Model-Aided navigation for a quadrotor helicopter: A novel navigation system and first experimental results

This paper presents a novel navigation system for aiding a classic inertial navigation system with a Vehicle Dynamics Model in a quadrotor application. The navigation system is based on the previously presented Unified Model technique for optimal fusion of the two prediction models. The key point of the approach in this paper is that the required modeling and implementation effort is minimized by incorporating a translational dynamics model only. Even though no rotational vehicle dynamics are modeled, it is shown that the filter is able to estimate roll and pitch angles and even IMU biases with bounded errors. Indoor and outdoor flight experiments confirm that all navigation states except heading angle are significantly improved compared to the single systems. For the first time, this demonstrates in real flight experiments that model-aided navigation allows estimating a full navigation solution for several minutes even without any aiding sensors.