Improved LAS detector for MIMO systems with imperfect channel state information

Abstract

Likelihood ascent search (LAS) detector is a neighbourhood search algorithm and one of the simplest schemes for low-complexity near-optimal detection in massive multiple-input multiple-output systems. LAS detector design under the assumption of perfect channel state information has been an area of research for decades. However, channel estimation errors have never been taken into account by conventional LAS detectors when calculating the maximum-likelihood decoding metric. As a result, the bit error rate performance of LAS detectors can be significantly degraded. This study proposes robust LAS detectors which take channel estimation errors into account in the computation of the ML decoding metric. The proposed approach involves the computation of the equivalent noise covariance matrix inverse, which may increases the computational complexity. Therefore, the authors also propose a low complexity method to inverse the covariance matrix. Simulation results show that the proposed schemes outperform the conventional LAS detector.