MOBILE AGENT ORIENTED ONTOLOGY BASED POLICIES FOR HONEY-BEE TEAMWORK ARCHITECTURE

Abstract

With exponential growth in applications complexity, software agents’ domain is making a remarkable progress. Complex and distributed applications need a group of software agents to perform the intended tasks, instead of standalone isolated agents. In software agents’ domain, limited work has been done in case of teamwork among mobile agents. Mobile agents which have the capability to roam around on various machines need special communication and coordination strategies as well as goal sharing and task accomplishment mechanisms. In this thesis, a teamwork strategy is proposed to enhance the task execution performance as well as mechanism to efficiently address the communication and coordination issues among group of mobile agents. The proposed technique is developed by conceptualizing the Honey-Bee teamwork strategy and named after it as “Honey-Bee teamwork architecture”. In association with it, the goal oriented ontology based policies concept is proposed for improving the mobility mechanisms of mobile agents. It also addresses the aspects of goal definition, task creation and execution on distributed machines by mobile agents. This goal oriented approach reinforces the teamwork architecture as well as its goal and task division characteristics. The ontology based policy technique facilitates in convenient assignment of goals and associated tasks setup information for mobile agents which were earlier difficult to create and execute as desired especially in dynamic environments. The proposed strategy is evaluated using combination of teamwork strategies on single and multiple machines scenarios. The proposed honey-bee technique was found more efficient in the distributed infrastructure especially for higher number of member agents in a team. It shows linear behavior as compared to exponential increase in task accomplishment time in case of conventional strategy. The combination of simplified task execution as well as reduced size of mobile agent requires less time in performing the desired functionality. Additionally, the earthquake management system (EMS) is proposed and discussed as proof of concept application. It highlights the usage of software agents in disaster management and expresses its teamwork capabilities to handle complex application scenarios. This leads to conception of major application domain and test bed where software agents may execute viin combination of various teamwork fashions. The major structure and behavior of proposed application is expressed and analyzed using Pi Calculus and Pi ADL formal techniques.