SmartBlackBox: enhancing driver's safety via real-time machine learning on IoT insurance black-boxes

The Internet of Things (IoT) is shifting from a purely technical perspective to being a technology with implications on society and its economy. Responding to the global rise of awareness on how the IoT impacts on important themes such as health, safety and social responsibility, we propose and evaluate an IoT device in a field where safety is critical: a smart black-box for the automotive sector, starting from the concept popularized by insurance companies. The SmartBlackBox is a device that supports on-board machine learning to classify the drivers' behavior and supply valuable insight on how to enhance their driving styles. It features reconfigurable hardware within an embedded System-on-a-Chip that is programmed to transform what is usually a simple IoT data-ingestion node into an intelligent companion that learns the drivers' behavior, supporting them in achieving a safer driving style. Compared to traditional black-boxes, thanks to accelerators synthesized on reconfigurable hardware, the SmartBlackBox enters the domain of cyber-physical systems as it supports faster data input streams coming from multiple sensors, ad-hoc data compression for edge-cloud communication, and, especially, realtime classification of driving maneuvers.