| In wireless networks, we can improve system performance by exploiting cooperative communications, where neighbor nodes may interact to jointly encode, decode, or relay information. In this thesis, we show that cooperation can improve capacity, but only when the right cooperation strategy is chosen based on the network topology, signal-to-noise ratio (SNR), channel state information (CSI), and power allocation assumptions. In particular, when we consider the deployment of relay nodes in a wireless network, CSI at the transmitter is necessary for effective transmitter cooperation, while transmission power allocation between the cooperating nodes is crucial for receiver cooperation. The benefits of cooperation are further studied when the relay and receiver can iterate to exchange information over orthogonal, finite-capacity channels known as conference links. We show that a two-round iterative decoding scheme can achieve a higher capacity over non-iterative cooperation.; This thesis also investigates cross-layer resource allocation where network resources such as power and bandwidth are optimized jointly based on conditions of the wireless channels, user requirements, and the interaction among the network layers. We consider the joint compression and transmission of a source over a slowly fading channel with decoding delay constraints, when CSI is not available at the transmitter. The source is described by a low-fidelity base layer, combined with high-fidelity refinement layers that are only decodable when the channel fading realization is favorable. We allocate the transmission power optimally among the layers to minimize the expected distortion. Under Rayleigh fading the optimal power allocation is conservative: most of the power is allocated toward the base layer. Cross-layer resource allocation is combined with cooperative communications when we consider the scenario where, in addition to an encoder, a helper node transmits correlated side information to a decoder through a fading analog channel. The decoder knows the fading realization but the encoder knows only the fading distribution. In this case the expected distortion is minimized by optimally allocating the source coding rate among the possible fading states, which we show can be formulated as a convex optimization problem. |
