Candidate: Mahsa Parsapour
Title: Understanding Controller Structure using Value Function
Date: August 16, 2022
Time: 4:00 PM
Place: online
Supervisor(s): Kulic, Dana (Adjunct) - Mombaur, Katja (SYDE)
Abstract:
The goal of this research is to study the structure of controllers for nonlinear systems which may or may not be subject to a disturbance. We would like to understand whether the motion of nonlinear systems was generated by a feedforward controller or a combination of feedforward and feedback controller. Specifically, we are interested in understanding if the feedback is optimal or non-optimal. Our proposed method is based on the value function and Bellman's optimality condition. We formulate a constrained optimization problem to estimate the control signal, and the decision on the structure is made based on comparing the estimated and the reference experimental trajectories. We assume four structures as a reference to decide about the controller. The performance of the proposed algorithm is illustrated on nonlinear systems, and the final goal is to apply the proposed algorithm for human motion analysis.