Mobile-Projected Trajectory Algorithm With Velocity-Change Detection for Predicting Residual Link Lifetime in MANET

Abstract

We study the estimation of residual link lifetime (RLL) in mobile ad hoc networks (MANETs) using the distances between the link's nodes. We first prove that to compute uniquely the RLL, at least four distance measurements are required. We also demonstrate that random measurement errors are the dominant factor in prediction inaccuracy and that systematic errors are negligible. We then propose a mobile-projected trajectory (MPT) algorithm, which estimates the relative trajectory between two nodes from periodical measurements of the distances between them. Using the relative trajectory, the algorithm estimates the RLL of the link between the two nodes. For comparison purposes, we derive a theoretical upper bound on the achievable prediction inaccuracy by any distance-based RLL prediction algorithm with unknown but finitely bounded measurement-error distribution. To account for velocity changes, the MPT is enhanced with a velocity-change detection (VCD) test. Performance evaluation demonstrates robustness in RLL prediction for piecewise-linear trajectory and multiple velocity changes during the link lifetime.