01590574

Component

This paper is focused on developing machine learning algorithms to extract useful traffic information from crowdsourced data. In particular, high-resolution accelerometer data collected by smartphones onboard vehicles are analyzed, and advanced classification algorithms are developed to reliably detect vehicle stops (e.g., at traffic signals). Support Vector Machines (SVMs), Hidden Markov Models (HMMs), and Changepoint Detection Methods (CDMs) are employed to develop reliable algorithms for stop detection. These algorithms are optimized based on the field data collected by a custom Android application. The results demonstrate that both SVM-HMM and CDM models are effective in detecting stops with minimal false alarms. Overall, this research demonstrates the feasibility of collecting useful performance measures at arterials which are important for improving system operations.

This paper was sponsored by TRB committee ABJ70 Standing Committee on Artificial Intelligence and Advanced Computing Applications.

01584066

16-3993

English

14p

2016

Transportation Research Board 95th Annual Meeting

Digital/other

Figures; References; Tables

Accelerometers; Algorithms; Classification; Crowdsourcing; Machine learning; Smartphones

Data and Information Technology; Highways; I72: Traffic and Transport Planning

Transportation Research Board Annual Meeting 2016 Paper #16-3993

TRIS, TRB, ATRI

Jan 12 2016 5:45PM