Loading...

Designing a Distributed Controller for an Under-Actuated Running Robot with Learning Ability

Ehteshami, Vahid | 2020

616 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53420 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Salarieh, Hassan
  7. Abstract:
  8. Walking-robots have received much attention because of the variety of motion maneuvers they can produce and the many applications they have in various areas including rehabilitation. One of these maneuvers is running. In this study, a time-invariant controller is designed to dynamically stabilize a five-link and a seven-link running-robots in two dimensions. In the simulation of the seven-link model, an attempt has been made to measure the impact of the circular foot on the stepping pattern. Also, it is tried to present a dynamically simple model similar to the geometry of the human body. Simulations are performed in MATLAB software. The running is modeled with three phases of stance, flight and collision. The dynamic equations of the robot in each phase were extracted by the Lagrange method. The robot's heel-strike is also rigidly modeled. The controller in each phase guides the robot by zeroing predetermined outputs in the form of paths trajectories using feedback linearization method. The outputs are designed so that a rhythmic running will be presented by the robot on a limit cycle. The provided limit cycle has been determined and its stability has been investigated using the Poincare return map. A hierarchical structure control method for robot control is also presented. The performance of this structure is divided into high-level and low-level controllers. The high-level controller is a feedback linearization controller and keeps the robot on a stable limit cycle by following the virtual constraints. For the low-level control, the actuator of each joint with local information from the same joint controls the movement following pre-learned behavior. This behavior is estimated by the learning algorithm from the motion algorithm generated by the high-level controller given to the low-level controller. A series of critiques monitor the performance of the low-level controller and, if there is any problem, push the robot's controller to a high-level to re-train the low-level controller and return to the lower level if possible
  9. Keywords:
  10. Dynamic Stability ; Poincare Map ; Limit Cycles ; Learning Ability ; Underactuated Biped Robot ; Running Robot ; Distributed Control ; Hybrid Zero Dynamics

 Digital Object List

 Bookmark

No TOC