Deep Reinforcement Learning Based Optical and Acoustic Dual Channel Multiple Access in Heterogeneous Underwater...

Area of Science:

• Underwater sensor networks
• Wireless communication protocols
• Machine learning applications

Background:

• Heterogeneous hybrid optical and acoustic underwater sensor networks face challenges in efficient channel bandwidth utilization.
• Diverse Media Access Control (MAC) protocols used by sensor nodes can lead to suboptimal network performance.

Purpose of the Study:

• To propose and evaluate a new MAC protocol, Optical and Acoustic Dual-Channel Deep-Reinforcement Learning Multiple Access (OA-DLMA), for efficient bandwidth utilization in hybrid underwater networks.
• To enable sensor nodes to learn optimal channel access strategies without prior information by leveraging deep reinforcement learning.

Main Methods:

• Development of the OA-DLMA protocol utilizing deep reinforcement learning (DRL) where sensor nodes act as agents.
• Implementation of a differentiated reward policy prioritizing the optical channel for enhanced data transmission.
• Analytical derivation of optimal short-term sum throughput and channel utilization.

Main Results:

• The OA-DLMA protocol demonstrates near-optimal performance in simulations.
• OA-DLMA significantly outperforms existing protocols in terms of short-term sum throughput and channel utilization.
• The differentiated reward policy effectively prioritizes optical channel access.

Conclusions:

• The proposed OA-DLMA protocol offers an efficient solution for bandwidth utilization in hybrid optical and acoustic underwater sensor networks.
• DRL provides an effective framework for adaptive and optimal channel access in complex network environments.
• The OA-DLMA protocol represents a significant advancement in underwater wireless communication network performance.