This paper considers the multiantenna multiple-access relay channel (MARC), in which multiple users transmit messages to a common destination with the assistance of a relay. In a variety of MARC settings, the dynamic decode-and-forward (DDF) protocol is very useful due to its outstanding rate performance. However, the lack of good structured codebooks so far hinders practical applications of DDF for MARC. In this work, two classes of structured MARC codes are proposed: 1) one-to-one relay-mapper-aided multiuser lattice coding (O-MLC); and 2) modulo-sum relay-mapper-aided multiuser lattice coding (MS-MLC). The former enjoys better rate performance, whereas the latter provides more flexibility to tradeoff between the complexity of the relay mapper and the rate performance. It is shown that, in order to approach the rate performance achievable by an unstructured codebook with maximum-likelihood decoding, it is crucial to use a new K-stage coset decoder for structured O-MLC instead of the one-stage decoder proposed in previous works. However, if O-MLC is decoded with the one-stage decoder only, it can still achieve the optimal DDF diversity-multiplexing gain tradeoff in the high signal-to-noise ratio regime. As for MS-MLC, its rate performance can approach that of the O-MLC by increasing the complexity of the modulo-sum relay-mapper. Finally, for practical implementations of both O-MLC and MS-MLC, practical short-length lattice codes with linear mappers are designed, which facilitate efficient lattice decoding. Simulation results show that the proposed coding schemes outperform existing schemes in terms of outage probabilities in a variety of channel settings, especially when the users-to-relay links are better than the other channel links.

The dynamic decode-and-forward (DDF) protocol for the multiple access relay channel (MARC) with quasi static fading is evaluated using the Zheng-Tse diversity-multiplexing tradeoff (DMT). We assume that there are two users, one half-duplex relay, and a common destination, each equipped with single antenna. For the Rayleigh fading channel, the DDF protocol is well known and has been analyzed in terms of the DMT with the infinite block length assumption by Azarian et al. However, to make the protocol feasible, the practical constraint of finite block length must be enforced, which may result in a loss in the DMT. Another practical effect not considered in the infinite block length DDF protocol is the possible decoding error at the relay. By carefully dealing with these practical issues due to finite block length, we characterize the finite block length DMT of the DDF protocol. We further consider the situation where the destination does not have a priori knowledge of the relay decision time at which the relay switches from listening to transmitting, and show that the optimal DMT is still achievable as if there is no decoding error at the relay. Therefore, the assumption of error-free decoding at the relay and additional protocol overhead to communicate the decision time are not needed for the DDF to achieve the optimal DMT.

Linear-assignment Gel'fand-Pinsker coding (LA-GPC) is a coding technique for channels with interference known only at the transmitter, where the known interference is treated as side-information (SI). As a special case of the LA-GPC, dirty paper coding has been shown to be able to achieve the optimal interference-free rate for SI channels with perfect channel state information at the transmitter (CSIT). In the cases where only the channel distribution information at the transmitter (CDIT) is available, LA-GPC also has good (sometimes optimal) performance in a variety of fast and slow fading SI channels. In this letter, we design filters in the nested lattice based coding to make it achieve the same rate performance as the LA-GPC in multiple-input multiple-output (MIMO) channels. Compared with the random Gaussian codebooks used in previous works, our resultant coding schemes have algebraic structures and can be implemented in practical systems. Simulations in slow-fading channels are provided, and near interference-free error performance is obtained. The proposed coding schemes can serve as the fundamental building blocks to achieve the promised rate performance of MIMO Gaussian broadcast channels with CDIT or perfect CSIT.

In this paper, we present the design of cognitive radio in the Rician channel with partial channel state information at the transmitter (CSIT). We replace the dirty paper coding (DPC) used in the cognitive radio with full CSIT by the linear assignment Gel'fand-Pinsker coding (LA-GPC) which can achieve better error performance when there is only partial CSIT. Based on the achievable rate derived from the LA-GPC, two optimization problems under the fast and slow fading channels are formulated. We derive semi-analytical solutions to find the relaying ratios and precoding coefficients. We also show that the parameters derived by the proposed methods converge to the optimal full CSIT solutions in the asymptotic cases. This result verifies the correctness of the proposed methods asymptotically. Moreover, a new coding scheme is proposed to implement the LA-GPC in practice. Simulation results show that the proposed semi-analytical solutions perform close to the optimal solutions found by brute-force search, and outperform the systems based on naive DPC. Simulation results also show that the proposed practical coding scheme can effectively approach the theoretical rate performance.

A result of Zheng and Tse states that over a quasi-static channel, there exists a fundamental tradeoff, known as the diversity-multiplexing gain (D-MG) tradeoff. The quest for space-time codes that achieve D-MG tradeoff has generated numerous works. However, these space-time codes generally result in a pretty high peak to average power ratio (PAR) value on each antenna. In this paper, we propose two general ways to reduce PAR without affecting code structures and without any side information being transmitted. Therefore, the modified space-time codes still achieve the D-MG tradeoff. The two methods are both based on constellation shaping. The first method is similar to the method proposed by Kwok except we employ the Hermite Normal Form (HNF) decomposition instead of the Smith Normal Form (SNF); the second one takes the idea of integer reversible mapping. Both techniques lead to an asymptotical PAR value equal to 3 as the constellation size becomes large. Sphere decoding is performed to verify that the proposed PAR reduction methods do not affect the performance of space-time codes.