Embedded Real-Time Nonlinear Model Predictive Flight Control for Energy-Efficient UAS

Model Predictive Control (MPC) is a powerful control framework that recurrently provides a control signal to optimally control a system. However, in the application of an Unmanned Aerial System (UAS) with a multirotor configuration, MPC is often used more as a high-level guidance system than as a low-level flight controller due to the extensive computation required to solve for an optimal solution given the UAS's relatively fast dynamics. As a result, the eventual performance of the UAS is degraded from the optimal measure predicted by the MPC. This paper addresses this challenge by formulating the MPC problem for both position and attitude controls, specifically for UAS with a multi-rotor configuration, and developing an optimization solver tailored to the problem, enabling the framework to be executed in real-time on a commercial-off-the-shelf embedded hardware platform. Experimental flight tests have been conducted, and the resulting performance is compared with that of the baseline flight control in terms of energy consumption.