Reinforcement-Learning Based Dynamic Transmission Range Adjustment in Medium Access Control for Underwater Wireless Sensor Networks

In this paper, we propose a reinforcement learning (RL) based Medium Access Control (MAC) protocol with dynamic transmission range control (TRC). This protocol provides an adaptive, multi-hop, energy-efficient solution for communication in underwater sensors networks. It features a contention-based TRC scheme with a reactive multi-hop transmission. The protocol has the ability to adjust to network conditions using RL-based learning algorithm. The combination of TRC and RL algorithms can hit a balance between the energy consumption and network performance. Moreover, the proposed adaptive mechanism for relay-selection provides better network utilization and energy-efficiency over time, comparing to existing solutions. Using a straightforward ALOHA-based channel access alongside "helper-relays" (intermediate nodes), the protocol is able to obtain a substantial amount of energy savings, achieving up to 90% of the theoretical "best possible" energy efficiency. In addition, the protocol shows a significant advantage in MAC layer performance, such as network throughput and end-to-end delay.