Declarative specifications for the development of multi-agent systems

摘要

The designing and implementation of a multi-agent system (MAS), where autonomous agents collaborate with other agents for solving problems, constitute complex tasks that may become even harder when agents work in new interactive environments such as the Semantic Web. In order to deal with the complexities of designing and implementing a MAS, a domain-specific language (DSL) can be employed inside the MAS's development cycle. In such a manner, a MAS can be completely specified by programs written in a DSL. Such programs are declarative, expressive, and at the right abstraction level. In this way the complexity of MAS development is then partially shifted to DSL development and the task herein can be much more feasible by using a proper DSL development methodology and related tools. This paper presents and discusses our methodology for DSL development based on declarative formal specifications that are easy to compose, and its usage during MAS development A practical case-study is also provided covering an example of a MAS's development for expert finding systems. By using denotational semantics for precisely defining the language, we show that it is possible to generate the language automatically. In addition, using attribute grammars makes it possible to have modular methodology within which evolutionary language development becomes easier.