Control of liquid level in coupled tanks «CT-100» using reinforcement learning

Abstract

This Document proposes an application where Reinforcement Learning (RL) could be applied to maintain the liquid on a desired level in a system with a known non-linear dynamics. In Artificial Intelligence (AI), Reinforcement Learning (RL) is one of the most important technics of Machine Learning (ML), an initial introduction to these fundamental technics is given in Chapter 1, which serves as an introduction before exploring deeper reinforcement learning techniques. In Chapter 2, an exploration of (RL) methods including Model-Based (MB) and Model-Free (MF) learning is introduced, then digging into two important algorithms; Policy Iteration (PI) and Q learning. Since this document contains a simulation, it is essential to develop a comprehensive coupled tanks system model using physical and mathematical laws such us the principle of mass conservation. This model will outline the key components and dynamics of the system, providing a foundation for accurate simulations. Chapter 3 details the construction and parameters of the system model. Finally, Chapter 4 introduces the application of reinforcement learning (RL) algorithms to control the liquid level in our coupled tank system. It includes a description of the MATLAB implementation results, followed by an overview of the results.