A CYBERNETIC ARCHITECTURE OF PRACTICAL REASONING AGENT

摘要

During the last ten years, agent technology has been widely discussed in various research areas. An agent is a computer system that is situated in some environment, and that is capable of autonomous actions in this environment in order to meet its design objectives. There are at least two kinds of reasoning methods applied in constructing an agent, namely practical reasoning and theoretical reasoning. Practical reasoning directed towards actions ― the process of figuring out what to do by weighing different acting options against with agent desires and believes. While theoretical reasoning is directed towards beliefs. In this paper, we just focus on practical reasoning. A widely used BDI model for practical reasoning agent will be introduced, based on which our cybernetic-BDI architecture is discussed. 'Intelligence' and 'autonomy' are perhaps the most important aspects of agent system. Attempts to model intelligent behaviors of an agent, especially a practical reasoning agent, have been made from areas of computer sciences, psychology, sociology, and many others. Cybernetics provides a concrete mechanism for this purpose, namely by 'feedback', 'feedforward', and 'sociocybernetics'. We discuss first intelligent behaviors of agent systems in terms of reactivity, proactity, and social ability based on cybernetic concepts of feedback, feedforward, and sociocybernetics. Then based on the Belief-Desire-Intention (BDI) model and cybernetic principles we build up our Cybernetic-BDI architecture. With a pseudocode we validate the architecture for its practical implementation and fulfillment of required intelligent behaviors. In the last, a scenario of healthcare agent for diabetes patients is provided to show how the agent works according to the Cybernetic-BDI architecture.