An Efficient Probabilistic Model for Anomaly Prediction in Aerial Ad Hoc Networks

The increasing use of Unmanned Aerial Vehicles (UAVs) in various applications of Aerial Ad Hoc Networks, such as precision agriculture and aerial remote sensing, is fast contributing to their adoption in many civilian applications. Generally equipped with multiple sensors, such as cameras and movement detectors, UAVs are also deployed in hostile environments such as disaster zones or military fields. Therefore, and in order to ensure the success of any deployment of Aerial Ad hoc Networks, detecting and isolating failures is of great importance to allow a high level of security and reliability. In this research, we first introduce a model using a Bayesian network that copes with this type of issues and tries to detect any faulty UAV. Second, we develop a probabilistic predictive scheme to avoid the unexpected failure of a UAV. The proposed approach is validated using realistic synthetic datasets provided by the UAV laboratory at the University of Minnesota.