User Selection for Multiuser MIMO Downlink With Zero-Forcing Beamforming

Abstract

In this paper, we propose a greedy user selection with swap (GUSS) algorithm based on zero-forcing beamforming for multiuser multiple-input-multiple-output (MIMO) downlink channels. Existing user selection algorithms such as zero forcing with selection (ZFS) have the flaws of "redundant users" and "local optimum," which compromise the achieved sum rate. GUSS improves the performance by adding the "delete" and "swap" operations to the user selection procedure of ZFS to eliminate "redundant users" and escape from "local optimum," respectively. In addition, an effective-channel-vector-based effective-channelgain-updating scheme is proposed to reduce the complexity of GUSS. With the help of this updating scheme, GUSS has the same order of complexity as ZFS with only a linear increment. Simulation results indicate that over the range of transmit signal-tonoise ratios (SNRs) considered, on average, the sum rate of GUSS reaches 99.3% of the upper bound that is achieved by exhaustive search, with only 1.51 to 2.29 times the complexity of ZFS.Index Terms-Broadcast channel, multiuser multiple-inputmultiple-output (MIMO), user selection, zero-forcing beamforming (ZFBF).

A novel low-complexity user selection scheme for multiuser MIMO downlink with zero-forcing beamforming

“…The complexity reduction method in ZFS updates γ via iteratively updating H † U and is only applicable when a new user is added to U. In [3], the authors presented an efficient γ updating strategy which is suitable for swapping via iteratively updating ECV ν defined in (2). This updating strategy will be adopted in our USS algorithm to further reduce the complexity.…”

“…• Update w u j , ν u j and γ u j utilizing the updating scheme presented in [3], denote the achieved rate R…”

“…The 'swap' operation in step 2) requires evaluating K − M sum rates R k 1 . According to equations (45)- (48) of [3], each R k 1 involves NM +4N complex multiplications. Repeating this over K − M users in U, we obtain the complexity of step 2) as (K − M )(NM + 4N ) complex multiplications.…”

“…The 'delete' operation in step 3) requires evaluating M − 1 sum rates R k 2 . According to equations (40)- (41) of [3], each R k 2 needs M − k 2 γ u j updates in total, which requires M − k 2 vector-vector multiplications, i.e., involves…”

“…For example, the zero-forcing with selection (ZFS) algorithm proposed in [2], starts with an empty user set and adds one user in each iteration to maximize the sum rate increment. In [3], S. Huang et al proposed a greedy user selection with swap (GUSS) to improve the sum rate performance with a bit higher complexity than ZFS. Semi-orthogonal user selection (SUS) based on searching for a user set with near-orthogonal channel vectors is proposed in [4].…”