Cascaded Nonlinear MPC for Realtime Quadrotor Position Tracking ?

摘要

In this paper we present a cascaded control approach using nonlinear model-predictive controllers for both stages. Using a quadrotor platform as an exemplary target plant with fast nonlinear dynamics, a realtime capable design is proposed that does not require the plant dynamics to exhibit clearly separable time constants as in classical cascaded control. In contrast to similar work we employ NMPC for the inner control loop instead of classical control approaches such as PID control, allowing to keep predictive properties as well as explicit constraint handling. We demonstrate via Monte-Carlo simulations that our design is able to achieve a significantly better position tracking performance of the quadrotor while being equally computationally expensive compared to monolithic position tracking NMPC.