Nonlinear model predictive control for hexacopter with failed rotors based on quaternions —simulations and hardware experiments—

Abstract

This work applies real-time nonlinear model predictive control (NMPC) to fault-tolerant control problems of an unmanned aerial vehicle (UAV) with failed rotors. In the control problem, a hexacopter with up to three failed rotors out of the six available rotors is considered. The NMPC approach includes a quaternion-based nonlinear model of the hexacopter as well as constraints in the thrusts to consider the inherent nonlinearities of UAVs. The proposed method aims to achieve real-time optimization of the NMPC in the on-board computers without any linearization. We explore all possible scenarios in up to three rotor failures and demonstrate control designs in the NMPC for these scenarios. The simulation results indicate that by using the quaternion model, the position and attitude of a hexacopter can be controlled from a large inclined initial state with a non-zero angular velocity and falling velocity. Moreover, the results reveal that the quaternion model is superior to the Euler angle model in terms of the computation time. We also conduct hardware experiments using an actual hexacopter with a failed rotor to demonstrate the real-time NMPC optimization. The results of the simulations and hardware experiments demonstrate that the NMPC can deal with various operation conditions of a hexacopter in a unified manner, with only minor modifications in the performance index.

Keywords

• nonlinear control
• predictive control
• optimal control
• unmanned aerial vehicle
• fault-tolerant control
• quaternion