Loading...

System Identification and Control of Space Explorer Robot Arm Using Neural Network

Bahmanabadi, Hossein | 2020

429 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53296 (45)
  4. University: Sharif University of Technology
  5. Department: Aerospace Engineering
  6. Advisor(s): Asadian, Nima
  7. Abstract:
  8. In this study, two conventional methods for controlling flexible two-link robots based on artificial neural networks is improved. The first method is implemented by combining fuzzy logic and reinforcement learning in the form of a neural network. This method is modified in two steps. In the first step, the system feedback is changed and in the second step, the system Jacobin is used. This Jacobin can be the result of system identification. In the second method, an optimal controller is proposed for the system, which is also implemented in the form of a neural network. The performance of the neural-optimal controller is further improved by redefining the Bellman's principle of optimality. This redefinition manifests itself in terms of self-consistency constraints. Finally, the neural fuzzy method is compared with the optimal neural method. In other words, two approaches based on reinforcement learning, in one of which the cost function is defined at any time and in the other the cost function is defined in a time interval, are compared. It should be noted that in this study, only the planar movements of the arms are considered
  9. Keywords:
  10. Space Robot ; Reinforcement Learning ; Flexibility ; Artificial Neural Network ; System Identification ; Flexible Arm ; Bellman’s Optimality Principle ; Self Consistency

 Digital Object List

 Bookmark

No TOC