A Controlled Experiment on Coverage Maximization of Automated Model-Based Software Test Cases in the Automotive Industry
Ladda ner
Typ
Examensarbete för masterexamen
Master Thesis
Master Thesis
Program
Software engineering and technology (MPSOF), MSc
Publicerad
2016
Författare
Darwish, Rashid
Gwosuta, Nakyanzi Lynnie
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
Background: In the automotive industry, as the complexity of ECU’s increase, there is need for creation of models that facilitate early tests to ensure functionality; but there is little guidance on how to write these tests in order to achieve maximum coverage. Objective: We evaluated our prototype CANoe+ which uses the CANoe and GraphWalker tools Vs CANoe with regard to coverage maximization of generated test cases from the viewpoint of both software developers and software testers. The possibilities and limitations of this approach are also stated. Method: We conducted a controlled experiment using a nested design with the authors executing sample functions using the prototype and CANoe tools multiple times (240 runs) for each tool. The coverage data from the experimental runs of the two treatments i.e. CANoe+ and CANoe was collected and statistically analyzed. Results: CANoe+ was significantly more effective than CANoe at an alpha level of 0.05 for a one-tailed test while using the Mann-Whitney-Wilcoxon statistical test. Limitations: Using the presented approach could be unfeasible if one attempts to test the whole system in one go. It is best suited for when a specific module of the system needs to be tested after which one can move to the next module and then cover the whole system in the long run. Conclusion: The results reinforced the existing evidence regarding the superiority of using model-based testing techniques like CANoe+ over using testing methods like CANoe in automotive systems.
Beskrivning
Ämne/nyckelord
Data- och informationsvetenskap , Computer and Information Science