Feedback-Topology Designs for Interference Alignment in MIMO Interference Channels

Abstract

Interference alignment (IA) is a joint-transmission technique that achieves the maximum degrees-offreedom (DoF) of the interference channel, which provides linear scaling of the capacity with the number of users for high signal-to-noise ratios (SNRs). Most prior work on IA is based on the impractical assumption that perfect and global channel-state information (CSI) is available at all transmitters. To implement IA, each receiver has to feed back CSI to all interferers, resulting in overwhelming feedback overhead. In particular, the sum feedback rate of each receiver scales quadratically with the number of users even if the quantized CSI is fed back. To substantially suppress feedback overhead, this paper focuses on designing efficient arrangements of feedback links, called feedback topologies, under the IA constraint. For the multiple-input-multiple-output (MIMO) K-user interference channel, we propose the feedback topology that supports sequential CSI exchange (feedback and feedforward) between transmitters and receivers so as to achieve IA progressively. This feedback topology is shown to reduce the network feedback overhead from a quadratic function of K to a linear one. To reduce the delay in the sequential CSI exchange, an alternative feedback topology is designed for supporting two-hop feedback via a control station, which also achieves the linear feedback scaling with K. Next, given the proposed feedback topologies, the feedback-bit allocation algorithm is designed for allocating feedback bits by each receiver to different feedback links so as to regulate the residual interference caused by the finite-rate feedback. Simulation results demonstrate that the proposed bit allocation leads to significant throughput gains especially in strong interference environments.

Limited Feedback Design for Interference Alignment on MIMO Interference Networks With Heterogeneous Path Loss and Spatial Correlations

“…However, we start with matrix analysis and overcome these difficulties to get a comparatively more general result. Second, the LB provided in Lemma 2 is tighter than the conventional analysis result in [21] (The LB in [21] is equivalent to the right side of (20) by setting the first E{I j } = 0.). The numerical comparison of the two bounds is also illustrated in Section V.…”

“…can better bound the DFS than conventional LB derived according to [21], especially in low feedback bits regime.…”

“…First, it is applicable to general d (d ≥ 1) data stream cases. Note that many previous works on limited feedback system [7], [8], [21] focus on single stream case (d = 1) due to the mathematical difficulty to analyze matrix functions. However, we start with matrix analysis and overcome these difficulties to get a comparatively more general result.…”

Performance analysis of MAX-SINR algorithm under CSI mismatch

“…Since the performance of IA under quantized feedback has been widely investigated in the literature (see e.g., [10][11][12][13][14]), in this paper we consider a rather generalized imperfect CSI model where the variance of the measurement error is a function of SNR. We especially evaluate the performance of maximum signal-to-interference-plus-noise ratio (Max-SINR) algorithm described in [2] under CSI mismatch.…”

Adaptive MMSE-based beamformer design for multiuser MIMO interference channels

“…Most of the efforts have been focused on beamformer design for constant MIMO IC under the availability of perfect CSI (see e.g., [2]- [7]). Those which have been done regarding the beamformer design under imperfect CSI are mainly related to quantized feedback scenarios with channel-aware receivers (see e.g., [8], [9], [11], [12]). Moreover, in [13], beamformer design has been considered when the channel estimation error variance is a constant, and in [14], robust transceiver design for multiuser MIMO IC has been considered under ellipsoidal CSI uncertainties.…”