ADMSv2

Abstract

This paper presents ADMSv2, an end-to-end data-driven system that enables real-time and historical data analytics and machine learning tasks over big, streaming, spatiotemporal data. ADMSv2 employs a unified multi-layered architecture that integrates several open-source frameworks to collect, store, manage, and analyze a variety of data sources, including massive traffic sensor data, bus trajectory data, transportation network data, and traffic incidents data. ADMSv2 enables numerous applications in intelligent transportation, urban planning, public policy, and emergency response, all of which are critical for city resilience. Here, we demonstrate three application scenarios running on top of ADMSv2 to showcase the efficiency of its capabilities of query processing on real-world streaming and historical data as well as real-time data analysis using deep learning for traffic forecasting.