Quality of software test design: design platform and test process.

Summary Electronics in cars is becoming more and more complex and represents more than 30% of the overall cost of a car. For example, in a 2008 BMW 5 series, there can be up to 80 electronic computers communicating together and representing around 10 million lines of software code. Faced with this increasing complexity, automotive electronics manufacturers and suppliers are increasingly interested in efficient methods for developing, verifying and validating electronic modules. More precisely, they are focusing their efforts on the software part of these modules since it is at the origin of more than 80% of the problems detected on these products. In this context, we have conducted a research project whose objective is to propose a global approach to improve the quality of software embedded in vehicles. Our research starts with an audit of the processes and tools currently used in the automotive electronics industry. This audit allowed us to identify potential levers for improving software quality. Based on the results of the audit and taking into account the literature in the field of software quality, we have proposed a global approach to test case design for software products. Indeed, we have developed an automatic test generation platform for a software product. This platform consists in modeling the specifications of the software product to simulate it during tests, to focus on critical tests (having a high probability of detecting defects) and to drive the automatic generation of tests by quality criteria such as the coverage of the code and the specification but also the cost of the tests. The generation of critical tests is made possible by the definition of real use profiles per software product, as well as by the reuse of defects and tests capitalized on former projects. In addition to the algorithmic aspects of software testing, our approach takes into account organizational aspects such as knowledge and skills management and software project management. Our approach has been implemented on two real-life case studies of an automotive electronics supplier, with historical test data. The results of our experiments reveal significant quality gains: more defects are found earlier and in less time.