A Machine-Learning Enabled Estimation Approach for Real-Time Plume and Source Tracking with a Network of Autonomous Underwater Vehicles

Abstract

The research project at Worcester Polytechnic Institute (WPI) develops a highly-dynamical system estimation approach enabled by machine learning and addresses the NPT-02 theme. The highlydynamical system considered consists of a moving underwater unknown source that releases a liquid or gas trace, resulting in a spatiotemporally varying plume where a network of autonomous underwater vehicles (AUVs) performs plume estimation and source tracking. The objective of the project is to develop an approach that guides and optimally repositions the AUVs, so that the onboard estimator provides in real-time a prediction of the plume concentration, the source strength and source localization. The approach also performs under conditions of limited operability of AUVs. The project advances the state-of-the art. The estimation approach is physics-inspired and incorporates the plume dispersion modeled by the 3D advection diffusion equation, the motion of the unknown source modeled as an exosystem, the motion of the guided AUVs modeled by dynamical equations, and the concentration sensor modeled with bandwidth and noise. The estimation approach is data-driven because through adaptive sampling in the plume, the burden of processing "big-data" is replaced by significantly reduced "smart data" taken by the limited number of AUVs. The estimation approach is also machine-learning enabled and provides via a physics-informed scheme the unknown ocean currents, and source strength and location. The estimator is implemented with advanced computational methods bridging the multiple scales of the highly-dynamical system and is real-time executable onboard AUVs. The estimation approach provides in real-time human interpreted results, plume concentration and source tracking, that can lead to effective decision making. The project outcomes have impacts to applications of interest to the NAVY such as detection of underwater intruders, search-and rescue operations, and environmental monitoring. The project engages graduate and undergraduate students contributing to their education for potential future careers in the DoD. *This abstract publicly releasable

Document Details

Document Type
DoD Grant Award
Publication Date
Mar 18, 2025
Source ID
N001742210004

Entities

People

  • Nikolaos Gatsonis

Organizations

  • United States Navy
  • Worcester Polytechnic Institute

Tags

Readers

  • Computational Modeling and Simulation
  • Neural Network Machine Learning.
  • Robotics and Automation.

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • AI & ML - Bayesian Inference