Seamless Interactions Between Humans and Mobility Systems

Abstract

As mobility systems, including vehicles and roadside infrastructure, enter a period of rapid and profound change, it is important to enhance interactions between people and mobility systems. Seamless human—mobility system interactions can promote widespread deployment of engaging applications, which are crucial for driving safety and efficiency. The ever-increasing penetration rate of ubiquitous computing devices, such as smartphones and wearable devices, can facilitate realization of this goal. Although researchers and developers have attempted to adapt ubiquitous sensors for mobility applications (e.g., navigation apps), these solutions often suffer from limited usability and can be risk-prone. The root causes of these limitations include the low sensing modality and limited computational power available in ubiquitous computing devices. We address these challenges by developing and demonstrating that novel sensing techniques and machine learning can be applied to extract essential, safety-critical information from drivers natural driving behavior, even actions as subtle as steering maneuvers (e.g., left-/righthand turns and lane changes). We first show how ubiquitous sensors can be used to detect steering maneuvers regardless of disturbances to sensing devices. Next, by focusing on turning maneuvers, we characterize drivers driving patterns using a quantifiable metric. Then, we demonstrate how microscopic analyses of crowdsourced ubiquitous sensory data can be used to infer critical macroscopic contextual information, such as risks present at road intersections. Finally, we use ubiquitous sensors to profile a driver’s behavioral patterns on a large scale; such sensors are found to be essential to the analysis and improvement of drivers driving behavior.