Specifying reuse concerns in agent system design using a role algebra

Abstract

During the design of an agent system many decisions will be taken that determine the structure of the system for reasons that are clear to the designer and customers at the time. However, when later teams approach the system it may not be obvious why particular decisions have been taken. This problem is particularly acute in the case of designers attempting to integrate complex “intelligent” services from many different service providers. In this paper a mechanism for recording these decisions is described and grouping functionality into Roles which can then be combined using the recorded design knowledge is subsequent development episodes. We illustrate how design decisions can be captured, discuss the semantics of the constructs we introduce and how these abstractions can then be used as the basis of reuse in an extension of the Zeus agent toolkit.