Reinforcement-Learning-Based Adaptive Neighbor Discovery Algorithm for Directional Transmission-Enabled Internet of Underwater Things
Jinfang Jiang, Shuaihui Wang, Guangjie Han, Hao Wang
Abstract
In the Internet of Underwater Things (IoUT), nodes are usually deployed randomly. Effective discovery of randomly deployed neighbor nodes is the basis for network topology self-configuration, data routing, transmission, etc. Especially in the IoUT with the directional transmission, how to efficiently discover neighbors is a major challenge to be solved at present. Hence, in this study, the neighbor discovery problem is investigated. The proposed algorithm consists of two parts: 1) a basic quorum system-based neighbor discovery (QSND) algorithm and 2) an adaptive reinforcement learning-based neighbor discovery (RLND) algorithm. First, a directional transceiver beam scanning sequence is designed adopting a C-torus quorum system to complete the initial neighbor discovery. Then, a reinforcement learning-based adaptive beam adjustment method is investigated to adjust the number of directional beams to be scanned based on neighbor recommendations and prior knowledge, thereby reducing the number of time slots expected to be required for neighbor discovery. Finally, simulation results demonstrate that QSND and RLND outperforms other related algorithms in terms of neighbor discovery rate, neighbor discovery delay, energy usage, etc.