Sampling UAV Most Informative Diagnostic Signals

Detecting and diagnosing failures of Unmanned Aerial Vehicles during their mission is a key challenge for their effective deployment. On-board diagnostic systems are able to provide a huge amount of information about the state of the vehicle during the flight, by monitoring sensors, software, and hardware components. However, the ability of processing such data in an online manner is a serious obstacle to a timely detection and diagnosis of failures. This paper proposes a method to progressively focus the data collection on signals providing the most reliable information about the system failure probability, so as to reduce considerably the number of false alarms and/or undetected failures, and to ease the online data processing. We set a simulation experiment showing that the proposed approach is able to select the most informative subset of signals in few iterations in an effective and efficient way.