Towards Decentralized Fault Detection in UAV Formations

A decentralized fault detection (DFD) scheme warranting mission integrity for leader-to-follower formations of almost-lighter-than-air vehicles (ALTAVs) is proposed. Such DFD allows compensating for concurrent communication network and component-level faults that may significantly impact mission integrity despite the presence of on-board fault detection, isolation and recovery for actuator and sensor faults. The formation of unmanned ALTAVs is beforehand stabilized by a distributed formation guidance scheme that uses neighboring vehicle information. Each ALTAV is equipped with an observer based on simplified models of the neighboring ALTAVs. The observer and the guidance law use the same local information. Residues associated with neighboring vehicle states carry the information as to their normal or abnormal flight conditions, thereby determining whether or not there is a fault in the neighboring vehicles. The observer design is formulated as a mixed H₂/H_infin controller synthesis compensating for exogenous disturbance estimates and accounting for noisy measurements. The minimum detectable fault and the associated threshold are shown to be increasing functions of the relative distance between the formation leader and the follower vehicle at hand, which means that the value of the threshold may vary from one follower to another depending on its relative distance from the leader. Simulations based on high-fidelity nonlinear 6DOF ALTAV models show that the proposed team-level DFD scheme combined with a simple guidance adaptation technique enable quick hard fault/failure detection while preserving leader-to-follower formation geometry requirements.