Autonomous Artificial Intelligent Agents for Bayesian Robotics

This paper addresses the problem of working with uncertain or incomplete information in multiagent systems. When we work with real life systems (like a robotic agent) we normally use models to "insert" our knowledge to the robot. Nevertheless, any model of a real phenomenon will always be incomplete due to the existence of unknown, hidden variables that will influence the phenomenon, causing the model and the phenomenon to have different behavioral patterns. In this paper we provide a new intelligent Bayesian agent architecture oriented towards Bayesian robotics that provides a framework for developing intelligent agent applications using Bayesian theory. This architecture allows the programmer to use a common framework for developing agents that needs to work with uncertain or incomplete information, using one data type, join probabilistic distributions, which allows to use a common exchanging uncertainty assessments in multi-agent systems.