DriSafe: Developing Ubiquitous Systems for Safe Driving in Diverse Driving Micro-cultures

Drive tourism has become increasingly popular in the last decade; however, driving through a new city is challenging as the road, and traffic environments vary significantly across cities. A driver habitual to driving in one city may face severe difficulty adapting to a different driving environment, leading to road fatalities. Intrinsically, each city has its driving style, which is governed by the geospatial properties of the city, its traffic and road conditions, etc., that influence the actions taken by the driver. We developed a personalized driving guideline recommendation system while understanding the driving style specific to the driver and the city in which they are traveling. The system develops an explainable domain adaptation model to generate a set of critical guidelines and recommend the same to the driver while driving in a new city.

Another important aspect of driving across different environments is to correctly identify the driving anomalies. An anomaly in driving refers to any deviation from typical or expected driving behavior, manifesting in various forms and may indicate unsafe or irregular driving practices. Notably, anomaly signatures vary across various driving environments; so, we need a domain-adaptive method for extracting the anomaly information. In subsequent work, we developed DoAssist, a novel method that leverages driver-specific spatiotemporal multimodal data to detect driving anomalies in real-time and provides alerts to the driver. The core idea of the proposed solution is to explore additional sensing modalities available with modern wristbands, such as physiological signals, to augment the IMU data with rich contextual information.