Predicting node proximity in ad-hoc networks: a least overhead adaptive model for selecting stable routes

This paper presents a strategy for quantifying the future proximity of adjacent nodes in an ad-hoc network. The proximity model provides a quantitative metric that reflects the future stability of a given link. Because it is not feasible to maintain precise information in an ad-hoc network, our model is designed to require minimal information and uses an adaptive learning strategy to minimize the cost associated with making a wrong decision under uncertain conditions. After computing the initial baseline link availability assuming random-independent mobility, the model adapts future computations depending on the expected time-to-failure of the link based on the independence assumption, and a parameter that reflects the the environment. The purpose for defining this metric is to enhance the performance of routing algorithms and better facilitate mobility-adaptive dynamic clustering in ad-hoc networks.