To render AUV piloting in intuitive language amid challenges of marine autonomy, we propose a system OceanPlan to accomplish efficient and robust AUV missions. Specifically, we establish a three-level framework of LLM-task-motion planning assisted by replanning. The high-level LLM planner composes plan heuristics to fulfill the human command based on its common sense knowledge, semantic map and XML files, i.e., planning with semantic representations. The middle-level task planner generates a logical sequence of actions for the task structured from the plan heuristic, i.e., planning with logical representations. The low-level motion planner computes AUV control inputs for executing the actions in the ocean, i.e., planning with physical representations. The holistic replanner monitors both the ocean environment and the AUV status and sends the replanning signal to the respective level of planner in real time, i.e., planning in a closed-loop way.

We establish an ocean simulation platform HoloEco that offers a rich environment of diverse underwater activities. Included within it are a myriad of objects such as coral reefs, gliders, warships, underwater mountains and canyons, etc. We make available a suite of underwater applications in which a real-world AUV might engage, such as surveying coral reefs or navigating through unknown canyons. Additionally, our simulator involves an AUV model EcoMapper that closely aligns with its real-world counterpart. EcoMapper is meticulously crafted with efficient controllers like moving forward and multimodal sensors like depth maps and laser range sensors.