Sequential Optimization of Decision Feedback Equalizer Hyperparameters in Mobile Acoustic Channels

Mobile underwater acoustic communication (UAC) is affected by fast-changing propagation conditions, posing serious challenges to maintaining reliable data transmission. In mobile settings, receiver algorithms must be adaptive to track critical variations in Doppler and delay spreads. This adaptability can be achieved by tuning key hyperparameters of the receiver. In this paper, an online tuning approach for Decision-Feedback Equalizer (DFE) hyperparameters in mobile UAC channels is proposed. A Bayesian Optimization (BO) framework employing Gaussian Process (GP) regression is utilized to track the optimal hyperparameters at a single-evaluation-per-packet rate. Simulation results, obtained using channels generated by Bellhop combined with the Maximum Entropy model, demonstrate enhanced Packet-Error Rate (PER) and Signal-to-Noise Ratio (SNR) performance compared to a channel-estimation-based receiver tuning benchmark.