Genetic optimization of short block-length PAC codes for high capacity PHz communications

With the ongoing deployment of 5G cellular networks, the research and development of 6G technologies have recently triggered. The current studies begin to explore the higher frequency bands, such as Terahertz (THz) and Petahertz (PHz). In order to meet the requirements PHz communications, we focus on the novel coding schemes with short block-length and high code rate. Last year, Arikan proposed new concatenated codes called polarization-adjusted convolutional (PAC) codes. Under the Fano decoding, the performance of PAC codes with short block-length can approach Polyanskiy-Poor-Verdu (PPV) bound. In order to reduce complexity, list decoding of PAC codes is proposed. However, there is no general construction method for PAC codes with different block-lengths and code rates. In order to obtain an effective solution of PAC codes construction with higher code rates and longer lengths, we propose a genetic algorithm (GenAlg) based optimization scheme of PAC codes evaluated by list decoding. The simulation results show that a (512, 256) PAC code optimized by the proposed GenAlg scheme can achieve a noticeable performance gain of 0.25 dB at block error rate (BLER) of 10−3, when compared to the PAC code constructed by Gaussian approximation (GA). A (256,192) PAC code constructed by our scheme can lend to a coding gain of 0.3 dB at BLER of 10−2 when compared to the PAC code constructed by Gaussian approximation (GA).