Cross-layer design for MIMO ad hoc networks

This dissertation investigates the application of MIMO techniques to ad hoc networks. We first develop an analytical framework for performance evaluation of multi-hop ad hoc networks. The framework is capable of characterizing ad hoc networks at both the media access control (MAC) and physical layers. The performance evaluation can be applied to both omni-directional and directional networks under different signaling schemes. The framework is then applied to optimize the code rates for an ad hoc network under fast Rayleigh fading assumption. A performance criterion (the normalized expected forward rate) is also developed as a figure of merit for the design of ad hoc networks. Simulation results show that the proposed adaptive code rate scheme not only provides higher expected forward rates for MAC layer traffic but also provides higher end to end throughput.; For the physical layer, we generalize the concept of the Total Interference Function (TIF) in beamforming to full eigencoding. Based on the TIF function, we develop the Noncooperative Generalized Eigencoding with Taxation (NGET) for flat fading MIMO ad hoc networks. Simulations show that NGET provide higher power efficiencies than the greedy counterpart. A game theoretic interpretation of NGET and the proof of the existence of fixed points of NGET are also provided.; NGET is then extended to frequency selective fading channels. The algorithm is extended in two directions: the space-time eigencoding algorithm and the OFDM-constrained eigencoding. The iterative MMSE algorithm based on the idea of NGET is also developed. The algorithm can be implemented using training sequences and algorithms such as LMS or RLS without the requirement of explicit knowledge of the channel and interference covariance matrices.; The physical channel models of NGET are then used to develop the Directional Multiple Access with Collision Avoidance (DMACA) protocol. This is a MAC protocol for MIMO-OFDM ad hoc networks. With the assumption of perfect collision channel models, it is shown that by using appropriate parameters it is possible to achieve higher throughput using this protocol than a fully directional protocol or a protocol using omni-directional RTS-CTS. An optimization procedure to find the optimal number of control carriers and the optimal transmission rates based on the analytical framework developed earlier is also developed.