in many cases either removal or neutralization. In all but the shallowest waters, this requires intervention by divers trained in UXO remediation, a dangerous task requiring specialized training and extraordinary precautions.

Over the preceding half-century, the need to perform inspection and maintenance in the deep ocean, driven largely by the needs of the energy industries, has resulted in sophisticated tele-operated robotic systems designed for operation at pressures inhospitable to human life. Such remotely-operated vehicles frequently feature high-degree-of-freedom arms to perform manipulation tasks under remote control by an operator on the surface. Such technologies originating in subsea intervention could be adapted to perform remote remediation of UXO, reducing the risk to divers. However, despite the critical role such tele-operated manipulators play in maintaining critical deep ocean infrastructure, they are technologically relatively simple, with constrained sensory capabilities and limited manual dexterity. These challenges are highly relevant to teleoperated UXO remediation where precise handling is essential for safe handling of the UXO. However, advances in user-assistive robotic control motivated by the demands of work in the deep ocean can also improve the efficiency and effectiveness of tele-operated manipulation for UXO remediation.

This project incorporates three research elements. First is the construction of a testing platform which incorporates a commercial off-the-shelf subsea manipulator, appropriate sensors and a workspace for placing test objects. This test system is scaled to allow operation in the University of Washington School of Oceanography test tank as well as installation on two vessels at University of Washington Applied Physics Laboratory for testing in local waters. The second element, headed by Olis Robotics, concerns the development of software tools for visualizing and planning grasp maneuvers before execution, allowing delicate grasps to be planned in advance based on observed knowledge of the object in context rather than being planned in real-time by an operator. The third element, driven by the University of Washington School of Electrical and Computer Engineering, concerns the use of virtual reality in planning and visualizing grasping maneuvers.

The test system was constructed and deployed to the School of Oceanography test tank in the fall of 2018. This allowed completion of a full semi-autonomous grasp cycle, including collection of a 3D point cloud of the UXO in the test workspace using a subsea laser scanner, handoff to of that point cloud to the Olis control software, planning of an ideal grasp, then semi-automated execution of the grasp, allowing a controlled grasp of a UXO stand-in.

Once matured, robotic manipulation offers a compelling alternative to human divers for all forms of UXO handling, including removal as well as preparation for destruction-in-place. As shown within this project, integrating robotic perception and control allows more precise, more efficient manipulator control under human operation. It is also a necessary prerequisite for future, more autonomous control in which robotic intelligence might perform elements of UXO remediation under human oversight but without direct human control.