| What distinguishes 5G communication system from its previous generations is the “interconnection of everything”.Three major application scenarios in 5G,i.e.,enhanced Mobile Broadband(eMBB),Ultra-reliable low latency communication(URLLC)and massive Machine Type Communications(mMTC)focus on communication among people,communication between people and things,and communication among things,respectively.Channel coding,as a key technology in digital communication,its performance not only directly affects the efficiency of data transmission,but also affects the network coverage and the throughput of the whole communication system.In particular,polar code,as the only channel code that proves to achieve the channel capacity,takes less than ten years from its first proposal to becoming the channel coding scheme for eMBB control channel in 5G standard,which reflects its huge practical value.However,as a new coding technology,the application of polar code in different scenarios still needs to be studied.This dissertation studies the encoding and decoding schemes of polar codes under the scenarios of eMBB,URLLC and mMTC in 5G communications,and the contents of this dissertation are summarized as follows:Firstly,as the coding scheme for control channel in eMBB scenario,polar codes need to be of medium or even shorter block-length to meet the requirement of flexible coding granularity.However,the performance of polar codes with such block-lengths still needs to be improved under SC decoding.For this reason,this dissertation proposes a progressive bit-flipping SC decoder based on a critical set which flips at most independent erroneous hard decisions in SC decoding.This dissertation first studies the distribution of the first error in SC decoding and then propose a critical set which can include the first error in SC decoding with high probability.By iteratively improving such critical set,this dissertation constructs a search tree with a depth of,whose maximum value corresponds to the number of independent errors that can be corrected in SC decoding.In particular,when = {1,2,3},the proposed scheme can achieve the performance of the corresponding genie-aided bit-flipping decoder.Moreover,this dissertation also introduces a low-complexity implementation strategy for the proposed progressive bit-flipping decoder,which significantly reduces the complexity while not losing much BLER performance.Simulation results show that the proposed scheme can achieve the same decoding performance as the state-of-the-art CA-SCL decoders,and its average complexity and decoding latency are almost the same as those of SC decoding at medium and high SNR regime.Secondly,URLLC has a stringent requirement for error correction performance as well as decoding latency.However,traditional SC-based decoders must wait until all the previous bits have been decoded before it continues decoding the current bit,which makes the decoding latency inevitably increase with the block-length.In order to obtain a lower decoding latency while ensuring the decoding performance,a low-latency adaptive ordered statistic decoding(OSD)scheme is proposed by exploiting the highly parallelized structure of OSD.This dissertation first designs a search metric for each candidate codeword and then prove that the codeword that minimizes the search metric happens to minimize the decoding error probability.In this sense,if a codeword has a large search metric value,then it is largely unpromising and should be discarded without testing.In addition,this dissertation also proposes a concatenated adaptive OSD strategy,by decomposing the current candidate codeword into several independent subcodes of the same block-length,and decoding these subcodes with the concatenated adaptive OSDs,a good trade-off between decoding complexity and decoding latency can be achieved.Simulation results show that,for short block-lengths and high coding rates,the proposed scheme can achieve lower decoding delay and better decoding performance than the state-of-the-art CA-SCL decoders.In this regard,the propose scheme is more suitable for decoding packets of short block-lengths and high coding rates in URLLC scenario.Thirdly,coded modulation technique is of great importance to improve the spectral efficiency in eMBB scenario.However,the bit channels created by the higher-order modulation are neither identical nor independent,and this causes problems in applying polar coding,as polar code is designed for independent and identical channels.Different from the traditional polar-coded multi-level coding strategy and bit-interleaved coded modulation,this dissertation proposes a novel coded modulation scheme,i.e.,convolutional polar coded modulation(CPCM),by reconsidering the construction of polar codes.This dissertation first studies polar coding under independent and non-identical bit channels,and point out that a lower BLER can be obtained by combining two different channels at each polarization kernel.On this basis,a convolutional polar coded modulation structure is proposed,which maps the bit channels derived from one symbol to different polar codes such that each bit channel used by a polar code is independent from each other.Simulation results show that the proposed method can further improve the spectral efficiency of the existing polar-coded modulation schemes under a fixed number of hardware resources.Lastly,how to improve the network coverage of IoT is a main concern for 5G mMTC scenario.The polarization-adjusted convolutional codes(PAC Codes)can tolerate more transmission loss with its extremely high coding gain,which serves as a promising solution for improving the network coverage of IoT.PAC codes can be regarded as a kind of irregular tree code,thus they can be decoded with tree search heuristics.However,conventional tree search algorithms designed for memoryless channels can not be directly applied for PAC codes,as PAC codes exploit the polarized channels which have memories.As such,this dissertation first introduces the stack decoding procedure for an irregular tree code under memoryless channels,then the memoryless channels are substituted with the polarized channels,where the branch metric of stack decoding is designed accordingly.On this basis,a stack decoding algorithm for PAC codes is proposed,and the decoding performance and complexity performance are also studied.Simulation results show that,for PAC codes with different coding rates,the proposed algorithm virtually achieves the optimal decoding performance of the finite block-length codes.Moreover,when the current SNR is higher than the decoding threshold of the cutoff rate,the average complexity required by the proposed algorithm to decode one bit is almost a constant,and this dissertation also shows that under the same computational complexity,the performance of PAC code under the proposed stack decoding algorithm outperforms that of the conventional polar code under CA-SCL decoding. |
PDF Full Text Request