Software Testing: According to Plan!

摘要

Automated planning and scheduling represents a branch of classical artificial intelligence (AI) research. Although initially used in robotics and intelligent agents, the use of planning for testing purposes has increased over the years. There sequences of actions representing interactions with the system under test guide the test execution towards reaching a test purpose. A planning problem is formally defined as a model that resembles the interaction with a real system under test (SUT). The obtained solutions are generated, i.e., the plans, directly correspond to test cases. The planning model offers the possibility to generate test cases with a great variety of interactions without the need for an extensive model definition. Until now, planning has proven to be efficient in detecting both functional and non-functional issues. The second play a major role in uncovering vulnerabilities in software. In fact, testing of any domain can be specified as a planning problem. The purpose of this paper is to summarize previous research in the domain of planning for testing including discussing examples from multiple domains.