Automating Functional Test Design with Model-Based Testing 2012-01-0010

Classic problems of functional software test development industry include often poorly documented coverage of specification requirements, challenging delivery cycles, high software quality requirements, costly test development, inflexible test sets, and high maintenance costs due to ever changing requirements. Technology and commercial software exist today that can help to resolve all of these issues by automating test design with model-based testing (MBT), i.e., by automatically generating tests and test documentation. In numerous case studies over a wide range of application areas MBT has been independently confirmed [9, 10] to cut testing costs significantly by increasing productivity, to enable of up to 91% code coverage during test execution [8], and to locate numerous defects that have escaped in conventional testing approaches. But deployment of MBT in organizations is a shift in paradigm, and requires a different way of thinking about test development and an adaptation of processes.

In this paper, we introduce automated test design and compare it with conventional software test development. We outline a methodology for developing functional tests from models that describe the operation of the system (aspect) to be tested. Although the approach is illustrated here with an AUTOSAR basic software module [7] example, it can also be applied in driver, system integration and even end-to-end testing. Automated test design is presented by showing modeling, steering of test generation, test case review, all the way to test script generation based on one of the commercially available MBT tools called Conformiq Designer.