A burst-form CSI estimation approach for FDD massive MIMO systems

Abstract

Pilot and channel state information (CSI) feedback overhead in the downlink and uplink paths are two major implementation challenges in frequency-division duplex (FDD) based massive MIMO systems. When the massive MIMO channel satisfies the burst-sparsity property, we can acquire the channel with compressed pilots and CSI feedback in a more efficient approach. This paper proposes a burst-form estimation approach, referred to as the burst-form least squares (BFLS) algorithm, to fully utilize the burst-sparsity property of massive MIMO channels. The proposed algorithm is based on knowledge of the starting location of each burst at the user side. For situations where the starting locations change quickly or are otherwise initially unknown at the user, a starting point estimation (SPE) algorithm is proposed to provide the position of each burst in the channel vector. Numerical results demonstrate that the BFLS algorithm acquires the channel better than competing approaches and reaches the performance upper bound. It is shown that the SPE algorithm can find the location of bursts with high accuracy and using the estimated values do not significantly degrade the estimation quality.

Channel estimation for FDD multi‐user massive MIMO systems: a greedy approach based on user clustering

“…The BS side receives the feed-back signal from both nearby and faraway users. Therefore, the overall channel matrix S including the channel of K users in the cell does not satisfy a particular pattern in the zero or non-zero coefficients as defined by (6) and (7). In this section, we intend to introduce a clustering process based on mean AoAs received by the uplink signals, which imposes a certain number of groups with a specific sparsity pattern among the users inside the groups.…”

“…The predicted goals of this technology would be achievable if the base station (BS) can estimate channel state information (CSI) with high precision [3]. In frequency-division duplex (FDD) systems, pilot and CSI feedback overhead are the key challenges in the channel acquisition problem [4][5][6], while time-division duplex (TDD) protocol profits from the channel reciprocity property, and thus, simplifies CSI estimation operation [7,8]. In other words, the consumed training signals in FDD-based systems that use conventional channel estimation approaches are proportional to the number of transmit antennas, which could be a large array in the most real scenarios.…”

“…Due to the great commercial interest of FDD-based massive MIMO systems, many efforts have been devoted to addressing these issues in the literature, e.g. [5,6,[9][10][11][12].…”

“…However, the computational complexity of the optimisation algorithm would be heavy, especially when the size of antenna array grows up. The burst-sparse structure of massive MIMO channel has been studied in [6,19]; to handle the assumed channel model, the burst version of LASSO approach has been designed in [19] and a new version of LS approach has been proposed in [6], when the starting location of each burst is previously known by the users. Moreover, if the user needs to estimate the starting points, [6] introduces a novel starting point estimation algorithm to find the location of each burst in the channel vector.…”

“…The burst-sparse structure of massive MIMO channel has been studied in [6,19]; to handle the assumed channel model, the burst version of LASSO approach has been designed in [19] and a new version of LS approach has been proposed in [6], when the starting location of each burst is previously known by the users. Moreover, if the user needs to estimate the starting points, [6] introduces a novel starting point estimation algorithm to find the location of each burst in the channel vector. In [20], the slow variations of channel in the consecutive time slots and also, the spatially common channel support in the frequency domain are used to reconstruct the sparse channel.…”