Vehicular data acquisition and analytics system for real-time driver behavior monitoring and anomaly detection

Distracted drivers are causing a staggering number of accidents. Drivers deviate from their typical driving pattern due to reasons such as stress, distraction, drowsiness, and drunkenness. We present a vehicular data acquisition and analytics system for real-time driver behavior monitoring, anomaly detection, and alerting. On Board Diagnostic (OBD) unit available in most of the modern vehicles is used to collect the driver and vehicle related parameters. OBD-to-Bluetooth dongle is used to extract the data via a mobile app. Mobile app then transfers the data to a backend consisting of a Complex Event Processor (CEP). Then the proposed system first performs a historical analysis of completed trips to identify a driver's behavior using a Markov model and k-Means clustering algorithms. Moreover, Adaboost algorithm is used for safe driver-behavior monitoring. Based on these the CEP engine identifies multiple parameters to generate and classify a driver's driving style. Once a deviation from the typical driving pattern is detected the user is alerted. Experimental results show the proposed system can achieve more than 90% accuracy under various driving simulations.