UAV state and parameter estimation in wind using calibration trajectories optimized for observability

This letter develops a framework for online state, parameter, and wind estimation for a UAV equipped with an Inertial Measurement Unit (IMU) and a ground-velocity sensor, such as visual- or lidar-based odometry. Thrust and moment control inputs are used to steer the process model and the ground-velocity and IMU measurements are assimilated by a square-root unscented Kalman filter containing 23 states, 12 of which are constant parameters. Additionally, we characterize the system's observability and design optimal calibration trajectories to maximize observability of the model parameters. Simulations show the improvements obtained in tracking performance by coupling the estimation framework with a standard model-based controller. By estimating unsteady winds, the onboard controller's gust rejection improves.