Model predictive quadrotor indoor position control

This article addresses the control problem of quadrotors in environments where absolute-localization data (GPS, positioning from external cameras) is inadequate. Based on an attached IMU and an optical flow sensor the quadrotor's translational velocity is estimated using an Extended Kalman Filter. Subject to the velocity measurements, the roll, pitch and yaw (RPY) angles, the angular rates and the translational accelerations a switching Model Predictive Controller is designed. The quadrotor dynamics is linearized at various operating points according to the angular rates and the RP angles. The switching is inferred according to the various linearized models of the quadrotor. The controller is applied on a quadrotor prototype in low-altitude position hold maneuvers at very constrained environments. The experimental results indicate the overall system's efficiency in position/altitude set point maneuvers.