Design And Optimization Of Underwater Acoustic Reliable Communication Based On Forward Error-correction Coding

The low frequency of the sound wave and the complexity of the change in underwater acoustic channels(UACs) make the underwater acoustic channels have the following features: limited available bandwith, long transmission delay, serious noise and Doppler frequency shift, selectivity in both time domain and frequency domain. It is extremely difficult to achieve reliable and efficient communication in such underwater acoustic channels. Transmission errors over the underwater acoustic channel are inevitable and happen at a high probability, which needs error control scheme with good performance to improve the quality of underwater acoustic communication. FEC and ARQ are two basic methods for correcting error bits during the transmitted data. ARQ needs feedback information from receiver. The propagation delay of UACs is very long due to the low acoustic speed, which makes the ARQ scheme unavailable. FEC can detect and correct error bits by adding redundant information to original data, which just need little or no feedback information. Considering the limited bandwith and the long propagation delay in rapidly varying UACs, we prefer to choose FEC scheme to do the error correcting.In this paper, we proposed two FEC-like schemes to improve the goodput in UACs. The first one is a transport layer FEC scheme based on fountain codes to achieve the optimization of goodput, which uses fountain codes as packet-level FEC due to the rateless property and linear encoding and decoding complexity of the fountain codes. The second scheme is a improved version of the first one, which is a CL-FEC(Cross-layer Forward error correction) scheme. CL-FEC uses fountain codes as packet-level FEC in the transport layer and RS(Reed-Solomon) codes as bit-level FEC in the physical layer. The bit-level FEC improves the efficiency of the packet-FEC. Since the random variability of UACs, the optimization problems of the goodput in these two FEC schemes are formulated as stochastic optimization problems and solved by using discrete stochastic approximation approach. The simulation results show the iterative optimality in terms of goodput and statistic convergence of our proposals. Moreover, the simulation of the first scheme shows that our proposal can capture the dynamic changes of UACs, and the simulation of CL-FEC shows that CL-FEC is better than single-layer FEC(only packet-level FEC or only bit-level FEC).