Decentralized Detection of a Class of Non-Abrupt Faults With Application to Formations of Unmanned Airships

We propose a decentralized non-abrupt fault detection (DNaFD) scheme for leader-to-follower formations of unmanned airships. Non-abrupt faults are those that result in slow performance degradation and in undesirable drift, which can propagate from one vehicle to another, and therefore can adversely affect mission integrity, potentially destabilizing multivehicle formations, while being difficult to detect. As opposed to model-based fault detectors, which are typically insensitive to non-abrupt faults, the proposed signal-based DNaFD enables the detection of slowly degrading vehicle performance by performing a statistical test on heading angle trajectories. Here, the formation of unmanned airships is assumed stabilized by a distributed formation guidance scheme that uses neighboring vehicle information. High-fidelity, nonlinear 6-degrees-of-freedom (DOF) simulations of formation flying airships show that the proposed DNaFD scheme combined with a simple guidance adaptation technique enable detection of a class of non-abrupt faults and formation recovery, despite mild winds and parametric uncertainties, while preserving a requirement on formation geometry.