Adaptive Power Allocation for Non-coherent FSK in Time-varying Underwater Acoustic Communication Channels

Underwater acoustic channels are typically double selective. The multipath effect often leads to frequency selective fading, and the movement between transmitter and receiver leads to time selective fading. In this paper, focused on the non-coherent frequency shift keying (FSK) underwater acoustic communication systems, we proposed an adaptive power allocation based on reinforcement learning (RLAPA) algorithm. The proposed algorithm can find the optimal power allocation scheme according to the frequency selective fading of the underwater acoustic channel. While the underwater acoustic channel is constantly changing, the varying model of the underwater acoustic channel and the long-term optimal power switching scheme are learned through our proposed algorithm. Simulation results based on pool channels show that the proposed algorithm can achieve long-term optimized channel capacity and lower average bit error rate than the power averaging scheme, which improves the performance of the communication system.