In this dissertation, we considered the exceedingly general spherically invariant random process (SIRP) as a unifying framework for studying the wireless communication fading channels.
In addition, we studied the important issue of modeling uncertainty, where only limited knowledge of the underlying fading channel statistics is known.
The moment space methodology was proposed to characterize the uncertainty range of the system performance.
Since performance evaluation for wireless communication systems frequently involve Monte Carlo simulation, we introduce the Super-Efficient Monte Carlo simulation methodology and the concept of Approximate Super-Efficiency (ASE) to improve the convergence rate of Monte Carlo simulation.
While conventional Monte Carlo simulation yields the convergence rate $1/N$, our Super-Efficient Monte Carlo simulation has a superior convergence rate $1/N^2$ for integrands of the Super-Efficient type, and $1/N^{\alpha}$ for ASE algorithm, where $\alpha \in [1, 2]$.
Finally, we studied the downlink throughput maximization problem in cellular networks.
Inspired by the multi-armed bandit problem, we proposed several algorithms to solve the online throughput maximization problem, and provided convergence analysis.
The proposed algorithms achieved up to 99\% of the performance upper bound within 1000 time steps.