In this dissertation, various techniques utilizing network diversity are introduced to analyze the effect of the network diversity on the performance enhancement of heterogeneous networks (HetNets).
First, a vertical handover scheme is proposed to minimize total power consumption required in serving a traffic flow without affecting the constant service rate of each network by using dynamic power control. The proposed scheme uses a Markov decision process (MDP) to capture the system's stochastic behaviors as well as the power consumption during a handover execution.
Second, a traffic offloading method is proposed to minimize the energy consumption of 2-tier HetNets that support 2 classes of packets with different packet delay requirements. Under given base station (BS) densities, the optimal BS association biases that minimize the delay of both packet classes is analyzed. Then, by means of the analysis, the optimal BS densities that minimize network energy consumption are found, while satisfying the delay requirement of both packet classes.
Third, slot-by-slot joint resource allocation algorithms are proposed for uplink multi-homing networks, in which the objective is to maximize uplink network utility. According to difference sub-channel allocation assumptions (i.e., fractional and integer sub-channel allocation), different joint resource allocation algorithms are investigated. Under the assumption of fractional sub-channel allocation, an optimization problem for joint resource allocation is formulated as a convex optimization problem. To achieve an optimal solution of the problem, one adopts Lagrange duality theory and derive an iterative algorithm to achieve the optimal solution. In contrast, under the assumption of integer sub-channel allocation, an optimization problem for joint resource allocation is formulated as a mixed-integer nonlinear programming (MINLP) problem. Thus, a sub-optimal joint resource allocation algorithm is investigated, where the optimization problem is decoupled into three separate sub-problems: initial power allocation, sub-channel allocation for given initial power allocation, and power allocation improvement for given sub-channel allocation.
Fourth, a mathematical model is provided to evaluate the coexistence of Wi-Fi and Listen before talk (LBT)-enabled cellular networks sharing the unlicensed spectrum. Based on the proposed model, the throughput difference between the Wi-Fi access points (APs) and cellular BSs that results from the different medium access mechanisms of the two networks is investigated. In addition, because the two networks must coexist in a friendly manner, their \emph{graceful coexistence} is defined as the condition in which the performance of an individual node under a network scenario with $m$ Wi-Fi APs and $n$ cellular BSs is not worse than that under a network scenario with only $m+n$ Wi-Fi APs. Using this definition, we first examine whether this graceful coexistence is feasible through adjusting the LBT parameter of the cellular BSs, particularly the contention window (CW) size. Then, the optimal CW size that maximizes the total throughput of the two networks under the graceful coexistence condition is obtained for cases with various physical data rates and numbers of nodes in each network.
Finally, a mathematical framework is proposed based on queuing theory that models the time-domain behaviors of a Wi-Fi AP with the carrier sense multiple access with collision avoidance (CSMA/CA) mechanism and a cellular small BS without the regulatory requirement for LBT in the unlicensed spectrum. Based on the proposed framework, the mean packet delay of Wi-Fi AP and cellular small BS investigated as a coexistence performance metric, according to changes in their packet arrival rates. With help of our analysis, the maximum allowable packet arrival rates of Wi-Fi AP and cellular small BS are identified, under which the required Wi-Fi performance is satisfied without the coexistence mechanisms at the cellular small BS. This will serve as a guideline for the cellular small BS on when it needs to employ the coexistence mechanisms.