Design & Validation of Fault-Tolerant Integrated Navigation Systems for Small UAVs

Integrated navigation systems used in safety-critical flight control system are required to be fault-tolerant. This means that they must be able to quickly detect the onset of system faults. Once detected, they must be able to isolate them or issue a timely alarm so that operators can make effective contingency and recovery plans. After faults have been detected and isolated, the navigation system must be reconfigured for either for continued operation or recovery of the vehicle. For systems used in manned aircraft fault-tolerance is achieved, in part, by employing massive hardware redundancy whereby several replicas of components that are likely to fail are included in the system. This approach to fault-tolerance is untenable in many of the emerging safety-critical application of small Unmanned Aerial Vehicles (SUAVs). The severe cost as well as size, weight and power (SWAP) constraints encountered in SUAVs limits the level of hardware redundancy that can be used. This has led to the idea of analytical redundancy whereby intelligent sensor fusion algorithm designs are used to make up for the lack (or complement a limited amount) of physical redundancy.

In this presentation filtering approaches that can be used to incorporate analytical redundancy into the integrated navigation systems used for SUAV flight control systems are discussed. As a case study, we consider a synthetic air data systems--an integrated navigation system which estimates angle of attack, slide slip and airspeed without using the traditional pitot static system but rather information from inertial sensors, GNSS receivers, and equations of motion of the SUAV. As will be shown, the algorithms used for achieving analytical redundancy can be complex and often rely on non-linear filters. This makes safety-validation of these algorithms very difficult using traditional approaches such as overbounding. Thus, some potentially promising approaches for validation which leverage ideas from Extreme Value Theory (EVT) are discussed.