Deep learning-enabled polar code decoders for 5G networks and beyond
M. Meenalakshmi,Saurabh Chaturvedi,Vivek K. Dwivedi
DOI: https://doi.org/10.1016/j.aeue.2024.155220
IF: 3.169
2024-03-09
AEU - International Journal of Electronics and Communications
Abstract:The fifth generation (5G) wireless communication standard aims to integrate multiple networks, offering high throughput, large bandwidth, and ultra-low latency. Importantly, channel coding is now a crucial component of 5G wireless communication systems. Polar codes, introduced by Erdal Arikan, are linear block codes chosen for 5G control channels due to their capacity-reaching properties and efficient encoding/decoding algorithms. Polar codes are expected to be vital for achieving high reliability and low latency in 6G wireless communication systems, with current research focused on enhancing polar decoders. Two main polar decoder types exist: successive cancellation (SC) with lower complexity but limited throughput, and belief propagation (BP) with higher throughput but increased complexity. Addressing these challenges, deep learning (DL) techniques enhance both SC and BP decoders, leveraging fault tolerance, parallel and distributed processing ability, scalability, and better self-learning capability. The application of the DL technique in polar decoders lowers the cost of computation and significantly enhances the decoders' performance. This paper surveys novel DL-based SC/BP polar decoder approaches, evaluating error performance, decoding competence, learning efficiency, memory requirements, and hardware complexity. The review concludes that DL-based polar decoders hold great potential for enhancing performance in 5G networks and beyond.
telecommunications,engineering, electrical & electronic