基于多目标强化学习的抗强干扰Polar编码优化方法  

Anti-strong Jamming Polar Coding Optimization Method with Multiobjective Reinforcement Learning

作　　者：梁豪 叶淦华 陆锐敏 王恒[1] 魏鹏 LIANG Hao;YE Ganhua;LU Ruimin;WANG Heng;WEI Peng(Sixty-third Research Institute,National University of Defense Technology,Nanjing 210001,China)

机构地区：[1]国防科技大学第六十三研究所,南京210001

出　　处：《电子与信息学报》2023年第11期4092-4100,共9页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(62201596);国防科技大学科研计划(ZK22-45)。

摘　　要：为提升跳频(FH)通信系统信息传输的可靠性和抗干扰能力,该文基于新型Polar编码的慢跳频抗干扰通信系统模型,提出一种适应强干扰环境的Polar编码构造优化方法。首先,面向包含常态和干扰态的混合信道设计多目标强化学习算法,然后优化编码过程中的信息位比特信道序列,提升码字的纠错性能,并通过初始化预处理和理论计算回报值降低算法执行复杂度。仿真结果表明,在包含强干扰的混合信道条件下,所提编码优化方法的全局误码性能优于传统编码构造方法,相比于第5代移动通信系统(5G)第3代合作伙伴计划(3GPP)标准方案全局编码增益达0.5 dB,有效改善Polar编码跳频通信高可靠抗干扰传输性能。In order to improve the reliability and anti-jamming ability of information transmission for the Frequency-Hopping(FH)communication system,a Polar coding construction optimization method is proposed to adapt to the strong-jamming environment,which is based on a novel Polar coded slow FH communication system model.Firstly,the multi-objective reinforcement learning algorithm is designed for the hybrid channel containing normal state and jamming state,and then the information bit-channel sequence in the coding process is optimized.Consequently the error correction performance of the designed Polar codewords is improved.In addition,the complexity of algorithm is reduced by preprocessing the initialization and theoretically calculating the reward values.The simulation results show that the overall error performance of the proposed coding optimization method is better than those of conventional coding construction methods in the hybrid channel containing strong jamming.Compared with the 3rd Generation Partnership Project(3GPP)standard scheme in Fifth-Generation(5G)mobile communication systems,the obtained overall coding gain is up to 0.5 dB.Therefore the high-reliability and anti-jamming performance of Polar coded FH transmission is effectively improved.

关 键 词：信道编码 抗干扰 Polar码 强化学习 可靠性能 

分 类 号：TN911.2[电子电信—通信与信息系统]