Study on the Efficiency of Test-Driven Development for Automotive Systems using Model-Based Design

Abstract

Since the automotive industry uses model-based design and strives towards more effective and qualitative development, this thesis explores the effectiveness of testdriven development with Simulink for development of an automotive system. To this end, a system for an automotive motor installation has been designed using a model-based design tool, Simulink, by alternating between two different development approaches; test-driven development (TDD) and a test-after (TA) approach. The quality of the code generated by these different approaches was evaluated using four metrics to cover different steps of model-based development (identification, semantics, relationships). The metrics used were: 1) Weighted blocks per module, 2) Coupling between modules, 3) Response for a module and 4) Lack of cohesion in blocks. To address the last step of model-based development (implementation), coverage of tests and time spent developing were used as metrics. Trends could be observed showing that TDD improved the quality of larger modules, while TA improved the quality of smaller modules. The time spent on developing the modules, together with the above mentioned quality metrics, showed that the development was more efficient when using TDD. Simulink as a design tool does not restrict any of these approaches, however, the test environment of Simulink significantly hinders the use of TDD.