Design And Implementation Of SLABSp, An Agent-Oriented Programming Language

Agent-oriented programming is very essential for agent theory and technology. It's one of the most important programming methodologies beyond object-oriented programming, and it can be widely applied to many areas, such as Internet, human computer interface, autonomous craft and space probe control. Agent-oriented programming deals with how to build software agents with autonomous, pro-active, reactive, and social properties. In the formal specification level, SLABS proposed a number of novel language facilities that intended to support the development of agent-based systems, such as: organizing agent system in a modular structure, called castes, which naturally represents the structure and behavior characteristics of agents in encapsulated units; using scenarios to describe agent behaviors in the context of environment situations. These language facilities can make agent-oriented programs more readable, reusable and maintainable. But currently, they are hardly supported by any existing agent-oriented programming languages or platforms.This dissertation reports our attempts to support the language facilities from the view of programming languages, and presents an approach to facilitating the scenario mechanism and caste mechanism in agent-oriented programming. The basic idea is to extend object-oriented programming language to support agent-oriented programming, where Java is chosen as the base language. Firstly, the language framework of agent-oriented programming, SLABSp is presented, mostly conforming to SLABS, whose syntax is extended based on Java. Scenario mechanism and caste mechanism are introduced as the new features in the programming language. Secondly, the underlying object models are defined to serve as the semantics of the language, where agents are modeled by a couple of objects. A compiler has been built to compile the agent-oriented programs into Java. Finally, a running platform has been constructed as the multi-agent runtime environment of SLABSp, and some examples are given to demonstrate the usage of scenarios and castes in agent-oriented programming.