Algorithm Optimizations And Efficient Implementations For 5G Polar Codes

As the latest generation of mobile communication technology,the fifth-generation mobile com-munication technology is constantly evolving on the basis of the fourth-generation mobile commu-nication technology,which has led to an unprecedented rapid improvement in national and social development and the quality of life of the people.Basic service concepts such as mobile payment,online conferences,online classrooms,remote consultations,and cloud consultation rooms are e-merging in endlessly,changing people's production and lifestyles and improving people's happiness all the time.In addition to the basic life services of the people,the information processing capa-bilities and transmission efficiency required for the concept of unmanned driving,smart factories,augmented reality,and virtual reality have brought new challenges to the new generation of mobile communication technology,which means 5G The network has to face more application requirements and more diverse wireless mobile scenarios.Channel error correction code is an important technical part of the communication system during information transmission.Its function is to add appropriate redundant information to the information at the transmitting end to combat channel noise during transmission and improve the transmission reliability of digital signals.The 5G NR standard regulates the requirements of a new generation of low-latency,high-speed,and high-reliability mobile communications.At present,as the channel coding of the new generation of mobile communications,the polar code has been adopted as the short code scheme of the 5G Enhanced Mobile Broadband Scene(e MBB)control channel.And as the first channel code that has been theoretically proven to reach the capacity of Shannon's channel,polar codes have attracted more attention from communication experts and scholars at home and abroad.This paper will mainly study the optimization of polar code decoding algorithm and the problem of high-efficiency hardware implementation of the corresponding algorithm.At present,the two mainstream decoding algorithms for polar codes-successive cancellation(SC)algorithm and belief propagation(BP)algorithm.The two algorithms have their own advantages and disadvantages in actual decoding,so they are further studied by many scholars.In order to make up for the lack of performance of the SC algorithm,the successive cancellation list(SCL)algorithm was proposed,and then the simplified successive cancellation list(SSCL)and fast-simplified successive cancellation list(Fast-SSCL)algorithms were proposed to reduce SC and SCL.The decoding time delay of the algorithm further optimizes the decoding complexity.For the parallel BP algorithm,BP-based optimization algorithms such as the Belief Propagation List(BPL)Belief Propagation Flip(BPF)have been successively proposed to further improve its performance while ensuring the advantages of the original BP algorithm with low latency and high throughput.This paper discusses the decoding complexity of SSCL algorithm and studies the path splitting problem of SSCL algorithm for special nodes.For the original and accurate SSCL algorithm,it adopts the principle of maximum likelihood decision.All the information bits contained in the rate-1 node are flipped one by one to split the path.For a rate-1 node with a scale of N_vIn other words,2^Nvpaths need to be split and then sorted and selected.As a result,as the polar code N increases and the scale of the rate-1 node increases,the complexity of the accurate SSCL decoding algorithm will increase exponentially.The current decoding strategy tends to obtain the split times of special nodes in a specific decoding scene through simulation experiments,but its strategy lacks flexibility.In order to tighten the path splitting boundary and enable the number of splits to flexibly adapt to the polar code channel conditions,this paper proposes the worst path comparison splitting strategy.The SSCL decoding algorithm based on this strategy uses the PM value of each split path Features:Compare the PM value of each new candidate path split with the current worst path to determine whether it is necessary to split.This strategy can flexibly adapt to judge the pros and cons of the polarized channel,effectively suppress the splitting of the poor path at the rate-1 node,and provide a flexible split boundary for the rate-1 node to adapt to the channel conditions,and ensure the decoding performance of the original algorithm.At the same time,the computational complexity of the algorithm is greatly reduced.In view of the current existing BPF decoding algorithm,this paper summarizes its existing problems and the direction of further research.A total of three points can be summarized:flip set establishment experiments,sorting complexity,and the formulation of verification and error detection schemes.These three aspects are related to the performance and computational complexity of the algorithm itself,and also affect the optimization of the subsequent hardware design and implementation of the corresponding indicators.Based on the summarized three aspects of BPF algorithm performance,the pre-flip BPF algorithm-A-BPF algorithm is proposed.The A-BPF algorithm considers the timing of the BP decoding and flip set,the selection of the establishment of error-prone bits,and the influence of the check error detection scheme on the performance of the BPF algorithm,and the corresponding optimization scheme for the GBPF algorithm is given.It can be seen from the simulation experiment results that for the polar code P(1024,512),the A-BPF algorithm should flip the set at FER=10^-4When the scale size is T=16 and T=64,the A-BPF algorithm has more gains of 0.11d B and 0.17d B than the GBPF algorithm,respectively,which is better than BPL,and the L=16 algorithm is 0.23d B.And it can achieve the performance of CA-SCL L=8 under a certain complexity.Regarding efficient hardware implementation,this article discusses the hardware design of the proposed A-BPF algorithm and the design and optimization of each unique module.The top-level design structure of A-BPF decoder is given,including dual-row BP architecture,check error detection unit and bit flip management unit.The first two are basically designed according to the A=BPF algorithm.The most important thing is the research on the sorting network design in the bit flip management unit in the A-BPF decoder.Because the sorting network in the bit-flipping sorting unit faces the problem of increasing the size of the polarized code code length,this article discusses the optimization of its hardware complexity.The pruning optimization of the full bitonal sorting network and the folding optimization of the sorting network are adopted.According to the size of the search set and the flip set in the A-BPF algorithm,unnecessary CSUs in the full bitonal sorting network are rationally pruned and optimized,and the hardware is reduced.the complexity.In terms of folding optimization,according to the internal structural characteristics of the sorting network,we adopt the”divide and conquer”grouping sorting idea to disassemble the large-scale sorting network,and use the small-scale sorting network of the same structure to form a large-scale network that finally meets the functional requirements of the module.Furthermore,the multiplexing architecture of the sorting network is proposed to realize the folding optimization of the original large sorting network.The hardware complexity is greatly reduced,and the critical path of the sorting network is shortened.The final results show that the A-BPF decoder proposed in this paper guarantees the high throughput characteristics of the BP decoder itself and improves the decoding performance,making it better adapted to high-rate,low-latency actual decoding scenarios.In addition to mainstream SC decoding and BP decoding algorithms,this paper studies the performance optimization and hardware implementation of LDPC-like BP decoding(LDPClike-BP)algorithms.Two LDCP-like BP flipping(LDPClike-BPF)algorithms for wrong VN positioning strategies are proposed,and the advantages and disadvantages of the two in different channel sce-narios are compared.The experimental results show that under the preset maximum number of flips T_max=20,the performance of the P(1024,512)LDPClike-BPF algorithm reaches FER=10^-4is better than the traditional BP algorithm by 0.33 d B.Afterwards,it focused on the hardware design and implementation of LDPClike-BPF algorithm.Compared with BP decoders of different architectures,LDPClike-BPF decoders have good gains in throughput and hardware efficiency.