It has been some time since announcing the Central Data Service Playground (CDSP) v0.1.0 release, and the project has made significant progress since then. This blog highlights key developments, milestones, and the latest advancements shaping the future of CDSP.

CDSP is an initiative of COVESA’s Data Architecture team in the Data Expert Group. The team is using CDSP to investigate data architectures, but it is also intended to be useful more generally by COVESA members and adopters of the COVESA data models. In particular, CDSP provides a platform for other COVESA working groups to explore, document, and publish use cases.

What v0.1.0 Brought to the Playground: Core Features

• Docker-based deployment of the VISSR vehicle data server and Apache IoTDB time-series database.
• An online documentation site.
• An example data feeder for the RemotiveLabs virtual signal platform.
• Support for Apache IoTDB as a new VISSR state storage backend, which was contributed upstream.

Together, these components provided a foundation for storing, processing, and analysing vehicle data using the VSS data model, allowing users to explore different data architecture approaches for their own use cases.

The RemotiveLabs bridge demonstrated a shift-left approach by enabling fully virtualized signal sources, while also showing how to integrate other data sources. VISSR provided a standardized interface for accessing vehicle data in the VSS model, and Apache IoTDB delivered both operational and analytical processing of time-series data.

Key Changes Since v0.1.0

• C4 Model: The C4 Model was adopted for visualising the software architecture as it is easy to learn, developer-friendly, and is intended to be readable by the widest possible audience.
• DIKW Layered Data Architecture: A layered Data, Information, Knowledge, Wisdom data architecture was adopted. Along with the enriched semantics enabled by the COVESA data models, this provides a framework for transforming data into information, from which actionable insights can be derived.
• Information Layer: To support the DIKW architecture, an Information Layer Server has been introduced, which provides a get/set/subscribe API for information to the layers above. It provides an abstraction to the underlying data storage. Currently, Apache IoTDB is supported.
• Knowledge Layer: This new container (component) was introduced along with the Information Layer to support the DIKW architecture. The Knowledge Layer takes information from the Information Layer and converts it into a knowledge graph. An AI semantic reasoner within the Knowledge Layer can then use the knowledge graph to link, evaluate, and infer new facts based on IF-ELSE style rules. An example is provided that infers a person’s driving style from VSS data in the information layer. PoCs showing this running have been shown at recent COVESA AMMs. Initially on a host-based system and most recently at the Spring AMM in Porto with the reasoner running on the Renesas R-Car V4H SoC. The PoC won the AMM demo showcase innovation award at the Berlin AMM.
• Timeseries Database: We updated to newer upstream versions of the upstream production Apache IoTDB timeseries database. Moving from v1.2.2 to v2.0.6.
• VISS/VISSR: We have updated to newer upstream versions of the VISSR implementation of the VISS API, which enables support for VISS v3.0 and later v3.1.
• Kuksa CAN Provider: We integrated the CAN Provider from the Kuksa project by extending it to support writing to Apache IoTDB northbound. The Provider can playback CAN dumps and supports both real and virtual SocketCAN connections. It is extensible using the Python CAN library ecosystem.
• Documentation: Switched to the Hextra template for the Hugo static website generator used to create the documentation site. Hextra allows us to develop a more usable site in the future.

Looking Ahead: Current Development Focus

• New Reasoners: The initial Knowledge Layer used the production semantic reasoner RDFox, which is proprietary and requires a license beyond a trial period. Currently, work is underway to add two OSS reasoners: first RDF4J, then CQELS. There are open pull requests on GitHub showing the work to add RDF4J.
• AI Context Retrieval: Work has begun to address multimodal Context Retrieval from different sources.
• S2DM/VDM: Support for the S2DM semantic data modelling approach and the VDM data model that uses it is underway.

More broadly, we are working to add new use-case examples, improve the documentation, and make it easier to get started, for example, by automating database schema generation. We are also exploring opportunities for collaboration with other COVESA groups.

For more details on the CDSP’s current features and future roadmap, please see the COVESA Central Data Service Playground (CDSP) Introduction and Status Update presentation from the Spring COVESA All Member Meeting. Details of the AI-related work are available in the Integrating Open-source Semantic Reasoners to CDSP presentation.

Get Involved

CDSP contains examples to inspire and help get you started. For example, using IoTDB analytics to significantly downsample a signal time series to save storage, processing, or transmission costs. Another shows how to use the AI features of the Information and Knowledge Layer to reason on a VSS data stream to derive driving style. There is also a reference Docker Compose file for building and starting the core components.

The components are loosely coupled and deployed as Docker containers, allowing you to mix and match them with other containerized technologies to create a solution that fits your specific use case. Links to project organization resources such as the project planning and chat can be found in the project README.

Please get involved and join our weekly meetings every Tuesday at 3 pm CET / 9 am ET / 6 am PT, found on the Community Calendar.