Loading...

Model Predictive Control for Fast Descending Maneuver of Quadrotor

Nazeri Astanjin, Ahmad | 2022

170 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55503 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Fallah Rajabzadeh, Famida; Zohoor, Hassan; Nejat Pishkenari, Hossein
  7. Abstract:
  8. Quadrotors have limitation in performing fast descending maneuver due to entering a dangerous region in velocity space called Vortex Ring State(VRS) which cause considerable loss of lift that may lead to fall. The purpose of this project is to reduce time of fast descending maneuver without entering VRS region by using model predictive control. First, appropriate nonlinear dynamic model of quadrotor obtained by considering important aerodynamics effect. Then by considering velocity constraints in order to avoid entering the VRS region, a nonlinear model predictive controller(NMPC) has designed. This controller successfully reduce the time of the maneuver up to than ordinary descending. Because of high sensitivity of designed controller to inputs and in order to design stabilized and robust controller which can create capability of designing maneuver, a two layer hierarchical controller designed. Inner loop of the controller was a PD and outer loop was NMPC which was responsible for generating references. This robust controller successfully reduce the time of the maneuver up to than ordinary descending. Then by investigating different maneuvers, idea of omitting reference yaw which was produced by NMPC, has proposed. By implementing the suggested idea and considering bounded position for x and y up to , percentage of time reduction in performing fast descending maneuver increase up to . Although, this controller performance was great in order to fulfill the purpose of this project but the main problem was the huge computational cost of NMPC in a real-time application which solved by using deep learning techniques in order to imitate NMPC behavior. Then a deep robust LSTMNN network trained and implemented in flight simulation of Parrot Mambo project, which result in time reduction of fast descending maneuver up to without entering VRS region
  9. Keywords:
  10. Deep Learning ; Model Predictive Control ; Quadrotor ; Vortex Ring ; Fast Descending Maneuver ; Nonlinear Model Prediction

 Digital Object List

 Bookmark

No TOC