Distributive cooperation in multiple input multiple output systems

Abstract

The multiple-input multiple-output (MIMO) technology involving multiple antennas at both transmitter and receiver ends has been extensively studied for future wireless communication systems. Channel state information (CSI) is crucial for MIMO systems. In particular, CSI at the transmitter (CSIT) can be exploited to improve the system performance through the use of various techniques such as singular value decomposition (SVD), beam-forming and water-filling techniques. Cooperative techniques can further enhance the performance of MIMO systems. If global CSIT is available, several MIMO transmitters can operate jointly, which is much more efficient than they operate individually. However, global CSIT assumption requires backhaul transmission over the entire network and can be very costly to realize in practice. Also, the delay induced by the backhaul links can introduce CSIT error and degrade the system performance. In this thesis, the basic assumption is that each transmitter in a MIMO network knows CSIT of the channel it sees and does not know the CSIT (except some limited statistics) of the channels seen by other transmitters. This distributive assumption reduces the burden on the backhaul links, which is important in practical realization. However, the challenges are how to design the cooperative transmission scheme based on distributive CSIT and how to quantify the performance loss incurred by the distributive CSIT assumption compared with global CSIT. These two challenges are the motivation of this thesis. Three commonly considered MIMO scenarios are investigated in this thesis, and coherent effect, which coherently superimposes the signals from multiple distributed transmitters at the receiver, is shown to be essential for distributive transmission design. The first scenario is a distributed MIMO system in which multiple transmitters located at different places cooperatively send a common message to a single receiver. A linear Hermitian precoding technique is proposed to enhance the system performance. This scheme transforms the equivalent channel, including a physical channel and a precoder, into a Hermitian matrix form. The proposed scheme can provide coherent effect, and its optimality is also analyzed. Numerical results demonstrate that the proposed Hermitian precoding scheme can perform very close to the channel capacity with global CSIT in various settings, while only distributive CSIT is required. The next scenario investigated in this thesis is a parallel relaying system, in which a source node communicates with a destination node assisted by multiple parallel relays. A new amplify-and-forward (AF) relaying strategy, referred to as product Hermitian precoding (PHP), is proposed based on distributive CSIT. In this scheme, the equivalent channel matrix for each relay link, formed by the two-hop channels and the relay precoder, is transformed into the product of two Hermitian matrices. Product Hermitian precoding can be seen as an extension of Hermitian precoding, and can also provide the coherent effect in a relay network. Numerical results demonstrate that the product Hermitian precoding scheme significantly outperforms the existing AF schemes, and performs close to a performance upper bound with global CSIT. This loss converges to zero in some asymptotic cases. The final scenario is a multi-user cellular system, in which downlink cooperation transmission is discussed. Different from traditional cluster-centric cooperation schemes, a user-centric cooperation scheme is proposed in which each user selects several base-stations (BSs) with the best large-scale fading factors. The data of users is shared among the cooperating BSs, and only distributive CSIT is required at each BS, which greatly reduces the amount of backhaul communications. A precoding strategy deriving from the basic principle of Hermitian precoding and block diagonalization is proposed as the multi-user downlink transmission strategy at each BS. Numerical results demonstrate that, compared to the conventional cluster-centric cooperation schemes, the proposed user-centric cooperation scheme can provide higher sum rate, better user fairness, and meanwhile requires less CSIT. In summary, several novel cooperative MIMO transmission strategies based on distributive CSIT are developed in this thesis. These strategies perform close to the upper bound based on global CSIT and can be proved to be optimal under some system settings, which is theoretically interesting. It is revealed that very impressive performance gain is still available by cooperative transmission under distributive CSIT assumption. Furthermore, the distributive nature of these schemes significantly relives the burden on the backhaul, and thus is practically attractive.

Date of Award	3 Oct 2014
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Ping LI (Supervisor)