An AUV-Assisted Data Gathering Scheme Based on Deep Reinforcement Learning for IoUT

Abstract

The Underwater Internet of Things (IoUT) shows significant future potential in enabling a smart ocean. Underwater sensor network (UWSN) is a major form of IoUT, but it faces the problem of reliable data collection. To address these issues, this paper considers the use of the autonomous underwater vehicles (AUV) as mobile collectors to build reliable collection systems, while the value of information (VoI) is used as the primary measure of information quality. This paper first builds a realistic model to characterize the behavior of sensor nodes and the AUV together with challenging environments. Then, improved deep reinforcement learning (DRL) is used to dynamically plan the AUV’s navigation route by jointly considering the location of nodes, the data value of nodes, and the status of the AUV to maximize the data collection efficiency of the AUV. The results of the simulation show the dynamic data collection scheme is superior to the traditional path planning scheme, which only considers the node location, and greatly improves the efficiency of AUV data collection.

Keywords

• underwater internet of things
• deep reinforcement learning
• AUV
• data collection
• value of information