Non-Negative Least Squares Exploiting Multiple Measurements in Unsourced Random Access

Unsourced random access (URA) is considered as a key enabling technology to support a large number of user equipments (UEs) in massive machine-type communication (mMTC). As the number of devices increases, mMTC is demanding more efficient active detection (AD) technique. The requirement of determined AD signal model poses a bottleneck for latency reduction by shortened channel usage. Recently, non-Bayesian AD algorithms using non-negative least squares (NNLS) alleviates such bottleneck by exploiting the covariance information of received signal with improved AD capacity with limited observations. Although NNLS guarantees convergence to the local optimum point, there is still improvement space to be exploited. This work points out the limitation of NNLS framework under limited block length and short transmission period. Correspondingly, an iterative NNLS algorithm by multiple measurement vector (MMV-NNLS) mitigates inherent problems in NNLS. Numerical results demonstrate the validity of the proposed algorithm.