Driving Style Recognition Based on Naturalistic Driving: Volatilities, Decision-Making, and Safety Performances

Abstract

Driver’s personalities in decision-making and tactical maneuvering preferences developed over time is generally defined as “driving style”. Driving style recognition can be applied in many fields, such as driving behavior modeling and simulation. Since drivers’ driving styles vary among each other, applying a generalized decision-making model to understand microscopic driving behavior may overlook dangerous caused by risky driving behaviors. Despite individual differences are related with decision-making, existing studies of driving style recognition are mainly based on analyzing vehicle movements, such as speed profile and accelerations. Interactions with neighboring vehicles, such as relative velocity, gap, and drivers’ responses according to neighboring vehicles are rarely considered. Last but not least, safety performance is seldom included in the previous studies, the relationship between recognized driving styles and the driver’s safety performance was still unclear. This book chapter applies data mining approaches to extract driving styles based on naturalistic driving data. Three driving style identification approaches, volatility-based approach which utilities dynamic stability of vehicle movements, decision-based approach which focuses on decisions when interacting with neighboring vehicles, and safety-based approach which is based on safety performance are introduced and evaluated in the following sections.