Intelligent Control of a Quadrotor with Suspended Load

• Author / Creator

  Mohammadhasani, Arash

• This thesis targets output tracking problem for payload position and quadrotor yaw in an slung load system (SLS). In spite of its relatively extensive literature, full SLS
  control is still a challenging problem since its dimension, nonlinearity, and multiple sources of disturbances are not easy to handle using available nonlinear control tools. Based on the inherent property of SLS having a flat output, in this thesis, we
  tried to solve the robust output tracking using tools from differential geometry and reinforcement learning (RL).
  We first construct a control-affine model for SLS which is novel in that it is written in pendulum coordinates and considers disturbances in pendulum dynamics.
  Our payload is modelled as a single pendulum attached to the center of mass (CoM) of the unmanned aerial vehicle (UAV) with a spherical joint. Based on this model, we use dynamic extension algorithm (DEA) to reformulate SLS dynamics so that
  the derived system is feedback linearizable and yet complies with our original output tracking problem. Static state feedback linearization of the augmented system then
  provides linear time-invariant (LTI) tracking error dynamics in the linearizing state coordinates. These dynamics are exponentially stable on a well-defined and practical
  region of state space for constant disturbances. Moreover, we provide a Maple symbolic script which can be used to apply DEA to general nonlinear control-affine systems.
  Finally, we employ ideas from RL to further robustify our proposed DEA-based control scheme against disturbances and to fill its weaknesses in model dependency.
  We cover all required definitions and algorithms and highlight what distinguishes the proposed approach from conventional RL-based control. Closed-loop performance is validated in simulation and compared with a state-of-the-art method from the literature.

• Subjects / Keywords

  • Intelligent Control
  • Quadrotor control
  • Reinforcement Learning

• Graduation date

  Fall 2022

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-gn05-n579

• License

  This thesis is made available by the University of Alberta Library 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.