The Common Vehicle Information Model (CVIM) is AutoMat’s data model that enables the exchange of harmonized and brand-independent vehicle big data. AutoMat’s idea is to develop an open vehicle big data marketplace that gathers vehicle data from various brands and different cars all over Europe. One key success factor in such a marketplace is a common understanding of information between all participants. However, vehicles are equipped with a variety of sensors, which vary between different car manufacturers, suppliers or even within the same series of a model. Uniform vehicle big data standards did not exist. Within the scope of the AutoMat project the CVIM has been developed, which aligns the proprietary vehicle landscape. CVIM removes brand-dependent information and builds a common denominator between different car manufacturers and various sensors.
CVIM uses a three-layered approach to harmonize vehicle data, which is shown in the figure below. On the bottom layer are Signals, the raw in-vehicle information providers. Signals may be car- or supplier specific and proprietary. Their originate from any source within a vehicle, e.g. a vehicle speed signal captured from the Controller Area Network (CAN)- or On-Board Diagnostics (OBD)-bus. The data aggregation process from Signals is described by Measurement Channels on the middle layer. Measurement Channels introduce a standardized format using common sample rate definitions and data types. Hereby, they form a common ground between information originating from different brands or vehicles. Actual data samples are stored inside Data Packages. Data Packages are exchangeable messages, which can be transferred throughout the whole AutoMat ecosystem and beyond. Data Packages build the basis for AutoMat Marketplace and future automotive big data.
The figure above shows the structure of one CVIM Data Package. It is separated in descriptive metadata and the aggregated data. Metadata supports the Marketplace in indexing and sorting the data by leveraging the statistic properties, timestamps and geographic estimate in form of a bounding box. In addition, it provides ownership information, privacy levels and data stakeholders. For the sake of completeness and validity, OEMs can sign data packages using sequence numbers, checksums and signature. The second part of Data Packages stores aggregated data of one Measurement Channel. This may be either a time-series recording or pre-processed value distribution in form of (geo-coded) histogram.
All in all the CVIM provides a lightweight and scalable vehicle big data model. CVIM has been developed, implemented and evaluated within the scope of AutoMat. Here, it enables service providers to easily access brand-independent vehicle big data via a single point of access, the AutoMat Marketplace.