Prototipo Animat de interacción simple con el ambiente I: un experimento de aprendizaje maquinal
Autonomous artificial agents should have learning features that allow them to behave successfully in the presence of unexpected events which they are not programmed for. The more experience the agent have with its world, the more useful information it can acquire for enhance its behavior in front of...
| Autor Principal: | |
|---|---|
| Formato: | Artículo (Article) |
| Lenguaje: | Español (Spanish) |
| Publicado: |
Universidad Distrital Francisco José de Caldas
2001
|
| Materias: | |
| Acceso en línea: | http://hdl.handle.net/11349/19699 |
| Sumario: | Autonomous artificial agents should have learning features that allow them to behave successfully in the presence of unexpected events which they are not programmed for. The more experience the agent have with its world, the more useful information it can acquire for enhance its behavior in front of similar situations. A recent and interesting technique is reinforcement learning which is based in the agent seeking for rewards and avoiding for punishments like a trial-and-error approach. This paper describes the PAISA I an autonomous agent that can learn an adequate action-selection policy for maximize the amount of food it can find in a unpredictable world, in spite of a small state-action space. |
|---|