This paper investigates transmission schemes for fixed-rate communications over a Rayleigh block-fading interference channel. There are two source-destination pairs where each source, in the presence of a short-term power constraint, intends to communicate with its dedicated destination at a fixed data rate. It encodes its messages using a point-to-point Gaussian codebook. The two users' transmissions can be conducted orthogonally or non-orthogonally. In the latter case, each destination performs either direct decoding by treating the interference as noise, or successive interference cancellation (SIC) to recover its desired message. For each scheme, we seek solutions of a power control problem to efficiently assign power to the sources such that the codewords can be successfully decoded at destinations. However, because of the random nature of fading, the power control problem for some channel realizations may not have any feasible solution and the transmission will be in outage. Thus, for each transmission scheme, we first compute a lower bound and an upper bound on the outage probability. Next, we use these results to find an outer bound and an inner bound on the ϵ -outage achievable rate region, i.e., the rate region in which the outage probability is below a certain value ϵ.

This paper considers distributed transceiver design and power control for K-user multiple-input-multiple-output interference networks. Each source intends to send multiple independent data streams to its corresponding destination where the number of data streams coincides with the degrees of freedom of the network. Each data stream is encoded at a fixed data rate, whereas different streams can be encoded at possibly different rates. We assume that only local channel side information (i.e., knowledge related to channels directly connected to a terminal) can be acquired by each terminal. We propose iterative algorithms to perform both power control and transceiver design. Transmitter beamforming matrices and receiver filtering matrices are designed to maximize signal-to-interference-plus-noise ratio corresponding to each stream, and a power control scheme is performed to assign the minimum power to each encoded data stream such that successful communication can be guaranteed. The proposed algorithms exhibit a substantial performance improvement compared with the conventional orthogonal transmission schemes.

Considering the time-varying K-user single-antenna interference channel (IC), it has been shown that, when terminals have perfect global channel state information (CSI) and they can tolerate asymptotically long delay, applying an ergodic interference alignment (EIA) scheme can achieve half of the interference-free achievable rate. However, in practice, obtaining such CSI is challenging, and only a limited delay is acceptable. This paper addresses data transmission over the IC by taking these concerns into account. Specifically, we consider the case that each transmitter attains only quantized CSI via limited feedback signals. This causes imperfect interference alignment and a degraded performance. We propose adaptive schemes to compensate the impact of the CSI uncertainties. We first study a power control problem which is concerned with communicating at fixed rates using minimum transmit powers. A power control algorithm is used to reach the solution. Next, we address a throughput maximization problem when the transmit powers are fixed. Through the analysis of system outage probability, we propose a rate adaptation scheme to maximize throughput. Finally, we quantify the throughput loss in delay-limited systems. Our results show that, even with limited feedback, performing the EIA scheme with proper power control or rate adaptation strategies can still outperform conventional orthogonal transmission approaches.

This paper considers the ergodic block fading multi-user Gaussian interference channel (IC) in which each source desires to communicate to an intended destination. We assume that there is no CSI a priori available at terminals. We develop achievable rate results and compute the associated degrees of freedom by using a pilot-assisted interference alignment scheme. In this scheme, each source first sends known pilot symbols via which the destinations estimate channel gains, and the destinations then broadcast the estimated channel gains via orthogonal feedback channels. The estimated channel gains are used to perform interference alignment for data transmission. The pilot transmission power can be different from the data transmission power. By allocating more power to pilot transmission, channel gains can be estimated more accurately which implies less power left for data transmission. We find the optimum power allocation to pilot symbols and data symbols. Our study recommends, in large networks, to allocate more power to channel training instead of data transmission. In addition, our results reveal that for a K-user ergodic IC with a coherence time T, the total degrees of freedom 1/2 Kopt(1-Kopt/T) is achievable, where Kopt= min {K, T/2} is the optimum number of users selected to be active in the network. This recommends to perform a user selection in large networks (K > T/2), and apply channel training and interference alignment within the set of selected users.

We investigate the time-varying Gaussian interference channel (IC) in which each source desires to communicate to an intended destination. For the ergodic time-varying IC with global perfect CSI at all terminals, it has been known that with an interference alignment technique each source-destination pair can communicate at half of the interference-free achievable rate. In practice, the channel gains are estimated by transmitting known pilot symbols from the sources, and the channel estimation procedure is hence prone to errors. In this paper, we model the channel estimation error at the destinations by an independent additive Gaussian noise and study the behavior of the ergodic interference alignment scheme with the global noisy CSI at all terminals. Toward this end, we present a closed-form inner bound on the achievable rate region by which we conclude that the achievable degrees of freedom with global perfect CSI can be preserved, if the variance of channel estimation error is proportional to the inverse of the transmitted power.

This paper considers joint transceiver design and power control for K-user multiple-input multiple-output (MIMO) interference networks. Each source intends to communicate with its corresponding destination at a fixed data rate. Only local channel side information (i.e. knowledge related to the channels directly connected to a terminal) is available at each terminal. We propose iterative algorithms to perform power control to guarantee successful communication while designing transmitter beamforming matrices and receiver filtering matrices according to the interference alignment concept. The proposed algorithms can exhibit a substantial performance improvement compared to the conventional orthogonal transmission schemes.

We study the available degrees of freedom (DOF) of a class of wireless single-antenna multi-user relay networks. In these networks the communications between M unconnected source-destination pairs are provided by a large number of half-duplex relays. To conduct the communications we propose a cluster successive relaying protocol that divides the relays into two equal-size clusters. Unlike the conventional orthogonal relaying protocol that demands all the relays to simultaneously assist the sources, we require the two relay clusters to take turns forwarding the source messages to more efficiently use the channel. In a time-varying fading environment, through appropriate interference alignment the negative impact of inter-user interference can be effectively minimized. Thus the two clusters of half-duplex relays can mimic a cluster of full-duplex relays. When the number of relays is infinitely large, we show that the M-user half-duplex relay networks have M DOF, i.e. their sum capacity can be characterized as CΣ(SNR) = M log (SNR) + o(log (SNR)). This result implies that allowing only distributed processing and half-duplex operation is able to provide the same DOF performance as permitting joint processing and full-duplex operation in wireless relay networks.

Considering a single-antenna M-user interference channel with symmetrically distributed channel gains, when the channel state information (CSI) is globally available, applying the ergodic interference alignment scheme, each transmitter-receiver pair achieves a rate proportional to ½ of a single user's interference-free achievable rate. This is substantially higher than the achievable rate of the conventional orthogonal transmission schemes such as TDMA. Since the rigid requirement on the CSI may be difficult to realize in practice, in this paper we investigate the performance of applying the ergodic interference alignment scheme when the estimation of each channel gain is made globally known through exploiting only a limited feedback signal from the associated receiver of that channel. Under a block fading environment, we provide a lower bound on the achievable average throughput of the network. Our results imply that the better performance of interference alignment over TDMA may still exist even without the assumption of perfect CSI. Also, the trade off between allocating feedback rate of each receiver to the desired channel or the interference channels at deferent SNR region investigated.

We study the degrees of freedom (DOF) of a single- antenna M-user time-varying parallel relay network, where the communications between M pairs of unconnected sources and destinations are provided by a large number of half-duplex decode-and-forward (DF) relays. Unlike the conventional relaying strategy which demands all the relays to simultaneously assist the sources, we divide the relays into two clusters and permit them to take turns forwarding the source messages. With appropriate interference alignment design, it is proved that the M-user time-varying relay network has M DOF, provided that the number of relays is infinitely large.