Relay aided interference alignment in large interference limited networks

Relay aided interference alignment in large interference limited networks

In recent years, the number of active wireless devices increases exponentially and it is, therefore, to expect that the interference increases as well. Interference between communication links is the major performance limiting factor in today’s communication networks. Hence, the handling of the overall interference in a network is one major challenge in wireless communication networks of the future. If the interference signals are weak in comparison to the useful signal, they can be simply treated as noise. If the interference signals are strong in comparison to the useful signal, they can be reliably decoded and subtracted from the received signal at the receivers. However, in multiuser communication networks, the interference and the useful signal are often of comparable signal strength. The conventional approach to handle these interference signals is to orthogonalize the useful signal and the interference signals using, e.g., time division multiple access (TDMA) or frequency division multiple access (FDMA). In the past few years, instead of orthogonalization, interference alignment (IA) has been developed as an efficient technique to handle interference signals, especially in the high signal to noise ratio (SNR) region. The basic idea of IA is to align multiple interference signals in a particular subspace of reduced dimension at each receiver. The objective is to minimize the signal dimensions occupied by interference at each receiver. In order to perform IA, the receive space is divided into two disjoint subspaces, the useful signal subspace and the interference signal subspace. Each transmitting node designs its transmit filters in such a way that at each receiving node, all interference signals are within the interference subspace and only the useful signal is in the useful subspace.

In this project, the focus is on large interference limited wireless communication networks. In contrast to the conventional use of relays, for extending the coverage, in this project, the relays are used to manipulate the effective end-to-end channel between the transmitters and receivers to perform IA in the network. Since the relays are used to assist the process of IA and not interested in the data streams transmitted by the nodes, amplify and forward relays are sufficient to support the process of IA. Therefore, the main focus of this project is on amplify-and-forward relays. Throughout this project, it is assumed that all nodes and relays are multi-antenna half-duplex devices. When considering large networks, the assumption that all nodes are connected to all relays does not hold due to physical propagation phenomena, e.g., high path loss and shadowing. In such large networks, the distances between different nodes may differ a lot, leading to links of considerably different signal strengths, where sufficiently weak links may be neglected. Hence, large networks are in general partially connected. In this project, three important interference-limited relay aided wireless network topologies are investigated, the partially connected relay aided multi-pair pair-wise communication network, the fully connected multi-group multi-way relaying network and the partially connected multi-group multi-way relaying network. For each of these topologies, new algorithms to perform IA are developed in this project.

First, a large partially connected relay aided pair-wise communication network is considered. The concept of an appropriate partitioning of a partially connected network into subnetworks which are themselves fully connected is introduced. Each of these subnetworks contains a single relay and all nodes being connected to this relay. Some nodes or even communication pairs may be connected to multiple relays. The bidirectional pairwise communication between the nodes takes place via the intermediate relays, using the two-way relaying protocol. Only relays which are connected to both nodes of a communication pair can serve this pair. Hence, it is assumed that all communication pairs in the entire network are served by at least one relay. The most challenging part of such a partially connected network is the handling of nodes which are connected to multiple relays. Hence, techniques called simultaneous signal alignment (SSA) and simultaneous channel alignment (SCA), are proposed to perform signal alignment (SA) and channel alignment (CA) with multiple relays simultaneously. SA means that all nodes transmit to the relay in such a way that the signals of each communicating pair are pair-wise aligned at the relay. For CA, which is dual to SA, the receive filter of each node is designed such that the effective channels between the relay and both nodes of a communicating pair span the same subspace. A closed-form solution to perform IA in this network topology is obtained and the properness conditions for SSA and SCA are derived. It is shown that local channel state information (CSI) is sufficient to perform IA in partially connected networks, whereas in fully connected relay aided networks, global CSI is required in general. Through simulations, it is shown that the proposed closed-form solution achieves more degrees of freedom (DoF) than the reference algorithms and has better sum-rate performance, especially in the high SNR-region. Especially in large wireless networks, it may happen that not both nodes of a communication pair are connected to the same relays. If a single node of a communication pair is in addition connected to a relay which, therefore, cannot assist the communication, this node receives only interference and no useful signal from this relay. Such a node suffers from inter-subnetwork interference, due to the connection by an inter-subnetwork link to the additional relay. Hence, in this project, a closed form algorithm which minimizes the inter-subnetwork interference power in the whole partially connected network is proposed and the properness conditions are derived. The condition under which an interference free-communication can be achieved by the proposed inter-subnetwork interference power minimization algorithm is derived. Further, it is shown that the proposed inter-subnetwork interference power minimization algorithm achieves a higher sum rate in comparison to the considered reference algorithm.

Secondly, a fully connected multi-group multi-way relaying networks is considered. In such a network, multiple nodes form a group and each node wants to share its message with all other nodes in its group via an intermediate relay. The group-wise communication between the nodes inside a group takes place via the intermediate relay, using a transmission strategy considering several multiple access (MAC) phases and several multicast (MC) phases, in general. In this project, a multicast IA algorithm to handle the interference in such a network is proposed. The idea of the proposed algorithm is that in each of the MC phases, a multiple input multiple output (MIMO) interference multicast channel is created by separating the antennas of the relay into as many clusters as groups in the network. Each of these clusters serves a specific group of nodes and transmits in such a way that the signals transmitted from different clusters are aligned at the receiving nodes of the non-intended multicast groups. It is shown that the minimum required number of antennas at the relay is independent of the number of nodes per group, which is an important property since the number of antennas available at the relay is limited in general. Furthermore, the properness conditions for the proposed multicast IA algorithm are derived. It is shown that the proposed multicast algorithm outperforms a reference algorithm for a broad range of SNR values, while still requiring less antennas at the relay.

Finally, a large partially connected multi-group multi-way relay network is considered. In contrast to the fully connected multi-group multi-way relaying network, multiple relays are considered in this partially connected network. Such a partially connected network can be partitioned into subnetworks that are themselves fully connected. Hence, such a partially connected network consists of multiple subnetworks, where each of these contains a single relay and all groups of nodes which are connected to this relay. Each group of nodes may be connected to one or multiple relays. This means that not all groups of nodes are connected to all relays in the network. However, any group is connected to at least one relay which serves this group of nodes. The group-wise exchange of data between the nodes inside a group is performed via the multi-way relaying protocol. The most challenging part of such a partially connected network is the handling of the nodes inside groups which are connected to multiple relays. To overcome this challenge, new techniques called simultaneous group signal alignment (SGSA) and simultaneous group channel alignment (SGCA) are introduced to perform SA and CA in partially connected multi-group multi-way relaying networks. A closed-form IA solution for this network topology is obtained and the properness conditions for the solvability of SGSA and SGCA are derived. It is shown that the proposed IA algorithm outperforms the reference algorithm in terms of sum rate and DoF.