Efficient Data Collection Scheme for Multi-Modal Underwater Sensor Networks Based on Deep Reinforcement Learning
Shanshan Song, Jun Liu, Jiani Guo, Bin Lin, Qiang Ye, Jun-Hong Cui
Abstract
Autonomous Underwater Vehicles (AUVs) with multi-modal transmission can achieve high efficient data collection for underwater sensor networks. However, multi-modal transmission and trajectory planning impose great challenges on data collection in complex underwater environments. Most prior studies focus on design of multi-modal architecture, but lack of available implementation and consideration of AUVs' trajectory. Meanwhile, existing trajectory planning research cannot work well on data collection with multiple complex tasks simultaneously. In this paper, an efficient Data Collection scheme for Multi-modal underwater sensor networks based on Deep reinforcement learning (DCMD) is proposed to solve the above challenges. We first propose an optimal multi-modal transmission selection algorithm that provides an implementation to improve transmission efficiency. Then we propose a distributed multi-AUVs' trajectory planning algorithm based on deep reinforcement learning by AUVs' collaborations, considering transmission situation, ocean currents and underwater obstacles, to maximize collection rate and energy efficiency. In addition, we joint transmission and trajectory planning in a protocol to improve collection efficiency. Extensive experimental results show that DCMD achieves better performance on efficiency and reliability than four state-of-the-art methods, demonstrating its great advantage on collecting data for USNs.