Soft Information Based Protocols in Network Coded Relay Networks

Future wireless networks aim at providing higher quality of service (QoS) to mobile users. However, phenomena such as multipath fading and limited resources, namely bandwidth, still constrain almost every design. The emergence of relay technologies has shed light on new methodologies through which the system capacity can be dramatically increased with low deployment cost. In this thesis, novel relay technologies have been proposed in two practical scenarios: wireless sensor networks (WSN) and cellular networks. The research includes theoretical analysis and simulations performed to show the advantages of the proposed protocols over existing relay protocols. In practical WSN designs, energy conservation is the single most important requirement; being powered by batteries and deployed in distributed areas, designs aim at extending the life span of the sensor network by reducing its energy consumption. Often, replacing or recharging batteries is too difficult or even impossible; thus extending the life span of the network would reduce maintenance costs. The introduction of relay nodes to the WSN has much relieved the burden of each source node. This is because the relay nodes are usually more powerful and being deployed in the middle of the remote transmissions, which can assist to forward messages from the source nodes. At the same time, because adjacent source nodes are observing the same phenomenon, there should be a lot of redundancy within the data they have gathered. Relay nodes can thus eliminate the redundancy through processing the received signals together. This thesis draws attention to a multiple access relay channels (MARC) model in the WSN, where two correlated source nodes transmit their messages to a common destination via the help of a relay node. Network coding (NC) is performed at the relay node by merging two received signals into one signal using predetermined techniques. In this thesis, the network coded symbol for the received signals from correlated sources has been derived; the network coded symbol vector is then converted into a sparse vector, after