The performance of routing protocols in wireless ad-hoc networks is determined by a number of factors, among which the path durations are of much importance. Path durations affect the reliability of the network service provided to the applications and the routing overhead incurred. In this dissertation, we study the distribution of path duration in wireless ad-hoc networks and its impact on routing. We focus on identifying a scheme that selects a path with the largest expected duration for data transmission when multiple paths are available between a source-destination pair. To this end, we first study the distribution of path duration. Our analytical result also reveals the relation between link level and path level statistics, which can be used to estimate expected path durations.

Our main results show that in a large scale wireless ad-hoc network, as long as the local dependency among link excess lives is not too strong, the path durations can be well approximated by an exponential random variable for paths with sufficiently large hop count. Furthermore, the inverse of the expected duration of a path can be estimated using the sum of the inverses of the expected durations of the links along the path.

Based on these analytical results, we propose a new path selection scheme referred to as ``Maximum Expected Duration" (MED) path selection. This scheme can be easily incorporated into existing routing protocols. Information needed for stimating the expected path duration can be collected during path discovery phase. Our simulation results demonstrate that under routing protocols with the MED path selection scheme added, the median value of the path durations can be increased up to $60$ percent compared to those without using the scheme. Moreover, we reduce the delay and overhead during local path recovery by using cached paths that are likely to be available when the primary path breaks down.