Reproduce “Embodied at Sea: Imitation Learning for Multi-Task Underwater Manipulation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Greta Perez-Haiek, Luca Macesanu, Kevin Angers, Mohammed Imam, Timothy Gambin, Christopher Clark
The authors present a multi-task vision-language-action policy for underwater remotely operated vehicles, using images, language instructions, and vehicle-state observations to control tasks. In pool tests, DynamicVLA with roll, pitch, and yaw reached 93.3% overall when partial successes counted, versus 60.0% for images alone. Field trials demonstrate grasping and mooring, but do not isolate architecture effects or establish open-water generality.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.