Aprenentatge per reforç aplicat a un cas de recursos compartits

Abstract

[en] This thesis explores the theoretical concepts needed to generate an ethical embedding, as well as the development of prior theoretical knowledge for understanding. Ethical embedding involves generating a Markow decision process where optimal policies are ethical based on a multi-objective Markow decision process where at least one of them follows an ethical criterion. Finally, it includes the implementation of the knowledge through the adaptation of the Common Game problem proposed by the company DeepMind and its subsequent resolution through the algorithms previously seveloped in a theoretical way.