Ingrid Nunes

Developing and Evolving a Multi-agent System Product Line: An Exploratory Study

Software Product Line (SPL) approaches motivate the development and implementation of a flexible and adaptable architecture to enable software reuse in organizations. The SPL architecture addresses a set of common and variable features of a family of products. Based on this architecture, products can be derived in a systematic way. A multi-agent system product line (MAS-PL) defines a SPL architecture, whose design and implementation is accomplished using software agents to address its common and variable features. This paper presents the evolutionary development of a MAS-PL from an existing web-based system. The MAS-PL architecture developed is composed of: (i) the core architecture represented by the web-based system that addresses the main mandatory features; and (ii) a set of software agents that extends the core architecture to introduce in the web system new optional and alternative autonomous behavior features. We report several lessons learned from this exploratory study of definition of a MAS-PL.

On the development of multi-agent systems product lines: a domain engineering process

Multi-agent System Product Lines (MAS-PLs) are the integration of two promising technologies: Multi-agent Systems (MASs), which provides a powerful abstraction to model features with autonomous and pro-active behavior, and Software Product Lines (SPLs), whose aimis to reduce both time-to-market and costs in the development of system families by the exploitation of commonalities among family members. This paper presents a domain engineering process for developing MAS-PLs. It defines activities and work products, whose purposes include allowing agent variability and providing agent features traceability, both not addressed by current SPL and MAS approaches.

Extending web-based applications to incorporate autonomous behavior

Web applications are popular nowadays due to the ubiquity of the client and also because user experience is becoming each time more interactive. However, several tasks of these applications can be automated. Agent-oriented software engineering has emerged as a new software engineering paradigm to allow the development of applications that present autonomous behavior. In this work, we present two case studies of web-based systems, on which we added autonomous behavior by means of software agents. We also discuss some design and implementation issues found on the development of those systems and propose an architectural pattern as a consequence of our case studies.

Automating the product derivation process of multi-agent systems product lines

Agent-oriented software engineering and software product lines are two promising software engineering techniques. Recent research work explores the integration between them to allow reuse and variability management in the context of complex systems. However, the automatic product derivation process is not addressed in the current literature. In this paper, we present our approach to deal with multi-agent systems product lines (MAS-PL) variability management and automatic product derivation. Our approach is implemented as an extension of the GenArch product derivation tool. A case study illustrates how the proposed approach can be used to derive products (instances) from a MAS-PL.

BDI-Agent Plan Selection Based on Prediction of Plan Outcomes

The agent technology arises as a solution that provides flexibility and robustness to address dynamic and complex domains. Such flexibility can be achieved by the adoption of existing agent-based approaches, such as the BDI architecture, which provides agents with the mental attitudes of beliefs, desires and intentions. This architecture is highly customisable, leaving gaps to be fulfilled in particular applications. One of these gaps is the plan selection algorithm that is responsible for selecting a plan to be executed by an agent to achieve a goal, having an important influence on the overall agent performance. Thus, in this paper, we propose a plan selection approach, which is able to learn plans that provide possibly best outcomes, based on a current context and agents preferences. Our approach is composed of a meta-model, which must be instantiated to specify plan metadata, and a technique that uses such metadata to learn and predict plan outcomes. We evaluated our approach experimentally, and results indicate it is effective.

Developing multi-agent system product lines: from requirements to code

Many modern software systems have autonomous, open, context-aware and highly-interactive properties. The agent abstraction with its autonomous and pro-active characteristics and the related discipline of agent-oriented software engineering (AOSE) are promising paradigms to address these types of systems. Even though agents are frequently being adopted, little effort has been directed in AOSE methodologies toward extensive software reuse techniques, which can provide both reduced time-to-market and lower development costs. Multi-agent system product lines (MAS-PLs) are the result of the integration of AOSE with software product lines (SPLs). SPLs bring many reuse benefits to the agent domain through the exploitation of common characteristics among family members. In this context, this paper presents a domain engineering process for developing MAS-PLs. It defines activities and work products, whose purposes include supporting agent variability and providing agent feature traceability, both not addressed by current SPL and AOSE approaches.

Investigating explanations to justify choice

Many different forms of explanation have been proposed for justifying decisions made by automated systems. However, there is no consensus on what constitutes a good explanation, or what information these explanations should include. In this paper, we present the results of a study into how people justify their decisions. Analysis of our results allowed us to extract the forms of explanation adopted by users to justify choices, and the situations in which these forms are used. The analysis led to the development of guidelines and patterns for explanations to be generated by automated decision systems. This paper presents the study, its results, and the guidelines and patterns we derived.

Comparing Stability of Implementation Techniques for Multi-agent System Product Lines

Multi-agent systems (MAS) are increasingly being exploited to support autonomous recommendation of products and information to contemporary application users. Multi-agent system product lines (MAS-PL) promote large-scale reuse of common and variable agency features across multiple MAS applications. The development of MAS-PLs can be achieved through alternative MAS-specific frameworks (JADE and Jadex), and general-purpose implementation techniques, such as aspect-oriented programming (AOP). However, there is not much evidence on how these techniques provide better modularity, allowing the conception of stable MAS-PL designs. This paper reports an empirical study that assesses the modularity of a MAS-PL through a systematic analysis of its releases. The study consists of a comparison among three distinct versions of this MAS-PL, each one implemented with a different technique: (i) Jadex platform and configuration files; (ii) JADE platform and configuration files; and (iii) JADE platform enriched with AOP mechanisms. Our analysis was driven by fundamental modularity attributes.

Pattern-based explanation for automated decisions

Explanations play an essential role in decision support and recommender systems as they are directly associated with the acceptance of those systems and the choices they make. Although approaches have been proposed to explain automated decisions based on multi-attribute decision models, there is a lack of evidence that they produce the explanations users need. In response, in this paper we propose an explanation generation technique, which follows user-derived explanation patterns. It receives as input a multi-attribute decision model, which is used together with user-centric principles to make a decision to which an explanation is generated. The technique includes algorithms that select relevant attributes and produce an explanation that justifies an automated choice. An evaluation with a user study demonstrates the effectiveness of our approach.

Improving the Design and Modularity of BDI Agents with Capability Relationships

The belief-desire-intention (BDI) architecture has been proposed to support the development of rational agents, integrating theoretical foundations of BDI agents, their implementation, and the building of large-scale multi-agent applications. However, the BDI architecture, as initially proposed, does not provide adequate concepts to produce intra-agent modular software components. The capability concept emerged to address this issue, but the relationships among capabilities have been insufficiently explored to support the development of BDI agents. We thus, in this paper, propose the use of three different types of relationships between capabilities in BDI agent development — namely association, composition and generalisation — which are widely used in object-oriented software development, and are fundamental to develop software components with low coupling and high cohesion. Our goal with this paper is to promote the exploitation of these and other mechanisms to develop large-scale modular multi-agent systems and discussion about this important issue of agent-oriented software engineering.