@TECHREPORT{Día_Apr_2019,
author = "Díaz Latorre, Andrés Steven",
title = "Aprendizaje por refuerzo para control de sistemas dinámicos",
abstract = "Reinforcement Learning or RL - by its initials in English, is a branch of artificial intelligence that deals with an agent that receives information from an environment or environment in the form of states and actions, in addition to acting in the environment, resulting in a new state, the agent receives a reward as payment when taking an action. This reward is assigned to the new state, therefore, as the agent takes action the reward will have both positive and negative value. In the project a series of algorithms in Python language were used for the control of classic dynamic systems, using the Gym and Tensorflow libraries. Python was used because it is one of the most used programming languages ​​because it is open source, object oriented and because of the ease of installing packages. The learning methods used in the algorithms are available in Q-Learning, Deep Q-Learning and actor - critic, better known as A2C. In addition, a guide mode is presented and for educational purposes the step-by-step process for the creation of our own environments with the gym library as well as how to implement these algorithms in our own environments since much of this theme is available in English, even in some universities the RL usually summarize it with Q-Learning, but this branch of intelligence is larger",
year = 2019,
institution = "Universidad Autónoma de Occidente",
url = "https://red.uao.edu.co/handle/10614/11694",
}