Interference Resilient Techniques in Wireless Communication Systems

The common thread of the research in this dissertation is to introduce novel interference resilient techniques in multiuser wireless communication systems where the goal of these techniques is to improve system throughput or efficiency and guarantee reliable communication. More specifically, in the first part, we consider the problem of multiplexing wireless data transmission in two-way communication (TWC) channels. Using traditional wireless packet switching methods bidirectional communication tends to double the interference (or time) needed to complete the session. We propose new interference resilient schemes by combining the so called ``two-way relaying'' scheme with a distributed randomized space-time block coding (RSTBC) strategy. RSTBC is a decentralized cooperative technique that ensures diversity gains through the recruitment of multiple uncoordinated relays, with virtually no signaling overhead to enlist relays. In this problem, RSTBC is applied to two-way relaying wireless networks which, when two terminals want to send a message to each other, can potentially improve the network throughput by allowing them to exchange data over two or three time slots via bidirectional relay communications. Specifically, two decode-and-forward (DF) relaying strategies are considered which take up only two time slots. In the first slot the two sources transmit simultaneously. In the former scheme which we refer to as decode and forward both (DFB) RSTBC, only relays which can reliably decode both source blocks via joint maximum likelihood (ML) decoding cooperate, and do so by modulating the bit-level XOR of the decoded data through a single RSTBC. In the latter scheme called decode and forward any (DFA) RSTBC, the relays cooperate in the second slot also when they can decode only one of the two source data. In this case each source data that is decoded is mapped into an independent RSTBC. If the relay decoded reliably both sources, after cancellation of the strong interference, then it sends the two RSTBCs encoding the symbol vectors from each of the sources. A randomized forwarding scheme is also proposed for three-time-slot relaying, which is also a DFA strategy, although without joint decoding or interference cancellation after the first slot.;In addition to more judicious transmission and relaying policies, a key ingredient of future wireless network will be enhanced spectrum sensing. This is motivated by the wide adoption of Multiple Input Multiple Output (MIMO) Orthogonal Frequency Division Multiplexing (OFDM) systems, which offer significant degrees of freedom in encoding data, but at the same time result in a complex spatio-temporal landscape of interference. We combine recent advances in array processing and compressed channel sensing, to solve the problem of estimating the covariance of an asynchronous network of MIMO-OFDM sources. We specifically focus on spectrum sensing (SS) in cognitive radio (CR) systems, since it is of paramount importance to approach the capacity limits for the Secondary Users (SU), while ensuring the undisturbed transmission of Primary Users (PU). In this problem, we formulate a cognitive radio systems spectrum sensing problem in which SU, with multiple receive antennas, senses a channel shared among multiple asynchronous PUs transmitting MIMO-OFDM signals.;Lastly, we study online learning and stochastic optimization in cognitive radio system at MAC layer. The typical Compressive Sensing problem models an observer that wants to recover a sparse N dimensional vector from K linear projections of the vector. We combine the ideas of opportunistic spectrum sensing with compressive sensing and introduce the Cognitive Compressive Sensing (CCS) problem which models a cognitive receiver that uses Bayesian beliefs on a dynamically changing sparse vector representing measurements from the signals occupying the sub-channels. The CR objective is to optimally choose the K linear projections with the objective of maximizing a reward from the inference of the sparse vector support. We formulate CCS as a Restless Multi-Armed Bandit problem, generalizing the popular Cognitive Spectrum Sensing model, in which the CR can sense K out of the N sub-channels. We derive the myopic sensing policy that leverages on the beliefs to its inferences. We also propose a greedy counterpart of the myopic sensing policy algorithm with considerably less required computations. While in general the optimum policy remains elusive, we provide sufficient conditions in which in the limit for large K and N the greedy policy is optimum.