| Channel coding is one of the most important technologies in the wireless communication sys-tems,which helps to improve the reliability of the communication systems and maximize the capac-ity during interference.As the first channel coding that theoretically reaches shannon limit,polar code has caused widespread concern by many academic researchers.In practice,there is high re-quirement user data throughput and latency in the Enhanced Mobile Broadband(e MBB)scenario of5 GNR,and it is necessary to develop high throughput polar coded techniques.Incremental Redun-dancy Hybrid Automatic re Quest(IR-HARQ)is a technology combining the channel coding and Automatic Re Quest(ARQ)to achieve reliable communication.The polar coded IR-HARQ scheme has the potential to provide higher throughput performance.In this thesis we mainly research on maximizing the throughput of polar coded IR-HARQ scheme based on deep reinforcement learn-ing,and we further study a Turbo-like iterative decoding algorithm based on stage-permuted tanner graph of polar code.This thesis first introduces the basics of polar code.Specifically,we introduce the principle of polarization of polar code,and the construction method of Gaussian approximation in detail.Then We introduce and the encoding algorithm of polar code.Later,we described several decoding al-gorithms of polar code,including Successive Cancellation(SC)algorithm,CRC-Aided Successive Cancellation List(CA-SCL)algorithm and Cancellation of the Soft Output(SCAN)algorithm,and we compare the performances and the complexity of different algorithm mentioned above.Secondly,we proposed an incremental bits optimization algorithm of the polarizing matrix extension IR-HARQ scheme by using Deep Deterministic Policy Gradient(DDPG)algorithm.In this thesis,the process of polar coded IR-HARQ is modeled as a Markov Decision Process(MDP),the transmitter is viewed as the Agent of MDP,the receiver and channel are jointly viewed as the environment,and a novel reward function that is negatively correlated with the number of incre-mental redundant bits is designed to ensure the reasonability of the learning process.Simulation results show that the throughput of the DDPG based IR-HARQ scheme significantly outperform the traditional polar coded HARQ scheme.Finally,we research on the Turbo-like decoding framework of polar code.we firstly intro-duce the basics of the Turbo-like decoding algorithm,and then the construction algorithm named progressive RRBP freezing algorithm is introduced.In order to solve the problem of insufficient performance gain of progessive RRBP freezing algorithm,this thesis combines the spatial coupling technology with Turbo-like iterative decoding algorithm to improve the decoding performance un-der code rate.In order to further improve the throughput of hardware implementation,a semi-parallel spatial coupling decoding scheme is proposed subsequently in detail.Compared with the original forward-backtracking coupling decoding scheme,the proposed semi-parallel decoding al-gorithm reduced the latancy and increase the data throughput.Finally,the simulation results show the effectiveness of the proposed algorithm. |
