Disturbance Observer-Based Motion Control and Visual-Inertial-Actuator Odometry for UAVs

  • Author / Creator
    Moeini, Amir
  • Motion control of multirotor Unmanned Aerial Vehicles (UAV) is an area of research which continues to generate significant interest in the community. Being able to accurately follow a broad class of trajectories clearly improves the mission capabilities of the vehicle. Model uncertainty and external disturbances are important factors reducing motion control performance. Improving the robustness of motion control can clearly broaden UAV capability. In this thesis we propose a number of trajectory tracking motion controls which are based on a backstepping design method. By incorporating disturbance observers for external force and torque the proposed methods provide exponentially stable tracking error dynamics for the constant disturbance case. For time-varying disturbances tracking error is proven to be ultimately bounded. The stability analysis accounts for the full nonlinear vehicle model which includes translational and rotational dynamics. This avoids having to make common simplifying assumptions typical of designs with inner outer loop structure (e.g., linear approximation of the rotational dynamics) during the closed-loop stability analysis. The proposed design provides a model-based partial compensation of rotor drag. Software-in-the-loop (SITL) simulation and experimental flight testing results are presented. These results show the effectiveness of the proposed method using the commonly used open-source PX4/Pixhawk development framework. The results demonstrate the methods' practical usefulness including their robustness and tracking error performance.

    The developed motion control algorithms require an accurate knowledge of the system's state, in some applications a description of the environment and an estimate of external forces (e.g., aerial manipulation and load transport). Having an algorithm that only depends on onboard sensors can increase autonomy and reliability. There has been an increasing amount of work developing state estimation algorithms using vision and inertial measurements. However, incorporation of the dynamics modelling and actuation data, which are already available on UAVs and can provide more information about the vehicle's motion, are normally ignored or just an approximate model with unrealistic assumption on force modelling is used. In this thesis, we include an accurate dynamical modelling of a multirotor by considering the effect of rotor drag and also a disturbance observer developed with the assumption of constant force disturbance into an existing open source state estimation approach. The effect of rotor drag is proved to be significant in control and state estimation and its consideration can improve both the estimation and control tasks. In addition, the proposed disturbance observer which is reformulated as a residual term can assist the estimator to differentiate between the constant or slowly time-varying component of the external force and the accelerometer bias providing a more accurate force estimate which is a need in many UAVs applications. Furthermore, this structure increases the odometry accuracy. We evaluate the performance of our proposed method by integrating it with an open source Visual-Inertial Odometry (VIO) system and testing it on benchmark datasets. The results show a significant improvement in the estimation accuracy.

  • Subjects / Keywords
  • Graduation date
    Fall 2021
  • Type of Item
  • Degree
    Doctor of Philosophy
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.