In the past few years, research in Wireless Sensor Networks (WSN) has grown at an unprecented rate. This is due to the large number of potential applications and environments WSNs can be used in. Nodes in WSNs communicate in multihop fashion to deliver the sensory information to a central processing unit, such as a base station or a sink node. This form of communication requires a degree of network connectivity which might not be always achievable, either due to the sensor deployment strategy, or due to sensor node failure, which can be malicious, or otherwise. In this thesis, we study the problem of data delivery in disconnected WSNs. A special class of disconnected sensor networks called "Fragmented wireless sensor networks (FWSN)" is considered. A FWSN consists of several groups of connected sensor nodes that we call "fragments". We propose a mobility based approach that exploits resource rich, in terms of power and buffer size, mobile agents that move in the network and operate as data relays between fragments to eventually deliver data to the base station. The movement of the mobile nodes and their role as relay stations is modeled using a closed queueing network approach, which is used to obtain steady state results. Building on these results, we derive the distributions of the fragment-to-fragment and fragment-to-sink delays. The results show that this model accurately captures the system behavior. Using the same model, the effect of the movement policy, the number and speed of mobile relays, and the service time at each fragment on the end-to-end delay has also been studied. The proposed queueing model can also be used to model other roles of the mobile nodes, including their roles as either data collectors or data sinks. We also study some practical issues, including mobility control in large networks and engineering the service time, i.e., the time that an MR spend in relaying data between fragments.