Fault-tolerant Control of Formation Flying Satellites Using Machine Learning

Farhang Fallah, Raouf | 2021

142 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54743 (45)
  4. University: Sharif University of Technology
  5. Department: Aerospace Engineering
  6. Advisor(s): Assadian, Nima
  7. Abstract:
  8. In this thesis, a fault-tolerant method for controlling the relative position and attitude between two satellites in a leader and follower formation is proposed. The follower satellite is equipped with twelve thrusters which are installed on the satellite in a particular pattern. These thrusters are assumed to be afflicted by faults. The satellites are subject to external disturbances – such as the ellipsoidal gravity of Earth (J2), drag force, solar radiation pressure, and the third body, and a controller is designed to attain the desired formation under these disturbances.For this purpose, six separated neural networks are trained, one for each of the position or attitude channels. Since conventional thrusters are not able to perform well under fault conditions, neural networks are utilized to achieve fault tolerance. Two important features of a neural network are adaptability to different situations and the ability to learn a dynamic phenomenon. Accordingly, such structures can calculate the commands needed to deal with the fault after it occurs. Neural networks are trained to overcome up to 90% of the fault of each thruster by making appropriate commands. Also, these networks are optimized so that the mission is done with optimal energy consumption. Finally, the efficiency of the controllers has been measured despite the uncertainties and in different initial and final conditions and in the presence of different faults
  9. Keywords:
  10. Satellite Formation Flying ; Fault-tolerant Control ; Neural Networks ; Six Degree of Freedom Satellite Control ; Position Control ; Leader-Follower Based Formation Control

 Digital Object List