The Quadrotor Position Control Based on MPC with Adaptation

Abstract

In recent years, quadrotor drones have gained more and more attraction both in industry and research. Position control is critical to quadcopter flight control. Its performance is highly related to collision avoidance, and therefore safety critical. Among various control approaches, model predictive control gained attention by its systematic way of addressing optimal performance objective and system constraints in both states and inputs. However, the model predictive control (MPC) approach is inherently model based. It cannot guarantee good performance of the system under the condition of inaccurate or purturbed model of the plant. In this paper, we design position controller based on model predictive control for the quadrotor with adaptation. Model reference adaptive control is used to recover the nominal model used by MPC. To further improve the performance, the state dependent position dynamics are captured by Linear Parameter Varying (LPV) model, and comparison the performance to LQR controller using single point model shows the advantage of this approach. The simulation result shows that the MPC controller is better than LQR when the model of quadrotor is nonlinear with state dependency. When there are uncertainties in the model, the MPC controller with adaptation has a better performance than MPC controller.