项目名称: 基于交替方向乘子法的高效译码理论与算法研究
项目编号: No.61471286
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 无线电电子学、电信技术
项目作者: 焦晓鹏
作者单位: 西安电子科技大学
项目金额: 80万元
中文摘要: 交替方向乘子法(Alternating Direction Method of Multipliers,ADMM)通过分解-协作的方式将大规模优化问题分解为可分布式求解的多个小规模优化问题。ADMM的高效性和分布式特点,使其适用于LDPC码的译码。本项目旨在研究基于ADMM的译码理论及算法,设计具有极低复杂度且性能良好的实用译码方案。具体内容包括:研究ADMM译码算法动态消息调度策略,设计合理的消息量化机制;针对循环/准循环LDPC码,设计高效的ADMM译码算法,降低运算和存储开销;借助重要性抽样理论,估计ADMM译码在错误平层区域的性能,研究降低错误平层的各种方法;基于凸优化和抽象代数理论,设计多元LDPC码的ADMM译码算法;研究基于ADMM的LDPC码联合检测译码算法。通过本项目的研究,期望能够进一步探索基于优化技术的译码理论与算法,为未来的通信和存储系统提供可靠、高效的译码方案。
中文关键词: 信道编码;交替方向乘子法;LDPC码;联合检测与译码;线性规划译码
英文摘要: By using a decomposition-coordination procedure, the alternating direction method of multipliers (ADMM) decomposes large scale optimization problems into small problems which can be solved distributely. The high efficiency and distribution features of ADMM makes it suitable for decoding LDPC codes. This project aims to study the ADMM-based decoding theory and algorithms, and design practical decoding schemes with low complexity and excellent performance. The research contents of this project are summarized as follows: (1) Investigate dynamic message scheduling methods and design efficient message quantification schemes for ADMM decoding algorithms. (2) Design efficient ADMM decoding algorithms for cyclic/quasi-cyclic LDPC codes,and lower the computation and stroage overhead. (3) Estimate decoding performances of ADMM algorithms at error floor region by using the importance sampling technique, and investigate various methods to lower the error-floor performance. (4) Based on convext optimization and abstract algebra, we design efficient ADMM decoding algorithms for non-binary LDPC codes. (5) Study ADMM algorithms for joint detection and decoding problems with LDPC codes. It is expected that new theories and methods obtained during this research will explore intensive applications of optimization theory in the area of channel coding, and provide effient and reliable decoding schemes for future communication and storage systems.
英文关键词: channel coding;alternating direction method of multipliers (ADMM);LDPC codes;joint detection and decoding;linear-programming decoding