Learning algorithms based state estimation, optimization and control of nonlinear processes

Abstract

Machine learning (ML), including deep learning and reinforcement learning, offers powerful tools for addressing complex problems. This thesis leverages ML to enhance state estimation, system identification, and optimization in non-linear systems, where traditional methods often fall short. Key focus areas include improving accuracy in capturing complex system dynamics, extracting system characteristics directly from data, and solving non-convex problems. The thesis demonstrates these methods through applications in aircraft dynamics and smart sensor networks for IoT technologies, highlighting the potential of ML to enhance the performance, reliability, and adaptability of control systems.