A big data-as-a-service architecture for naturalistic driving studies

Abstract

Naturalistic driving studies (NDS) collect driving data from various vehicles in order to observe driving behavior in an unobtrusive setting. Using an array of collection devices, NDS result in kinematic real-time data, but are also often enriched with additional data sets from surveys and external information from weather, road accidents, etc. This results in inevitable huge amounts of data that becomes challenging to handle due to its sheer volume and heterogeneity. Building big data systems from scratch requires high costs, and skilled labor and time, which slows down the progress of NDS. The aim of this paper is therefore to present a hybrid architecture based on big data-as-a-service (BDaaS) for NDS. The proposed architecture handles all aspects of big data challenges in NDS and inherently eases the deployment and maintenance of such systems. This enables NDS project members to focus more on the objective of the data collection rather than getting drowned in the big data management process.