A belief function-based forecasting link breakage indicator for VANETs

In Vehicular Ad-Hoc Networks, a link failure may occur due to non-optimal channel conditions, congestion or node mobility which causes data loss. Common proposed approaches try to overcome this problem by anticipating link disruptions with MAC layer indicators. Such methods, particularly in urban environments (i.e. highly dynamic) are ineffective. Our aim is to setup an indicator that detects at the PHY level an upcoming link breakage before it causes packet loss at the NET layer. To this end, we use Orthogonal Frequency-Division Multiplexing decoding events that are combined thanks to the Dempster–Shafer Theory (DST). The proposed indicator, called Link Breakage Forecasting Indicator performs for a given link, an analysis based on decoding error density measurements, in order to maintain the link history. The adaptation of the DST to the analyzed phenomena relies on using mass functions controlled by the reception power, the relative speed and the error density. The link failure probability is obtained thanks to the fusion of these heterogeneous information using the cautious combination rule. The later allows to consider data even if it is dependent without providing biased results. Simulation results show that detection times are suitable and robust against mobility related characteristics, such as vehicle speeds and urban environment variability.