Experiences on Using Software Experiments in the Validation of Industrial Research Questions

摘要

Experimentation in software engineering is difficult. One reason is the large number of context variables [1] and the impracticality of experiments in an industrial setting. Considering the budgets of comprehensive projects, it is apparent that a company cannot double its effort executing a project twice, in order to compare two different approaches concerning process or method improvement. Performing experiments on the basis of small projects seldom offers solutions valid for industrial settings. Our commendation is a cooperation between industry and academic education. This approach offers multiple advantages. In this paper, we outline our experiences in experimental software engineering gained in about 20 experiments over the past 10 years by the Ulm University cooperating with Daimler AG, Research & Development, Ulm. Additionally we provide an insight into a current experiment and present our approach to experimental software engineering in further detail.