Transformación del Q-Learning para el Aprendizaje en Agentes JADE

Abstract

Increased interaction between computer systems has modified the traditional way to analyze and develop them. The need for interaction between the system components is increasingly important to solve joint tasks, which individually would be very expensive or even impossible to develop once. Multi-agent systems offer an interesting and complete distributed architecture to execute tasks cooperate. The creation of a multi-agent system or an agent requires great effort so methods have been adopted as the deployment patterns. The pattern creates Proactive Obsever_JADE agents and include in each endowed with intelligence behaviors can evolve using machine learning techniques. The reinforcement learning is a machine learning technique that allows agents to learn through trial and error interactions in a dynamic environment. Reinforcement learning in multi-agent systems offers new challenges arising from the distribution of learning, such as the need for coordination between agents or distribution of knowledge, which should be analyzed and treated.