Formal Approaches to Model Based Testing

点击翻译按钮获取中文摘要

Summary***State-based formalisms such as Finite State Machine (FSM), Communicating FSM (CFSM), State-Charts, UML and their extensions have been widely used to specify the behavior of systems in diverse areas such as machine learning, interactive systems, reactive systems, web services, and telecommunications as well as to describe the control structure of concurrent and distributed systems.***Within this context, the objectives of the proposed research are to develop formal methods for generation of effective test suites from specifications/requirements of state-based systems. Since state-based formalisms are commonly used to capture specifications/requirements of such systems, model based testing techniques are often employed during their development. This study addresses the problem of the construction of test suites from system specifications/requirements given in a state-based formalism. Main objective is to improve the state of the art in test generation from specifications and requirements of state-based systems. Formal approaches will be sought that will yield tests that can be used to determine the conformance of the implementations to their specifications and/or requirements. It is anticipated that the techniques developed for the construction of test suites may also be used for the reduction of test suites obtained manually or by semi-automated means. This study will also include generation of tests from multi-process specifications and designs such that the resulting tests can be applied in distributed test architectures without encountering controllability and observability problems. Requirements based regression testing is also a part of this study where differences in the requirements of evolving systems will be utilized to i) generate regression test suites, ii) select regression test suites from existing test suites, iii) reduce regression test suites constructed manually. The differences in successive models of requirements will form a basis for identifying the effects of the modifications made in the models and the system under test. By utilizing these identified effects, regression test suites will be formed. Earlier works on requirements based regression testing consider the effects of each modification one at a time. This clearly is not realistic and all modifications made on a version to obtain its successor version need to be considered altogether. Hence, this study will revise the definitions of effects of modifications accordingly and the existing methods for regression test suite generation, reduction, and selection will benefit from the results of this study for their increased accuracy.***********