IMU-Based Inertia Estimation for a Quadrotor Using Newton-Euler Dynamics

In this letter, we demonstrate that a quadrotor's tilt, angular velocity, linear velocity and the parameters shown in Table II may be estimated using only an inertial measurement unit (IMU) and motor speed feedback for sensing. Motor speed commands are used to drive the process model and the motor speed and IMU measurements are used in the measurement model of an unscented Kalman filter (UKF) containing 32 states, 14 of which are constant parameters. We analytically show the observability of this system. Furthermore, we demonstrate through experiments that a blade flapping moment term is not only significant, but necessary to include in the rotation dynamics to get a sensible moment of inertia estimate. We also model the motor torque as a function of the angular acceleration and velocity of the motors in order to obtain a more accurate moment of inertia estimate.