UlTraMan: A Unified Platform for Big Trajectory Data Management and Analytics Xin Ding∗;y Lu Chenz Yunjun Gao∗;y Christian S. Jensenz Hujun Bao∗;y ∗State Key Laboratory of CAD&CG, Zhejiang University, Hangzhou, China yCollege of Computer Science, Zhejiang University, Hangzhou, China zDepartment of Computer Science, Aalborg University, Aalborg, Denmark fdingxin@, gaoyj@,
[email protected] fluchen,
[email protected] ABSTRACT also in real-world applications. As an example, DiDi, the Massive trajectory data is being generated by GPS-equipped largest ride-sharing company in China, utilizes trajectory devices, such as cars and mobile phones, which is used in- data to provide services such as travel time prediction, de- creasingly in transportation, location-based services, and mand forecasting, and carpool scheduling [6]. The explosive urban computing. As a result, a variety of methods have increase in data volumes and the rapid proliferation of new been proposed for trajectory data management and ana- data analysis methods expose three shortcomings of tradi- lytics. However, traditional systems and methods are usu- tional trajectory data management platforms. ally designed for very specific data management or analyt- First, in real-life applications, trajectory data is collected ics needs, which forces users to stitch together heteroge- at a rapid pace. For instance, the Daisy Lab at Aalborg neous systems to analyze trajectory data in an inefficient University currently receives some 100 million points per manner. Targeting the overall data pipeline of big trajec- day from about 40,000 vehicles in Denmark. Consequently, tory data management and analytics, we present a unified traditional centralized systems [13, 23, 35] are or will be platform, termed as UlTraMan.