The focus of this dissertation is modeling, analysis, and performance evaluation of wireless ad hoc networks. The first part consists of developing analytical models to compute the end-to-end delay in unsaturated IEEE 802.11-based multi-hop wireless networks, using two approaches: queueing theory and diffusion approximation. Using Queueing theory, we develop a Markov chain model to derive closed form expression for the end-to-end delay over a single-hop network by modeling each node as a tandem two-server queuing system taking into consideration an extensive list of key factors that have an impact on the performance of the delay, such as network layer processing time, network/MAC-layer queuing delay, retransmission delays, the IEEE 802.11 MAC-layer backoff, collision avoidance mechanisms, the duration of time the backoff timer is frozen, offered load, and network density. Using diffusion approximation, we developed our second analytical model for end-to-end delay over a single-hop. We formulated the end-to-end delay over a multi-hop path by computing the probability distribution of the total number of packets over a multi-hop path as the convolution of probability distributions of the number of packets over a single hop. The end-to-end delay average is computed as the expected value of the delay distribution over a multi-hop path.

The second part of this work consists of using statistical design of experiment to characterize the relationship between extensively used performance metrics, namely, end-to-end delay, packet delivery ratio, and jitter, and relevant independent factors, namely, the average speed, the offered load, the network size, and the routing protocol. The empirical models can be used to predict values of the mentioned performance metrics for any factor values within the analyzed design space, as opposed to running a large number of time and resource intensive simulations.

The third part of this work consists of developing a performance index that summarizes information available in multiple response metrics into a single scalar value. The performance index allows us to provide a reliable comparative study of the existing protocols. Using an analysis of variance, we derive an empirical model that describes the functional relationship between the proposed performance index and four independent influential factors.