基于KNN的MIMO-OFDM系统链路自适应研究

Research on link adaptation of MIMO-OFDM system based on KNN

针对传统模型难以建立信道状态与复杂MIMO-OFDM系统性能间的映射关系,结合监督学习在处理非线性问题中的优势,提出基于K-最近邻算法的自适应编码模型。以V-BLAST为基础模型,以处理信噪比SVDSNR作为信道特征,通过KNN对信道特征分类,从而建立信道特征与MCS切换间的映射关系。实验结果表明,在K=35,样本划分为7∶3时,MCS切换分类准确率最高;同时在相同实验条件下,本文提出的KNN自适应调制编码能快速适应信道环境,且BER和系统吞吐量都要明显优于传统查找表算法,说明此方法可行。

Based on the problem of the traditional model which is difficult to establish the mapping relationship between channel state and the performance of complex MIMO-OFDM system,and Combined with the advantages of supervised learning in dealing with nonlinear problems,an adaptive coding model based on K-nearest neighbor algorithm is proposed.V-BLAST is used as the basic model to deal with the signal-to-noise ratio SVD_SNR as the channel characteristics,and classifies the channel features by KNN,and establishes the mapping relationship between the channel characteristics and MCS handoff.The experiment results show that when K=35 and the sample is divided into 7∶3,the MCS handoff classification accuracy is the highest;at the same time,under the same experimental conditions,the KNN adaptive modulation and coding proposed in this paper can quickly adapt to the channel environment,and BER and system throughput are significantly better than the traditional look-up table algorithm,which shows that the method is feasible.