Pattern-based ontology learning technique for modeling intelligent agents in J2ME

To is an intelligent and mobile multi agent system (MAS) running on Java-enabled handheld devices. The universe of to agents is composed of a set of autonomous agents in interaction. The intelligence of to agents is carried out using a reasoning framework. This framework is currently specialized by a rule-based reasoning system. In this paper, we propose a learning framework to support ontology-based learning for intelligent agents. We mean by learning, the process of agent behaviour changing through its interaction with the surrounding world in order to improve its intelligence. Learning for agents is an important task to achieve full intelligence. The framework allows agents on mobile devices to acquire context data from environment, process ontological information and produce inference rules for the reasoner of the agents. We present our design principles and the implementation, and demonstrate the development of reasoning agents on reduced devices with learning capability. The ontology was developed in order to describe the knowledge of the surrounding computing environment of the intelligent agent. This research addresses the lack of a standard model for learning agents on mobile devices under J2ME. In this paper, we propose a framework for building an ontology-based learner producing and expanding inference rules for agent reasoning engine. The domain of ontologies to achieve learning of agents for mobile phones is an emerging area of research and perhaps even the key towards opening the intelligent interaction between agents on reduced mobile phones. We think that the development of agent learning software using design patterns, frameworks, and re-factoring for mobile devices in J2ME, will much improve software quality.