Deep Reinforcement Learning-Based Adaptive Transmission Control for Communication Systems in Autonomous Underwater Vehicles
Year: 2026
Type: conference-paper
Abstract: In the past few years, communication systems for autonomous underwater vehicles have gained significant importance in marine exploration, environmental monitoring, species tracking, offshore inspection and underwater surveillance applications. However, existing AI-driven adaptive communication approach struggled with several limitations such as communication adaptation depends on fixed predefined transmission rules that restricts real-time adaptability in varying underwater channel conditions and does not optimize communication performance based on transmission feedback. To overcome these challenges, Deep Reinforcement Learning-Based Adaptive Transmission Control (DRL-ATC) framework is proposed for autonomous underwater vehicle communication systems. Primarily, underwater image data and environmental sensor readings are collected from onboard sensing units. Then, collected data preprocessed using sensor synchronization, normalization and communication-state formatting for generating structured environmental feature vectors. Further, preprocessed image frames are processed through onboard species classification module for identifying marine species. Consequently, communication state vector is constructed through species classification output, signal-to-noise ratio, bit error rate, channel gain, Doppler characteristics, residual energy and packet delivery feedback. Further, based on this representation, based on this representation, proposed Adaptive Transmission Control module employs DRL for dynamically optimizing transmission parameters which involves modulation strategy, transmission power, packet size and retransmission interval according to real-time underwater conditions. Moreover, optimized parameters are applied at the time of acoustic communication which allows continuous learning-based adaptation throughout operation. Finally, proposed DRL-ATC framework obtained more reliable and adaptive underwater communication by dynamically adjusting transmission parameters according to real-time channel conditions. Therefore, proposed DRL-ATC framework attained better results in terms of throughput (92.3kbps), end-to-end delay (109ms), bit error rate (0.052) and minimized energy consumption (1.76 J) which indicates better communication efficiency compared to existing approach.
Institution: Malla Reddy Engineering College for Women
Language: English
Subfield: Ocean Engineering
Field: Engineering
Domain: Physical Sciences
SDG: Life below water
Open Access status: hybrid
OpenAlex ID: W7203464161