Online Navigation Planning for Long-term Autonomous Operation of Underwater Gliders

2025-11-05 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Robotics & Autonomous Systems, Artificial Intelligence & Machine Learning, Environmental Science & Earth Systems · Depth: Expert, extended

Summary

A new system, the Glider Long-term Autonomy Engine (GLAE), has been developed for online navigation planning of underwater gliders, addressing the challenges of long-term autonomous operation and fleet management. The system formulates glider navigation as a stochastic shortest-path Markov Decision Process (MDP) and employs a sample-based online planner using Monte Carlo Tree Search (MCTS). This MCTS planner is supported by a computationally efficient, physics-informed simulator that accounts for uncertain control execution and ocean current forecasts, with parameters calibrated using historical glider data. GLAE enables closed-loop replanning at each surfacing of a Slocum glider. The system was validated in two North Sea field deployments, MOGli and eSWEETS3, totaling approximately 3 months and 1000 km of autonomous operation. Results showed improved efficiency, with up to 9.88% reduction in dive duration and 16.51% in transect length in simulation, and a statistically significant 9.55% transect length reduction in one field deployment, compared to straight-to-goal navigation.

Key takeaway

For AI Scientists and Robotics Engineers developing autonomous marine systems, this work demonstrates a robust approach to long-term navigation under uncertainty. You should consider adopting a stochastic MDP formulation combined with Monte Carlo Tree Search and a data-calibrated physics simulator to enhance efficiency and reliability. This framework can significantly reduce mission duration and improve path accuracy compared to simpler navigation strategies, even with intermittent communication.

Key insights

GLAE uses MCTS and a physics-informed simulator for robust, long-term autonomous underwater glider navigation.

Principles

• Stochastic MDPs model uncertain marine environments.
• Online replanning adapts to real-time conditions.
• Data-driven calibration improves simulator fidelity.

Method

The GLAE system formulates glider navigation as a stochastic shortest-path MDP, using a custom physics-based simulator for transitions and an MCTS planner with root parallelism and double progressive widening for online action selection.

In practice

• Integrate MCTS with physics simulators for marine autonomy.
• Calibrate simulation parameters with historical field data.
• Implement backup waypoints for communication failures.

Topics

• Underwater Gliders
• Online Navigation Planning
• Markov Decision Process
• Monte Carlo Tree Search
• Physics-based Simulation

Code references

• smerckel/glidersim
• nocacuk/mars-oss

Best for: AI Scientist, Robotics Engineer, Research Scientist

Related on AIssential

• A Differentiable Composite Approximation Framework for Autonomous Underwater Vehicle Maneuvering Modeling from Sea-Trial Data — The framework combines polynomial and data-adaptive bases for AUV maneuvering, jointly calibrated with gradient-based methods and current compensation for improved prediction.
• An Infrastructure-less, Control-Independent Solution to Relative Localisation of a Team of Mobile Robots using Ranging Measurements — The MHDCL algorithm enables robust, infrastructure-less multi-robot localization by maintaining multiple pose hypotheses without motion control.
• NavWM: A Unified Navigation World Model for Foresight-Driven Planning — NavWM unifies perception, generation, and control for foresight-driven, robust visual navigation in complex environments.
• Towards End to End Motion Planning and Execution for Autonomous Underwater Vehicles Using Reinforcement Learning — An end-to-end hierarchical DRL system enables AUV navigation from raw sensor data to thruster commands.
• Featured video: Coding for underwater robotics — Underwater navigation algorithms can enable collaborative human-robot operations despite GPS limitations.
• Behavior Prediction under Occlusion — Accurate, uncertainty-aware behavior prediction is crucial for safe, efficient autonomous navigation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.