A RESTAURANT FINDER USING BELIEF-DESIRE-INTENTION AGENT MODEL AND JAVA TECHNOLOGY by

It has been becoming more and more important to design systems capable of performing high-level management and control tasks in interactive dynamic environments. Meanwhile the traditional software techniques are not sufficient enough to be applied to develop and maintain such complex systems. The agent-oriented/based systems, rooted in a different view of computational entities, offer prospects for a qualitative change in this aspect. In this work, we adopt the basic architecture of a BeliefDesire-Intention (BDI) agent model and develop a more intelligent and dynamic searching model for agent programming. In the BDI model, the desire is the goal to achieve or event to handle, the intention is a set of plans to realize the predefined goal or react to a specific situation, and the belief is the knowledge about the agent itself and the varying environment. Applying BDI concepts, we will build an experimental framework in our Restaurant Finder system. The belief, desire and intention are initially established as separate classes as in the object-oriented analysis and design, then these classes will be integrated as basic units for a BDI agent. The agent will possess the learning behavior based on the user's feedback and the principle of inferring preferences. In addition, the agent will also update its knowledge dynamically based on analysis of user's interaction with the system. The dramatic increase in the use and availability of mobile devices has resulted in the ability to access information anytime and anywhere. Our Restaurant Finder system is ix designed for use on mobile phones. This system has been constructed and implemented based on a three-tier architecture. The client tier is a J2ME (Java 2 Micro Edition) emulator for the primary interface. In the middle tier, the Apache Tomcat Server will be installed to process the client's request, update the system database, and send back the recommendation. The information tier is a database system using the IBM Cloudscape database. The Restaurant Finder system will demonstrate an example of BDI agent programming and the J2ME Mobile Information Device Profile (MIDP) client application design.