Title: Optimization Based Control for Dynamical Systems Using Output Prediction
Committee:
Dr. Abdallah, Advisor
Dr. Wardi, Co-Advisor
Dr. Coogan, Chair
Dr. Vamvoudakis
Abstract: The objective of the proposed research is to develop a class of control algorithms for dynamical systems based on output prediction and optimization. The prediction used in these algorithms provides an estimation of the system output in the future. This estimation establishes a direct relationship between system inputs and outputs, allowing the design of optimization algorithms without considering the system dynamics. The idea of prediction, together with the optimization algorithm (including the Newton-Raphson method), has already achieved success in tracking control. The proposed research aims at improving this tracking method and also attempts to extend the idea of optimization based control to multi-agent systems. Three research directions are mentioned in this proposal, which are correlated and have different emphasises. Firstly, this research attempts to enhance the performance of the existing tracking controller and complement its theoretical analysis. Secondly, the research will apply machine learning techniques on this tracking technique, creating a model-free tracking controller. Thirdly, the research considers solving control problems over multi-agent systems. Such problems include consensus and formation control. This proposal also reviews the existing results under each direction and provides an estimated research timelines. These tasks and timelines will serve as a guideline for the future researches.