Loading...

Navigation and Control of a Free-Flying Satellite for Cargo Transportation and Placement in Intra-Vehicular Environment

Moosavi, Farzan | 2022

163 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55250 (45)
  4. University: Sharif University of Technology
  5. Department: Aerospace Engineering
  6. Advisor(s): Kiani, Maryam
  7. Abstract:
  8. The main topic of this research is the navigation and control of a free-flying satellite with a robot manipulator while transferring and placing cargo on a space station. After connecting the cargo module, astronauts and robots need to deliver the available loads to the desired modules in a space station. Capturing an object or cargo is a critical issue studied extensively so far. However, delivery and placing of the cargo at the target point have not been investigated yet, to the best of the author’s knowledge; hence, this is a primary motivation for conducting the proposed research. After capturing the payload, the position and attitude of the manipulator with respect to the station’s internal components must be determined by the satellite navigation system and avoid collisions with obstacles, including dynamic and static.The robot’s attitude and position determination can be carried out by the Inertial Measurement Unit (IMU) and optical sensors. The robot manipulators consist of one arm to transfer and place cargo. Keeping the robot’s stability and the payload is another challenge that must be accomplished by the propulsion module placed in the base part of the satellite, and torque exerted by the robot’s joints. In addition, trajectory planning is required to define the optimal trajectory of the manipulator during the rendezvous phase. While approaching the target, the attitude and position of the robot are controlled to avoid colliding with obstacles. As it implies, this operation requires precise states of the free-flying system; therefore, state estimation of robot manipulator during transportation based on neural network is proposed with taking advantage of IMU as the main sensor.For the comparison and evaluation of the proposed method, an Unscented Kalman Filter is used, and the estimation robustness is investigated by mass margin of the whole system. Furthermore, the system is controlled via Deep Deterministic Policy Gradient, which is based on a data-driven method and requires no model for control. A proportional-Integrative-Derivative (PID) controller, whose gains are optimized by genetic algorithm, then is used as a reference controller to examine the performance of the Reinforcement Learning (RL) method. Finally, the same analogy is utilized for the placing phase of the mission.
  9. Keywords:
  10. Robot Manipulator ; Reinforcement Learning ; State Estimation ; Deep Deterministic Policy Gradient ; Satellite Attitude Control ; Free-Flying Satellite ; Space Cargo Transportation and Placing

 Digital Object List