Analysis and Optimization in Wireless Sensor Network Aided Cognitive Radio Systems

Abstract

To enhance cognitive radio (CR) technology, this thesis focuses on the wireless sensor network aided cognitive radio scenario. The main role of the wireless sensor network (WSN) in practice is to overcome the inherent sensing limitations of single independent nodes and provide better cognition capability to the CR entities. In that context, this thesis addresses some of the problems related to the radio resource allocation (RRA) and interference management (IM). The contributions can be summarized in threefold, as follows:

First, we revisit the widely investigated problem of maximizing the centralized uplink interference-limited sum-rate capacity with respect to a multi-cell underlay cognitive radio network. Until very recently only sub-optimal algorithms were proposed due to the inherent non-convexity of that RRA problem. Yet, the problem at hand has been usually neglected in the large-scale setting (i.e., large number of nodes and channels). To better cope with large systems we first propose an exact mathematical adaptation of a wellknown state-of-the-art algorithm. In addition, we also propose a novel low-complexity heuristic algorithm. When compared to two state-of-the-art algorithms, the proposed algorithm is relatively fast and offers very good sub-optimal performance in large-scale CR scenarios.

Secondly, to advocate the use of WSNs, a low operational cost is of paramount importance. In particular, the WSNs used for joint state estimation can also benefit from significant energy savings by proper RRA. The transmission power for each node is optimized to respond to the estimation accuracy constraint, the quality of the digital communication channel, and the energy available in the energy harvesting batteries. The use of energy harvesting shifts the nature of energy-aware protocols from simply minimizing energy expenditure to optimizing it. The goal is to prolong the life of the backup-batteries each sensor-node is equipped with, by strategically exploiting harvested energy when available. A centralized algorithm, based on a stochastic Lyapunov optimization used with a standard Kalman filter as the sate estimator, is first proposed. This algorithm can achieve arbitrarily close to optimal energy expenditure over time, without requiring a priori knowledge of the channel and harvesting statistics, but at the expense of slower convergence to the specified estimation accuracy. The trade-off is controlled via a tunable parameter. Various well-known energy fairness policies can be accommodated to achieve control over the battery replacement cycle. Moreover, a distributed approach is also proposed that is somewhat less energy efficient yet enables better network scalability.

Finally, one of the main purposes of the better cognition using WSNs is to better sense, model, and control the interference of the opportunistic unlicensed users toward the licensed ones. In contrast to the commonly used infinite area assumption for the interferers we focus on analytical performance in the more realistic finite area case, as well as for a more general spatial configuration. The parametric finite model we propose allows many practical combinations, enjoying wide applicability. We first characterize various statistical properties of the aggregate interference, then investigate the legacy users’ performance through different metrics such as: outage probability, bit error rate, and amount of fading, as well as the diversity order and coding gain.