Link Classification and Residual Time Estimation Through Adaptive Modeling for VANETs

Vehicular Ad hoc Networks design has drawn a lot of research attention. High node mobility, an inherent characteristic of VANETs, results in short lived links and rapid topology changes; this makes traditional ad-hoc protocols inefficient for vehicle to vehicle communication. In this paper we propose a cross-layer approach, where physical layer information can be used by upper layers to monitor the well-being of the various links used, and to estimate their residual time. The algorithm proposed comprises of forming a time series based on physical layer measurements. Utilizing adaptive non-linear parameter estimation methods, this time series can be used to estimate the current state of the link and its residual time. The environment in which VANETs usually operate together with the effect of relative node movement introduce considerable noise. For this reason, a data-driven signal processing technique, Empirical Mode Decomposition is used for denoising. The proposed algorithms are tested against real data and simulations.