免授权MIMO-NOMA系统中活动用户检测与信道估计算法

Active user detection and channel estimation algorithm in grant-free MIMO-NOMA system

针对免授权多输入多输出非正交多址接入(multi-input multi-output non-orthogonal multiple access,MIMO-NOMA)系统中的活动用户检测和信道估计问题,提出了一种Q学习辅助的空间相关块稀疏贝叶斯学习算法。该算法将活动用户检测和信道估计问题建模为多维块稀疏信号恢复问题,基于块稀疏贝叶斯原理推导代价函数,并将代价函数的优化过程描述为马尔可夫决策过程,把Q学习引入稀疏贝叶斯学习框架,以实现活动用户检测和信道估计。仿真结果表明,该算法信道估计的归一化均方误差低于0.01,活动用户检测错误率低于10-5。

A Q-learning assisted spatial correlation block sparse Bayesian learning algorithm was proposed to solve the problems of active user detection and channel estimation in grant-free multi-input multi-output non-orthogonal multiple access(MIMO-NOMA)systems.In this algorithm,the problem of active user detection and channel estimation was modeled as a multi-dimensional block sparse signal recovery problem.The cost function was derived based on the block sparse Bayesian principle,and the optimization process of the cost function was described as a Markov decision process.Q learning was then introduced into the sparse Bayesian learning framework to achieve active user detection and channel estimation.Simulation results show that the normalized mean square error of channel estimation of the proposed algorithm is less than 0.01 and the detection error rate of active users is less than 10-5.