Diseño e implementación de un brazo robótico controlado mediante técnicas de aprendizaje por refuerzo

Abstract

The main objective of this work is the creation of a simulation from the design of a real robot controlled by RL (Reinforcement Learning), solving this through the MuJoCo physics engine, to finally be contrasted with the real robot, taking care of a key aspect in its operation: the energy consumed. In this application we will see how the manipulator works with the reward, and how we address its value per action or task performed.