Model predictive control with real-time iteration for quadcopter motors on an embedded platform

Master's Thesis

Michael Floßmann

Monday, July 06, 2020, 10:00

Venue TBD

The effects of climate change in recent years has shifted social and political attitudes in favor of renewable energy. This has increased research and development efforts in airborne wind energy (AWE), which uses unmanned aerial vehicles, like quadcopters for power generation in high altitudes. To increase possible payload, flight-duration and range of these unmanned aerial vehicless (UAVs), more efficient control algorithms for their components are highly desired. For UAVs, rotor control is conventionally handled by PID-controllers, while increased computational power and efficiency of small-scale microcontrollers could enable more sophisticated approaches, like model predictive control (MPC).

This thesis proposes real-time NMPC with a real-time iteration (RTI) as an approach for controlling UAV rotors on an embedded platform. It covers the development of a real-time NMPC setup with RTI on embedded hardware, including modeling, parameter identification and implementation of a real-time embedded closed-loop controller.

Three different models of a UAV rotor with sinusoidal commutated BLDC motor for NMPC were derived and their parameters identified. A real-time closed-loop NMPC setup using a RTI scheme controlling the rotor was developed. The experiments in simulation and real-world setup show that each controller instance was able to control the rotor to a satisfying degree. The performances of the controllers were compared to each other and to an established motor controller for the same rotor setup. The developed controllers and models showed that UAV rotor control with NMPC is feasible on an embedded platform and lays the groundwork for future developments in this field.