Underwater robot navigates and grasps objects without training
AquaCap: A Training-Free Underwater Embodied Agent with Code-as-Policy
Robotics
Summary
Collecting data to train underwater robots is very expensive and difficult, which limits their abilities. The authors created AquaCap, a system that can understand instructions and its environment to plan actions and control an underwater robot without needing any pre-training. It uses detailed perception and remembers failed attempts to adjust its plans in real time. AquaCap was tested in simulations with a good success rate and showed it could grab and move objects underwater despite water disturbances.
What this means in practice
- •For underwater robot operators: Control underwater robots to perform grasping and transport of objects autonomously without needing extensive training data or tuning.
- •For marine environmental monitoring teams: Enable autonomous underwater manipulation tasks in challenging environments, improving data collection or sample handling with minimal setup.
Authors
Xiaoshi Li, Yule Xu, Chunghiu Kong, Yizhou Zhou, Yang Liu, Hao Yang, Zihao Huang, Yunxiao Shan
Abstract
Recent advances in vision-language-action models have stimulated growing interest in underwater embodied intelligence. However, their reliance on large-scale interaction data limits their applicability underwater, where data collection is costly and scarce. To address this challenge, we present AquaCap, a training-free Code-as-Policy framework for autonomous underwater navigation and manipulation. AquaCap employs a dual-layer agent that translates task instructions and environmental observations into condition-aware plans and executable control programs. Structured perception then provides the agent with semantic, geometric, and reliability-aware observations under degraded underwater conditions. A failure-aware memory diagnoses unsuccessful actions and supports closed-loop replanning and code revision. This design enables online adaptation without task-specific training or parameter updates. AquaCap achieves a 66.43% success rate in simulation. Real-world experiments further demonstrate autonomous grasping and object transport with an ROV, including the manipulation of targets displaced by hydrodynamic disturbances.