Symbolization for Large-scale Driving Corpus and Its Applications

Abstract

This paper presents symbolization approach of large-scale driving behavioral corpus and its applicability to various ITS problems. Recently, driving data has been collected for naturalistic driving study to analyze its potential risk based on distribution of natural driving. However, large-scale driving corpus is difficult to use effectively for actual ITS applications because of its hugeness and diversity of driving situations. In this paper, a data-driven symbolization was employed for summarizing and clustering time-series driving data, and by using more than 400 hour driving corpus, we evaluated its applicability to several ITS applications; similar-scene retrieval, and driving behavior detection.