Gustavo Augusto Lima de Campos

Automated test case prioritization with reactive GRASP

Modifications in software can affect some functionality that had been working until that point. In order to detect such a problem, the ideal solution would be testing the whole system once again, but there may be insufficient time or resources for this approach. An alternative solution is to order the test cases so that the most beneficial tests are executed first, in such a way only a subset of the test cases can be executed with little lost of effectiveness. Such a technique is known as regression test case prioritization. In this paper, we propose the use of the Reactive GRASP metaheuristic to prioritize test cases. We also compare this metaheuristic with other search-based algorithms previously described in literature. Five programs were used in the experiments. The experimental results demonstrated good coverage performance with some time overhead for the proposed technique. It also demonstrated a high stability of the results generated by the proposed approach.

MAS-ML 2.0: Supporting the modelling of multi-agent systems with different agent architectures

Abstract Multi-agent systems (MAS) involve a wide variety of agents that interact with each other to achieve their goals. Usually each agent has a particular internal architecture defining its main structure that gives support to the interaction among the entities of MAS. Many modelling languages have been proposed in recent years to represent the internal architectures of such agents, for instance the UML profiles. In particular, we highlight MAS-ML, an MAS modelling language that performs a conservative extension of UML while incorporating agent-related concepts to represent proactive agents. However, such languages fail to support the modelling of the heterogeneous architectures that can be used to develop the agents of an MAS. Even worse, little has been done to provide tools to help the systematic design of agents. This paper, therefore, aims to extend the MAS-ML metamodel and evolve its tool to support the modelling of not only proactive agents but also several other architectures described in the literature.

Rational Agents for the Test of Rational Agents

Intelligent agents are seen as a promising technology to deal with the development of complex systems. Aiming at improving the agent performance, the utility function is responsible for mapping a possible state ,or set of state, with a utility degree associated. Although there are some efforts to facilitate the development of agent based systems, little has been done to propose methods and techniques to test these systems. In this work, we propose the conception of an approach that uses rational agents who are able to test others rational agents. More specifically, in addition to the designer of the agent, the approach involves a tester agent, a monitoring agent and one or more agents for representing the task environment of the tested agent. The test result must indicate the performance of the agent and, especially, the goals that are not being met.

A management scheme of SRLG-disjoint protection path

With the emergence of new applications and requirements it became necessary to create new monitoring and reactive configuration mechanisms to try to meet the SLAs (Service Level Agreements). In WDM (Wavelength Division Multiplexing) optical networks, one way of trying to fulfill these agreements is by using pre-established protection paths. However, despite guaranteeing that traffic will be rapidly routed to its protection path in case of failure, there is no guarantee that the latter will be capable of meeting the contracted SLA in accordance with the bit error rate of its links. In this article we propose a scheme for monitoring and selecting the SRLG (Shared Risk Link Group) protection path disjointed from the main path using Genetic Algorithms, Fuzzy Logic in a PBM (Policy Based Management) platform denominated GAFUDI.

Towards the modeling reactive and proactive agents by using MAS-ML

The existence of MAS where agents with different internal architectures interact promotes the need for a language capable of modeling these applications. This paper aims to extend the MAS-ML language in order to support the modeling of proactive and reactive agents.

Monitoring and Diagnosis of Faults in Tests of Rational Agents based on Condition-action Rules

In theoretical references available to guide the design of agents, there are few testing techniques to validate them. It is known that this validation depends on the selected test cases, which should generate information that identifies the components of the agent tested that are causing unsatisfactory performance. In this paper, we propose an approach that aims to contribute to the testing of these programs, incorporating the ProMon agent in the testing process of rational agents. This agent monitors the testing and diagnosis of faults present in a tested agent, identifying the subsystem information-processing agent that is causing the faults to the designer. The first experiments are aimed at evaluating the approach by selecting test cases for simple reactive agents with internal states and in partially observable environments.

Artificial immune systems in intelligent agents test

Intelligent agents consist in a promising computing technology for the development of complex distributed systems. Despite the available theoretical references for guiding the designer of these agents, there are few proposed testing techniques to validate these systems. Its known that this validation depends on all the selected test cases, which should provide information regarding the components in the structure of the agent that show unsatisfactory performance. This article presents the application of Artificial Immune Systems (AIS), through Clonal Selection Algorithm (CLONALG), for the problem of optimization of selection of test cases for testing computing systems that are based on intelligent agents. In order to validate the use of CLONALG, comparisons between the Genetic Algorithms (GA) and Ant Colony Optimization Algorithms (ACO) techniques were performed. In the experiments with the approach testing intelligent agents with different types of architecture in partially and completely observable environments, the approach selected a group of satisfactory test cases in terms of the generated information about the irregular performance of the agent. From this result, the approach enables the identification of problematic episodes, allowing the designer to make objective changes in the internal structure of the agent in such a way to improve its performance.

Using artificial immune systems for intelligent agent testing

Intelligent agents consist in a promising computing technology for the development of complex distributed systems. Despite the available theoretical references for guiding the designer of these agents, there are few proposed testing techniques to validate these systems. It’s known that this validation depends on all the selected test cases, which should provide information regarding the components in the structure of the agent that show unsatisfactory performance. This article presents the application of Artificial Immune Systems (AIS), through Clonal Selection Algorithm (CLONALG), for the problem of optimization of selection of test cases for testing computing systems that are based on intelligent agents. In order to validate the use of CLONALG, comparisons between the Genetic Algorithms (GA) and Ant Colony Optimization Algorithms (ACO) techniques were performed. In the experiments with the approach testing intelligent agents with different types of architecture in partially and completely observable environments, the approach selected a group of satisfactory test cases in terms of the generated information about the irregular performance of the agent. From this result, the approach enables the identification of problematic episodes, allowing the designer to make objective changes in the internal structure of the agent in such a way to improve its performance.

A Problem-solving Agent to Test Rational Agents

Software agents are a promising technology for the development of complex systems, although few testing techniques have been proposed to validate these systems. In this paper, we propose an agent-based approach to select test cases and test the performance of rational agent. Interactions between agent and environment are realized in order to evaluate the agent performance for each test case. As a result, we obtain a set of test cases where the agent has not been well evaluated. Based on this result, the approach identifies the goals that are not met by the agent and reported to the designer.

Seleção de Casos de Teste para Agentes Racionais

Agentes racionais consistem em uma tecnologia da computacao promissora para o desenvolvimento de sistemas distribuidos complexos. Apesar dos referenciais disponiveis para orientar o projetista desses agentes, existem poucas tecnicas de testes propostas para validar esses sistemas. Sabe-se que essa validacao depende dos casos de teste selecionados os quais devem providenciar informacoes a respeito dos componentes na estrutura do agente que estao com desempenho insatisfatorio. Este trabalho apresenta uma formulacao para o problema de selecao de casos de teste de agentes racionais atraves de uma abordagem solucao fundamentada em um agente orientado por utilidade para selecionar casos de teste atraves de metaheuristicas baseadas em populacao para testar o desempenho dos agentes racionais. Para cada caso de teste sao realizadas as interacoes entre Agente e Ambiente para obter a avaliacao de desempenho correspondente. Como resultado, obtem-se um conjunto de casos de teste nos quais o agente nao foi bem avaliado, com a respectiva avaliacao. A partir desse resultado, a abordagem identifica os objetivos que nao estao sendo satisfeitos e as falhas apresentadas pelo agente e repassa ao projetista.