基于SARSA算法的水声通信自适应调制

Adaptive Modulation of Underwater Acoustic Communication Based on SARSA Algorithm

水声信道复杂多变,自适应调制系统中反馈信息存在较大的时延,实际信道状态与接收到的反馈信息无法匹配,带来反馈信道状态信息过时问题,发送端不能准确做出自适应决策进而导致传输误码高及吞吐量低等问题。针对该问题,利用强化学习中的SARSA算法学习信道的变化并进行行为策略的选择,根据信道的变化,择优选出最佳的调制方式,以改善系统的传输误码和通信吞吐量。对比固定调制方式和直接反馈情况下的系统的误码率和吞吐量,结果表明,经强化学习后的系统误码率和吞吐量均优于其他两种方式,可见,强化学习算法在时变水声信道自适应调制中改善传输误码和吞吐量的问题上是有效可行的。

The underwater acoustic channel is complex and variable.The feedback information in the adaptive modulation system has a large delay.The actual channel state cannot be matched with the received feedback information,which leads to the feedback channel state information outdated.The transmitter cannot accurately make adaptive decisions,leading to high transmission errors and low throughput.Aiming at this problem,the SARSA algorithm in reinforcement learning was used to learn the channel variation and selected the behavior strategy.According to the channel variation,the optimal modulation mode was selected to improve the transmission error and communication throughput of the system.The results show that the bit error rate and throughput of the system after reinforcement learning is better than the other two methods in comparison to the bit error rate and throughput of the system under fixed modulation and direct feedback.It can be seen that the reinforcement learning algorithm is effective and feasible in improving the performance of transmission error and throughput in adaptive modulation of time-varying underwater acoustic channels.