Demystifying Interactions Between Driving Behaviors and Styles Through Self-clustering Algorithms

Abstract

We argue that driving styles demand adaptive classifications, and such mechanisms are essential for adaptive and personalized Human-Vehicle Interaction systems. To this end, we conduct an in-depth study to demystify complicated interactions between driving behaviors and styles. The key idea behind this study is to enable different numbers of clusters on the fly, when classifying driving behaviors. We achieve so by applying Self-Clustering algorithms (i.e. DBSCAN) over a state-of-the-art open-sourced dataset of Human-Vehicle Interactions. Our results derive 8 key findings, which showcases the complicated interactions between driving behaviors and driving styles. Hence, we conjecture that future Human-Vehicle Interactions systems demand similar approaches for the characterizations of drivers, to enable more adaptive and personalized Human-Vehicle Interaction systems. We believe our findings can stimulate and benefit more future research as well.

Publication
In Proceeding of the 23rd International Conference on Human-Computer Interactions (HCII'21)
Zeyu Xiong
Zeyu Xiong
Incoming PhD Student

My research interests include Human-Computer Interaction, Human-AI Collaboration, Accessibility, Ubiquitous Computing and Computer Music.