GOBEAT: Towards a Methodology to Support MAS Test Case Definition

• Renato Andrade UNEB
• Ana Patricia Fontes Magalhaes Mascarenhas UNEB
• Marco Antonio Simões UNEB

Resumo

Multi-Agent Systems (MAS) is a branch of Artificial Intelligence (AI) that works with distributed systems whose components are autonomous entities called agents. The soccer game has been used as a test bed to stimulate research in the MAS. A soccer team, i.e., a MAS, is composed of a group of players, i.e., agents, who should coordinate their actions towards a goal. The testing process in MASs has proven to be challenging. Autonomous agents are programmed to learn during their execution. So, running the same test scenario successively can lead to different results. It makes difficult the application of conventional software testing techniques. This paper proposes the Goal Behavioral Agent Testing (GoBeAT) methodology to assist in specifying the MAS test suite applied to the robot soccer domain. GoBeAT was tested in a case study by a robot soccer team of the Robotic World Cup (RoboCup) and showed positive results in the definition of a test suit.

Palavras-chave: MAS, Software Testing, Robot Soccer