Thomas Juan

ROADMAP: extending the gaia methodology for complex open systems

This paper is concerned with improving the software engineering of agent-based open systems. It critiques the existing Gaia methodology in the light of a motivating example of intelligent home networks. It describes the ROADMAP1 methodology, which extends Gaia with four improvements - formal models of knowledge and the environment, role hierarchies, explicit representation of social structures and relationships, and incorporation of dynamic changes.

The ROADMAP Meta-model for Intelligent Adaptive Multi-agent Systems in Open Environments

In this paper, we introduce the ROADMAP meta-model, designed to describe intelligent adaptive systems in open environments, using agent concepts such as roles. Developing intelligent adaptive systems creates new challenges in engineering software quality attributes such as correctness and reliability. The ROADMAP meta-model captures our understanding of properties of intelligent adaptive systems and our perspective on organizational concepts such as roles. The meta-model does not solve specific engineering problems, but provides a clean high-level structure where engineering issues can be grouped and classified. Infrastructure to support these issues can then be put in place progressively with consistency. An informal evaluation of the meta-model and comparison to related work is also presented. We expect developers of AOSE methodologies, tools, programming languages and frameworks to benefit from understanding the design and structure of the ROADMAP meta-model. By adopting the meta-model, the resulting methodologies, tools and languages may inherit its desirable characteristics and better support the development of intelligent adaptive systems in open environments.

The Gaia Methodology

Gaia (Wooldridge et al., 2000b) was the first complete methodology proposed for the analysis and design of MAS. However, the original version of Gaia suffered from the limitations of being suitable for the analysis and design of closed MAS and of adopting non-standard notation techniques. Several extensions to the basic Gaia methodology have been recently proposed to overcome these limitations. In this chapter, we summarize the key characteristics of the original Gaia methodology and present three extensions that have been proposed to improve Gaia and make it more suitable for the development of open MAS in complex environments.

Customizing AOSE methodologies by reusing AOSE features

Future large-scale software development projects will require engineering support for a diverse range of software quality attributes, such as privacy and openness. It is not feasible to create one monolithic methodology to support all possible quality attributes. Instead, we expect AOSE methodologies to be created and reused in a modular way. A modular approach enables developers to build custom project-specific methodologies from AOSE features in the same way applications are built from reusable off-the-shelf components.In this paper, we provide a conceptual framework for creating and reusing modular methodologies. This conceptual framework is based on the concept of an AOSE feature, which performs one or more development activities, such as analysis, and addresses one or more quality attributes, such as privacy. An AOSE feature encapsulates software engineering techniques, models, supporting CASE tools and development knowledge such as design patterns.We illustrate the applicability of our approach by modularizing four existing methodologies, Prometheus, ROADMAP, CaseLP and the conventional OO approach, into AOSE features.

A meta-model for intelligent adaptive multi-agent systems in open environments

In this paper, we introduce the ROADMAP meta-model, designed to describe intelligent adaptive multi-agent systems in open environments. The meta-model captures our understanding of the properties of such systems and our perspective on organizational concepts such as roles. It defines agent constructs such as agents and roles, and their inter-relationships such as aggregation, in a way that best facilitates the development of intelligent adaptive systems in open environments.We intend to use the meta-model for knowledge sharing. We expect developers of AOSE methodologies, tools, programming languages and frameworks to benefit from understanding the design and structure of the ROADMAP meta-model. By adopting the meta-model, the resulting methodologies, tools and languages should inherit its desirable characteristics and better support the development of intelligent adaptive systems in open environments.

Achieving Dynamic Interfaces with Agent Concepts

Traditionally, interfaces of software entities, modules and components are immutable at runtime and carry no information on the meanings of the underlying implementation. We believe this definition of interface imposes a rigid view or context on the interaction of software entities, which impedes software re-use and the development of open / adaptive systems. We propose a novel analysis and design construct called a dynamic interface. Dynamic interfaces capture the social ability of agents and can be evolved consistently at runtime. Semantic information about the underlying implementation is also built into the dynamic interface, while preserving information hiding. We envisage the new dynamic interface construct to be complementary to traditional immutable interfaces. The two constructs can be used consistently in applications to address different requirements. We describe a prototype implementation of the dynamic interface construct. The implementation leverages the agent concepts of goals, roles, protocols, agents and services from the ROADMAP meta-model. The initial evaluation on its flexibility and performance indicates that dynamic interfaces have potential as an industry strength design and implementation construct.

The software engineering of agent-based intelligent adaptive systems

Future software systems will be intelligent and adaptive. They will have the ability to seamlessly integrate with smart applications that have not been explicitly designed to work together. Traditional software engineering approaches offer limited support for the development of intelligent systems. To handle the tremendous complexity and the new engineering challenges presented by intelligence, adaptiveness and seamless integration, developers need higher-level development constructs. Agent concepts are natural to describe intelligent adaptive systems. Agent-based technologies have been incorporating software engineering practices, and have matured to offer useful insights and concrete practices to mainstream software engineers. This tutorial presents the state of the art in agent development from a software engineering perspective, focusing on practices that are applicable today. We have walked the audience through analysis, design and verification of a portion of a real-world problem, a smart home network. We show how agent concepts more naturally match the engineering challenges of such systems like trust between adaptive components. The audience had a hands-on experience with analyzing, and designing parts of the smart home network and learn how to incorporate agent technologies into their current projects.