These pages present the initial version of the MESSAGE methodology as described in project deliverable D1. The full deliverable is available for download in PDF form. Click here to explore the hypertext version now, or read on for a brief synopsis.
This document represents the first deliverable (D1) from the MESSAGE project P907. The document contains an initial version of a methodology for developing Agent Oriented Software, in particular Multi-Agent Systems for Telecommunications applications. The methodology is intended to be of benefit to software engineering personnel in EURESCOM client organisations. This deliverable has concentrated on the Analysis activity due to its central importance within the development process. A later deliverable will make recommendations on how the methodology handles later activities such as Design, Implementation and Testing.
The document presents a set of five analysis models, which can be used by analysts to capture different aspects of an agent system. The models are described in terms of sets of interrelated concepts. The five models are:
Organisation Model: This captures the overall structure of the system. It specifies the number and types of agents within the system and how they are related in terms of "power relationships".
Goal/Task Model: This captures what the Agent System and constituent agents do in terms of the Goals that they work to attain and the tasks they must accomplish in order to do so. The model also captures the way that Goals and Tasks of the system as a whole are related to Goals and Tasks assigned to specific agents within the Multi-Agent System.
Agent Model: This model contains a detailed and comprehensive description of each individual Agent and Role within the MAS. This description provides an internal view including the Agent's Goals and the services they provide. This contrasts with the external perspective provided by the Organisation Model.
The Domain (Information) Model: This model acts as a repository of information (both entities and relations) concerning the problem domain.
The Interaction Model: This model is concerned with capturing the manner in which agents communicate with one another. It specifies interactions from both a high-level and low-level perspective.
The document also recommends notations for representing the model components in graphical or tabular form. Simple example models are also built in order to illustrate practical use of the notations.
Finally, a modelling process is recommended that describes the process by which the set of analysis models can be built. It describes the order in which the models are created and how information in one model acts as input to building of other models. It also describes how the models are progressively decomposed, starting with a Level 0 model in which the MAS is viewed as a single agent and progressing to Level 1 and so on, by successively decomposing each agent into component agents.
The entry point to the hypertext version of the Initial Methodology (Deliverable 1).
We actively encourage feedback on the initial version of the methodology to help us improve it. Please direct comments on the MESSAGE methodology in the first instance to Richard Evans, who is leading this Task within the project. Comments on the on-line hypertext version can be addressed to Paul Kearney.
The Methodology presented in Deliverable 1 is only an interim version. We are currently revising and extending it. A final version will be issued along with other project deliverables in the summer of 2001. Specific activities under way are:
If you are interested in receiving updates on progress, please subscribe to our mailinglist.