Online Assessment of Driving Riskiness viaSmartphone-Based Inertial Measurements

Assessing the driving-style from dynamic data is awell established line of research, which has tackled the descrip-tion of risky behaviours, the profiling of energy-consumptionhabits, and the detection of different driver’s characteristics fromthe analysis of motion data. In the last years, as smartphoneownership has become widespread, such an assessment has beenincreasingly relying on the measurements taken from the inertialsensors on board of the smartphone itself. This work stands inthis context, and it aims to design a 4-dimensional driving-styleassessment for insurance purposes. The main contribution isadding, to more common proxies of risky-driving, the dimensionof smartphone usage, the detection of which is performed throughan appropriate processing of smartphone-based inertial sensors,thus not relying on privacy-sensitive monitoring of phone usagebehavior. Physics-based, fine-grained dynamic features are usedto classify the overall riskiness of the driving-style, thus providinga comprehensive insight into the most discriminating features.The study is based on experimental data, collected over morethan five thousands kilometers of varied car trips