SmrtComp: Intelligent and Online CAN Data Compression


In modern vehicles, efficiently storing CAN bus data is crucial for on-board diagnostics, performance monitoring, analysis, and the investigation of failures and accidents. In past work, an EDR (Event Data Recorder) akin to the ones in aircrafts has been mooted. The state-of-the-art in this field comprises proposals that propose efficient lossless compression of CAN data for such analyses – this limits the data storage capacity. Our contribution, SmrtComp, aims to achieve a much larger storage efficiency by storing recent data in a lossless format and compressing older data in increasingly lossy formats. The system tries to approximately adhere to an expected accuracy vs time curve (a QoS metric) that is specified a priori. Our study evaluates SmrtComp’s performance using various metrics such as the compression gain (CG), root mean square error (RMSE), total data storage and the preservation of key features. SmrtComp was implemented on an ARM Beaglebone board; it was used to store realistic traces and synthetic traces obtained from CAN bus simulators. We achieve a 3.2× higher data storage efficiency as compared to the closest competing work and outperform a popular lossy algorithm by 94.33% in terms of the RMSE. SmrtComp achieves line-speed compression of CAN bus data making it a promising solution for managing large data volumes, and we also show that our compression method preserves anomalies. To the best of our knowledge, this is the first hybrid and tunable compression system in this domain.

International Conference on Intelligent Transportation Systems (ITSC)