Big data architecture for connected vehicles: Feedback and application examples from an automotive group

Nowadays, using their onboard built-in sensors and communication devices, connected vehicles (CVs) can perform numerous measurements (speed, temperature, fuel consumption, etc.) and transmit them, in a real-time fashion, to dedicated infrastructure, usually via 4G/5G wireless communications. This raises many opportunities to develop new innovative telematics services, including, among others, driver safety, customer experience, location-based services and infotainment. Indeed, it is expected that there will be roughly 2 billion connected cars by the end of 2025 on the world's roadways, where each of which can produce up to 30 terabytes of data per day. Managing this big automotive data, in real and batch modes, imposes tight constraints on the underlying data management platform. To contribute to this research area, in this paper, we report on a real, in-production automotive big data platform; specifically, the one deployed by Groupe PSA (a French car manufacturer known also as Peugeot-Citroen). In particular, we present the technologies and open-source products used within the different components of this CV platform to gather, store, process, and leverage big automotive data. The proposed architecture is then assessed through realistic experiments, and the obtained results are reported and analyzed. Finally, we also provide examples of deployed automotive applications and reveal the implementation details of one of them (an eco-driving service). (c) 2022 Elsevier B.V. All rights reserved.

Automated summary

Connected vehicles can perform various measurements and transmit data to dedicated infrastructure in real time. The global number of connected cars is expected to reach 2 billion by the end of 2025, generating up to 30 terabytes of data per day. This paper presents an automotive big data platform deployed by Groupe PSA and evaluates its proposed architecture.