Ubiquitous sensing technologies make big data a trendy topic and a favored approach in transportation studies and applications, but the increasing volumes of data sets present remarkable challenges to data collection, storage, transfer, visualization, and processing. Fundamental aspects of big data in transportation are discussed, including how many data to collect and how to collect data effectively and economically. The focus is GPS trajectory data, which are used widely in this domain. An incremental piecewise regression algorithm is used to evaluate and compress GPS locations as they are produced. Row-wise QR decomposition and singular value decomposition are shown to be valid numerical algorithms for incremental regression. Sliding window–based piecewise regression can subsample the GPS streaming data instantaneously to preserve only the points of interest. Algorithm performance is evaluated completely as accuracy and compression power. A procedure is presented for users to choose the best parameter value for their GPS devices. Results of experiments with real-world trajectory data indicate that when the proper parameter value is selected, the proposed method achieves significant compression power (more than 10 times), maintains acceptable accuracy (less than 5 m), and always outperforms the fixed-rate sampling approach.
Real-Time Piecewise Regression
Application to Effective and Economical Collection of GPS Trajectory Data
Transportation Research Record
2017-01-01
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch