Ma, Song; Liu, Cunjia; Harvey, Christopher M; Bucknall, Richard; Liu, Yuanchang; (2025) Adaptive informative path planning for active reconstruction of spatio-temporal water pollution dispersion using Unmanned Surface Vehicles. Applied Ocean Research , 156 , Article 104458. 10.1016/j.apor.2025.104458.

Preview

Text Adaptive informative path planning for active reconstruction of spatio-temporal water pollution dispersion using Unmanned Surface Vehicles.pdf - Published Version Download (3MB) | Preview

Abstract

Real-time pollution monitoring is critical for marine environmental management, where accurate tracking and reconstruction of pollutant dispersion are essential to mitigate ecological impacts. This paper introduces an adaptive informative path planning (IPP) framework designed to address the challenges of reconstructing spatio-temporally varying dynamic environments, focusing on pollutant dispersion in marine environments. The framework combines a Finite Element Analysis (FEA)-based pollutant dispersion model with a recursive Bayesian estimator to capture spatio-temporal dynamics in complex marine environments accurately. Two information-based utility functions are developed to quantify and reduce system state uncertainty, enabling more effective and targeted data collection by Unmanned Surface Vehicles (USVs). Monte Carlo Tree Search (MCTS)-based path planning strategies are employed in the proposed framework with thorough comparative studies against the myopic path planning methods. The proposed framework is rigorously evaluated across three scenarios, investigating its performance under different starting positions and pollutant source terms. The testing scenarios, which reflects real-world conditions such as oil spills and chemical leaks, help assess the robustness and adaptability of the framework. A sensitivity analysis is conducted to guide the fine tuning of the path planning strategy. Comparative studies highlight the superior performance of the proposed framework, with the combination of expected information gain (EIG)-based utility and MCTS consistently achieving the lowest reconstruction error and demonstrating enhanced robustness in dynamic and uncertain environments. These results establish the framework as a significant advancement in autonomous environmental monitoring, offering a new solution for dynamic pollution tracking and management.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item