Autonomous Autorotation of Unmanned Rotorcraft using Nonlinear Model Predictive Control

Safe operations of unmanned rotorcraft hinge on successfully accommodating failures during flight, either via control reconfiguration or by terminating flight early in a controlled manner. This paper focuses on autorotation, a common maneuver used to bring helicopters safely to the ground even in the case of loss of power to the main rotor. A novel nonlinear model predictive controller augmented with a recurrent neural network is presented that is capable of performing an autonomous autorotation. Main advantages of the proposed approach are on-line, real-time trajectory optimization and reduced hardware requirements.