Diversity-multiplexing tradeoff improvement in interference networks

Abstract

In this dissertation, we propose the protocols that improve the performance of various interference networks in terms of diversity-multiplexing tradeoff (DMT).

We first present a spectrally efficient protocol for half-duplex multi-relay systems in block fading channels where a direct source-destination link is unavailable. The proposed protocol adaptively selects either successive interference cancelation (SIC) or joint decoding according to the causal decoding status of each relay. We also adopt dynamic refreshing that restarts the protocol whenever it is advantageous to do so, even if the relay decoding set (the set of relays that are able to decode the message) is not empty. The achievable DMT of the proposed protocol with m-antenna nodes is analyzed via a Markov chain whose states are related to the cardinality of a decoding set. This protocol strictly improves the DMT of the existing DF half-duplex relay-selection protocols without decoding delay, and in the low multiplexing gain region is able to meet the DMT upper bound. The main contributions of this work are the state-dependent decoding strategies in DF multi-relay systems and also the dynamic refresh for the flushing of residual interferences in the system, concepts that may find usefulness beyond the gains in the high-SNR regime.

Second, in order to improve diversity gain in interference networks of which links are quasi-static and hence to maximize diversity multiplexing tradeoff (DMT), we propose on-off filtered interference alignment (IA) that controls the time portion of IA usage in order to achieve the highest diversity gain for given multiplexing gain. The on-off filtered IA considered in this paper is based on the two schemes: (1) conventional IA and (2) IA with Alamouti coding, both of which require local channel state information at transmitter only. Each receiver adopts the simultaneous non-unique decoder. Analyzing and deriving DMT of the proposed schemes in closed form, we reveal that the on-off filtered IA can considerably improve DMT in the $2 \times 2$ X-channel. Both of the proposed schemes are shown to achieve diversity gain of 4 and DoF per user of 4/3. Consequently, the on-off filtered IA scheme based on conventional IA outperforms IA with Alamouti coding in terms of DMT for $0 \leq r \leq 1$, where r is the multiplexing gain. The on-off filtered IA scheme based on IA with Alamouti coding, to the best of our knowledge, surpasses any other existing schemes for the $2 \times 2$ X-channel with two antennas and approaches the DoF given by the linear function connecting maximum diversity gain and DoF per user, (0,4) and (4/3,0), respectively, i.e., d(r)=4-3r.

Third, we study a new interference channel called 4-user clustered Z-channel where 4 clusters consisting of two single-antenna nodes each construct a Z interference channel and the connected two clusters form a two-user interference channel. We analyze DMT in this channel by joint decoding. We also propose a new interference alignment scheme requiring only partial channel phase information and analyze DMT for the proposed interference alignment scheme. We also propose the time sharing scheme with both joint decoding and interference alignment. Eventually, we show that the optimized proposed time sharing scheme according to a time portion closely approaches to the DMT upper bound. It is shown that the proposed scheme is DMT optimal in the range between 0 and 0.2 and closed to DMT optimal in the remained multiplexing gain region, i.e., $0.2 \leq r \leq 0.5$.

Forth, we analyze DMT of K-user interference channels. We propose a time-sharing scheme using both joint decoding and IA, and derive a closed form of its achievable DMT. We show that the achievable DMT is not limited by the maximum of DMTs achieved by joint decoding and IA each but is much better than the maximum of the two. As K increases, the achievable DMT converges, but the DMT improvement by the time-sharing scheme becomes more remarkable. To the best of our knowledge, the time-sharing scheme outperforms any other known schemes in the K-user interference channel in terms of DMT.

Finally, we propose the two on-off filtered IA schemes with the delayed CSIT in 3-user MIMO interference channel with two antennas based on the retrospective IA with post-processing. We analyze diversity multiplexing tradeoff (DMT) of the proposed scheme. We derive DMT of the proposed scheme in closed form. The proposed scheme was shown to achieve diversity gain of 4 and DoF per user of 4/5 (i.e., sum DoF of 12/5 which is the optimal DoF with the delayed CSIT in 3-user MIMO interference channel, 6M/5) if the IA usage portion is optimized. The on-off filtered retrospective IA scheme outperforms both of referential schemes while achieving maximum diversity gain and multiplexing gain per user, i.e., (0,4) and (4/5,0), respectively.