A Novel Energy Adaptive Neural Network and Deep Q-Learning Network for Improved Energy Efficiency in Dynamic Underwater IoT Environment

Optimizing energy efficiency and communication reliability is essential for an underwater Internet of Things (IoT) network that utilizes hybrid optical-acoustic communication system. The proposed research work has an objective to reduce the energy consumption in a dynamic underwater IoT network by adjusting the transmission power and frequency selection using a novel Energy Adaptive Neural Network (EANN). The communication environment is continuously monitored by the proposed EANN to adjust the transmission power and frequency to adapt the dynamic changes in the network. The next objective of the research work is to enhance the reliability and robustness of the communication model by incorporating a Deep Q-Learning network for adaptive error correction and modulation (DQL-AECM). The reinforcement learning model optimizes the error correction and modulation schemes to ensure that the systems effectively adapt the variations while maintain high communication reliability in an underwater IoT network. The multiple deep learning models in the proposed work provides a complete solution to reduce the energy consumption and ensure reliable communication which is essential for an underwater IoT network. The experimental analysis of proposed model is considering the traditional LEACH, PEGASIS, TEEN and MTP models to comparatively evaluate the performances. The proposed model exhibits better performance with energy consumption of 22 Joules and an energy per bit of 2.0 × 10⁻⁶ Joules/bit, SNR of 21dB, low BER of 10-8, throughput of 6.2 Mbps and latency of 35ms which is superior than the existing LEACH, PEGASIS, TEEN and MTP models.