Optimizing the selection of representative configurations in verification of evolving product lines of distributed embedded systems

Electronics and computer science play a more and more prominent role in automotive technology. In the future the prevalence of those new technologies and the customers' demand for individuality leads to tremendously large configuration spaces of vehicle control systems. To cope with the resulting complexity in verification, new strategies need to be explored. One likely future challenge is to determine a set of vehicle configurations, such that the successful verification of this small set implies the correctness of the entire product family. This paper presents a method to address this task, based on exploiting communalities in architecture and requirements. We introduce efficient algorithms with provable quality guarantees for the optimization problems of choosing the minimum set of configurations necessary to verify all possible configurations and choosing the best k configurations to maximize the verification coverage of the entire product family. We discuss extensions of our method which allow requirement priorities and the consideration of configuration costs, and present a technique for automatically determining communalities in architecture and requirements which can be exploited by our optimization methods. We demonstrate the effectiveness of our method on an indicator light system product family. In this example a configuration reduction by 60% can be achieved