Reactionary envelope protection for autonomous UAVs

In this paper a new reactionary approach for automatic envelope protection in unmanned aerial vehicles is proposed and evaluated using software-in-the-loop simulations. The approach consists of three important steps: predict envelope violation, prescribe safe-response-profile and track safe-response-profile. An adaptive estimate of limit parameter dynamics is obtained by augmenting a linear model with the output from a single hidden layer neural network. The weights of this neural network are tuned on-line to capture system changes affecting limit parameter dynamics in real time. Limit boundary violations are predicted using estimate of the finite-time future limit parameter response. A safe-response profile is generated by treating the boundary of the limit parameter as an obstacle. Appropriate command or control corrections are computed based upon the adaptive estimate of limit parameter dynamics to track this safe-response profile. The approach is used for load factor limiting within the GTMax simulation architecture and flap angle limiting for the R22 vehicle model within the same simulation architecture. The reactionary envelope protection method is therefore shown to be applicable for both steady-state (for example load factor) and transient (for example rotor blade flapping) response critical limit parameters.