Study On Forward Error Correction Technology Based On Reliabilities

With the development of mobile Internet technology,new services such as virtual reality,ultra-clear live video streaming,the Internet of Things,and unmanned driving continue to emerge,placing ever higher requirements on the reliability and effectiveness of information transmission technology.As a classic error control method,forward error correction technology plays a very important role in digital communication,and it is also a research hotspot in recent years.The traditional forward error correction technology mainly works on the physical layer,such as Low Density Parity Check(LDPC)codes and polar codes used in the 5G-NR physical layer standard.In addition,there is a type of forward error correction technology that acts on the network layer and above,such as fountain codes represented by Luby Transform(LT)codes and Raptor codes,which have been adopted in some channels of 4G LTE,and is very promising to get more applications in B5 G solutions.For a high-performance forward error correction technology,the decoding algorithm should follow the principle of making full use of the correlation and reliability information within the codeword,so as to maximize the recovery of the correct codeword content.This paper focuses on the forward error correction technology based on reliability decoding of the physical layer and the application layer,and researches encoding and decoding algorithms of LDPC codes,polar codes,Polarization-Adjusted Convolutional(PAC)codes,and fountain codes,and gives detailed simulation results.First of all,the paper introduces the definition and encoding / decoding algorithms of LDPC codes,polar codes and fountain codes.At the physical layer,the paper describes the LDPC code with matrix and Tanner graph definition method,introduces the encoding method of Gaussian elimination,and several traditional decoding algorithms such as Belief Propagation(BP),MinSum(MS),and so on.The thesis analyzes the matrix generation method of polar code and the corresponding channel selection method of information bits,and gives the detailed steps of decoding algorithms such as Successive Cancellation(SC)and Successive Cancellation List(SCL).At the application layer,the paper introduces two classic fountain codes of LT code and Raptor code,studies the degree distribution function of LT code and the precoding matrix of Raptor code,and explores BP decoding of LT code and deactivation decoding of Raptor code.Secondly,the thesis studies the physical layer reliability decoding method represented by Ordered Statistic Decoding(OSD)algorithm.The paper introduces the basic steps of the OSD algorithm and its application methods in LDPC codes and polar codes.After explaining the CRC codes in the 5G standard,the paper proposes an OSD algorithm with partial CRC aided,and gives detailed matrix segmentation and stitching methods.The paper explores two approximate maximum likelihood decoding algorithms similar to the OSD algorithm,namely the List Erasure Decoding(LED)algorithm for LDPC codes and the CRC-aided Sphere Decoding(CA-SD)for polar codes,and compared their decoding performance and computational complexity with the OSD algorithm.Then,the thesis studies the reliability decoding method based on pre-transform modification.The paper expounds the calculation method of the performance approximation(PolyanskiyPoor-Verdú,PPV)bound under the limited code block length,and analyzes the current LDPC codes and polar codes that have a large performance improvement space in terms of short codes.This paper introduces a new type of Polarization Modified Convolutional(PAC)code.Its performance is very close to the PPV bound under the RM construction criteria and sequential decoder.The paper perfects the construction rules of PAC codes under multiple code rates and code lengths,proposes a Successive Cancellation Double List(SCDL)decoding algorithm with stable complexity,and gives relevant simulation results.Finally,the paper studies the forward error correction technology based on reliability feedback at the application layer.This thesis introduces three fountain codes with feedback—Shifted LT(SLT)codes,LT codes with Alternating Feedback(LTAF)codes,and Delete and Conquer(DC)codes.The encoding method optimizes the traditional LT code from the perspective of degree distribution function and feedback strategy,and reduces the decoding overhead while maintaining similar encoding and decoding complexity.The paper compares the effects of parameters in the Robust Soliton Distribution(RSD)for various coding schemes through simulation,and compares differences in key indicators such as the decoding overhead,storage requirements,decoding complexity and real-time property in unicast and multicast scenarios.