Ontology-based Test Generation for Multi Agent Systems

This paper investigates software agents testing, and in particular how to automate test generation. We propose a novel approach, which takes advantage of agent interaction ontologies that define content semantic of agent interactions to: (1) generate test inputs; (2) guide the exploration of the input space during generation; and, (3) verify messages exchanged among agents with respect to the defined interaction ontology. We integrated the proposed approach into a testing framework, called eCAT, which can generate and evolve test cases automatically, and run them continuously.