Learning and Sampling-Based Informative Path Planning for AUVs in Ocean Current Fields

Ying Yu

doi:10.1109/TSMC.2024.3370177

2025-01-01

PRE

Ying Yu

郑

郑华荣 (Huarong Zheng) *

许文 (Wen Xu)

DOI：10.1109/TSMC.2024.3370177

原文链接

摘要

En 中文

Autonomous underwater vehicles (AUVs) are widelyused in sampling on-site the seawater parameters, such astemperature, salinity and biomass for better understanding theocean. The AUV path needs to be carefully planned in order tomaximize the sampled information within the power constraints,which is known as the informative path planning (IPP). Theexistence of ocean currents further complicates the problem. Thisarticle proposes an IPP method for AUVs under the influenceof ocean currents via combining the probabilistic roadmap andQ-learning. Specifically, theQ-learning algorithm builds an infor-mative optimal AUV path by traversing a learned and updatedQ-table. TheQ-value in the table represents the expectation ofthe obtained reward if taking a certain action moving fromone position to another. Considering the characteristics of theIPP task, we design the reward matrix inQ-learning using theprior knowledge on the environment information. A convergentQ-table guarantees that only one complete training and learningis required to generate the path between any two positions.This feature facilitates converting the possible repetitive pathplannings into simple search problems, and thus the automaticreturn is easily realized whenever the AUV residual energy isinsufficient. Moreover, to improve the efficiency of theQ-learningalgorithm, a probabilistic roadmap with random sampling isgenerated and combined with theQ-learning. Various simulationsand comparisons are carried out. The results demonstrate theeffectiveness of the proposed IPP algorithm, showing that theconvergence of the path planning can be achieved quickly andsuccessfully. The superiority in terms of efficient return pathplanning over the traditional path planning method, RRT*, isalso demonstrated.

Keyword:

Path planning

Q-learning

Planning

Probabilistic logic

Ocean temperature

Task analysis

Search problems

Autonomous underwater vehicles (AUVs)

informative path planning (IPP)

ocean currents

probabilistic roadmap