GenAlg-SCL: Genetic Algorithm-Based SCL Polar Decoding Toward List Size Optimization

Thanks to its outstanding performance, the CRC-aided successive cancellation list (CA-SCL) decoding algorithm has gained widespread acceptance for polar codes. However, its complexity is directly tied to the list size L, posing challenges in meeting the escalating performance demands of the forthcoming 6G. Consequently, there is a pressing need to explore approaches that optimize L while upholding satisfactory error performance. Efforts to reduce L primarily depend either on the expertise of designers or on machine learning for online training. However, these strategies are hampered by a limited exploration of the near-optimal balance between performance and complexity, or they add considerable online complexity to the decoding process. Addressing this issue, this paper proposes a Genetic Algorithm based SCL (GenAlg-SCL) decoding coupled with an offline optimization framework. By automatically fine-tuning the list sizes of the SCL decoder through Genetic Algorithm, the proposed decoder achieves a lower complexity with an acceptable performance loss. Furthermore, a list size optimization scheme is proposed for GenAlg-SCL decoders in the rate matching scenario. For a (256,128) 5G-NR polar code, numerical results validate that the GenAlg-SCL-32 decoder achieves up to 43.9% complexity reduction with less than 0.5% performance penalization. Implementation results for software-defined radio (SDR) demonstrate that the proposed GenAlg-SCL succeeds in achieving 51.6% latency and 1.94-fold throughput compared with the adaptive CA-SCL decoder.