Environment-aware communication channel quality prediction for underwater acoustic transmissions: A machine learning method

Abstract Due to the limited energy supply of sensor nodes in underwater acoustic communication networks (UACNs), energy optimization for underwater acoustic transmissions is critical to prolong the network lifetime and improve network performance. Machine learning is a powerful and promising method that can be used to optimize energy consumption of UACNs. In this paper, we propose a machine learning based environment-aware communication channel quality prediction (ML-ECQP) method for UACNs. In ML-ECQP, the logistic regression (LR) algorithm is used to predict the communication channel quality (which is measured according to the bit error rate) between a transmitter and a receiver based on the perceived underwater acoustic channel environmental parameters (such as signal-to-noise ratio, underwater temperature, wind speed, etc.). Based on the predicted communication quality, each transmitter can optimize the acoustic data transmissions in order to minimize the energy waste caused by retransmissions, thus significantly reducing the energy consumption of UACNs. Extensive experiments are conducted in the Furong Lake at Xiamen University to demonstrate the performance (in terms of the feasibility, channel condition predication accuracy, and energy consumption reduction) of the proposed ML-ECQP method.