Cloud-assisted trajectory data management and analysis

Abstract

Massive trajectory data with spatio-temporal information are generated from various devices everyday. Due to a wide variety of devices and data collection methods, these data differ in terms of frequency and precision, which brings challenges to data management and analysis. This paper presents TrajMan, a cloud-assisted framework for trajectory data management and analysis, which adds an edge layer between clients and cloud servers. With the assistance of edge devices, TrajMan supports fast data acquisition of extensive trajectory data, and reduces the WAN bandwidth and computation overhead for cloud servers by calibrating and compressing data locally. Cloud servers focus on data interpolation based on trajectory similarity to handle data in various frequencies, deriving more accurate trajectory features. To demonstrate the feasibility and performance of TrajMan, a preliminary case study of commute behavior analysis is conducted using real dataset, which shows 88% of the calculated commute distance has the relative deviation less than 20% and 80% of the calculated commute time deviates less than 10 minutes.