Methodological and software artifacts support for agent-oriented component reuse and engineering

Systematically producing quality software systems that are distributed, heterogeneous, autonomous, and collaborative is the charter of agent-oriented software engineering. This combination of software characteristics is uniquely pursued by the agent-oriented paradigm since traditional distributed systems solutions lack the degree of autonomy and collaboration that typify multiagent systems. While traditional distributed systems have achieved mainstream deployment, agent-based systems and the frameworks they are built from are not industrial strength. A mature programming paradigm and complimentary reusable component infrastructure is not available to engineer agent systems. The object-oriented programming and software engineering communities have developed clearly delineated of reusable and compatible components, design patterns, interface (standards) specifications, frameworks, interoperability between frameworks, generators, and suitable architectures for connecting components together. The analogous mechanisms for agent-oriented programming and engineering are underdeveloped.; A comprehensive extension of existing software engineering techniques and the development of new agent-specific techniques are necessary to deploy large-scale and industrial strength multiagent systems that are dependable and efficient to build. This requires both developing programming and engineering methodologies while also developing a supporting collection of reusable software artifacts. The methodological contributions of this dissertation are to design agent-oriented programming. A decomposition framework specifies the layers components. A negotiation pattern template enables the comprehensive specification of communicative interaction between agents. The Reusable Interaction Protocols Modeling Language supports precise protocol descriptions for open agent environments that can be leveraged as reusable artifacts supporting the engineering of agent conversations. The software artifacts contributions of this dissertation provide extensions to the Java Agent Development Environment framework that demonstrate how hierarchical state-based programming of agent behavior can be used to achieve greater conversation control and programming flexibility. An internal agent architecture offers a novel separation of concerns structure that balances several design considerations, supports a high level of component reuse by modularizing agent behavior into managed subsystems and roles, and provides the infrastructure necessary for an agent to reason about its internal composition.