Polar code decoding using a learning-based rescue algorithm with a successive cancellation list decoder

Abstract

Low-complexity node-based successive cancellation list (SCL) decoding has gained significant attention due to its potential use in 5G communication systems owing to the low latency and high-reliability requirements of 5G. Although SCL decoding has a high error-correction capability compared to other standard successive cancellation (SC) decoding methods, SCL decoding faces complexity because of its list-based method which limits the practical implications. This research proposes a SCL decoder using the learning-based rescue (LBR) algorithm to address this limitation and enhance the error-correction capability of polar codes in terms of bit error and frame error rates. The proposed method identifies weak additive white Gaussian noise (AWGN) channels that degrade decoding performance. By employing the LBR approach, the bit error rate of the weak channel can be quickly identified and fixed before subsequent decoding attempts. With LBR, it is possible to achieve considerable improvement in error-correction capability and reduced computing complexity for AWGN channels. Compared with state-of-the-art node-based polar decoding techniques, a SCL decoder employing the LBR algorithm significantly reduces the bit and frame error rates. The improvements are realized by combining a hybrid-based decoding scheme and implementing the LBR algorithm tailored for 5G New Radio (NR) polar codes. The attained bit error rate, binary phase shift keying bit error rate, and frame error rate, respectively are $2.5\times 10^{-5}$, $2.6\times 10^{-2}$, and $4.6\times 10^{-4}$ for 4 dB signal-to-noise ratio (SNR) using the SCL decoder with the LBR algorithm.