Quality-aware scheduling of on-board and off-board data analysis in vehicle development

In automotive development, electronic control units (ECUs) in test vehicles generate large amounts of data that need to be analyzed efficiently. Usually, this data is evaluated afterward in a data center (off-board). Due to increasing data volumes, it is becoming more challenging to analyze this data in a timely manner, as required by reports for further development. For this reason, additional on-board systems are used that analyze the data while the vehicle is in motion. These on-board systems usually do not have full access to all data of the vehicle and are also prone to errors, which leads to a lower quality of the systems. Therefore, the distribution of jobs must take into account both, the availability and the quality of the system. In this work, we develop the algorithm DQSflex, which optimizes the distribution of jobs and provides a flexible tradeoff between data quality and timeliness. We compare DQSflex against quality-based and deadline-based scheduling, which is currently used. Our simulations of systems with realistic workloads show that DQSflex offers the best compromise between data quality and timeliness. The presented algorithm helps to better meet the diverse requirements in data analysis for vehicle development.