Research On Ordered Statistic Decoder Of High Reliable And Low Lantancy Fountain Codes

Ultra-reliable low latency communication（URLLC）is aimed at new services in the fifth generation（5G）mobile communication network with ultra-high reliability requirements and extremely low latency,such as factory automation and automation driving,etc.URLLC requires Block Error Rate（BLER）of 10^-510^-9 and an end-to-end latency less than 1 ms.Corresponding to the physical layer is equivalent to short code length channel coding requiring high reliability and low complexity.However,the existing URLLC coding candidate scheme can approximate the short code length performance boundary under Maximum Likelihood（ML）decoding,but most of them have the characteristics of fixed rate and high coding complexity,and fixed rate coding.The decoding failure of the scheme under the wireless time-varying channel will result in retransmission and cannot meet the latency requirement of the URLLC.Therefore,how to design a low-complexity codec scheme with flexible code rate is still a huge challenge.It can be seen that the URLLC requires a "rateless" coding and decoding scheme with high flexibility,high reliability and low decoding complexity,because the rateless code can adaptive channel state transmission and avoid the latency overhead caused by retransmission.On the other hand,the system Raptor code and the non-binary system Raptor Q code contained in the rateless code are already the standard for multimedia broadcast multicast services,and the Raptor code can achieve the performance close to the Shannon limit in the erasure channel,but it has limitation in wireless channel without high performance decoder.In response to the requirements of URLLC,we combine a finite-length non-binary Raptor Code（NBRC）that can approach to the channel capacity,and an ordered statistic decoder（OSD）that can approximate the ML decoding performance under a short code length to design a codec transmission scheme that can avoid retransmission.The specific research contents are as follows:Four codeword construction methods of NBRC are studied,and the feasibility of ultilizing OSD is analyzed.Comparing the performance of several existing URLLC candidate coding schemes under OSD,it proves that the finite length NBRC of OSD can reach the reliability requirement of URLLC.Based on this,firstly,the receiver of the finite length NBRC transmission scheme based on OSD can determine the required number of received decoding symbols according to the real-time channel conditions,so as to ensure that the expected BLER can be reached in a single transmission,and the fast feedback mechanism is proposed to predict number of received symbols.The upper bound of BLER of finite length NBRC under OSD in Additive White Gaussian Noise（AWGN）channel is derived and verified by Monte Carlo simulation.The number of received decoding symbols is calculated using the expected BLER.At the same time,the decoding algorithm of modified OSD（m OSD）is designed for the structural characteristics of NBRC.The simulation proves that the BLER drops below 10^-5 when the SNR is 4d B.The ratio interval is close to the short code length performance bound,and the required decoding order l at low SNR is analyzed.Based on the above work,a self-Adaptive OSD（S-OSD）decoding algorithm is designed to further reduce the decoding complexity of OSD.Combined with the BLER upper bound derivation,the key step simplification strategies in the S-OSD algorithm are given,including the most reliable base segmentation strategy and the corresponding test error pattern generation method,the test error pattern discarding criterion and the decoding stop criterion.The simulation results show that the performance of the S-OSD algorithm is similar to the designed m OSD performance,and the complexity is reduced by nearly 50%.