Research Statement

The widespread availability of inexpensive GPS devices has made it feasible to passively collect a large volume of GPS trace data by placing such devices in vehicles and by letting individuals carry them. Each trace represents the trajectory of a vehicle or a person over time. Using more sophisticated sensing devices, additional information (e.g. the speed/acceleration of a vehicle, temperature and humidity values inside a vehicle) can also be collected. A primary goal of this research is to develop techniques that can provide additional insights to transportation planners through the analysis and mining of trace data. A long term goal of this research is to develop a prototype system that can continuously process trace data and proactively notify transportation planners regarding potential bottlenecks and accidents. Benefits of the proposed research include the following:

(a) By a careful analysis of the trace data generated by GPS devices carried by people, one can obtain a model of transportation-related activities for people. For example, by overlapping the trace data on a map and suitably partitioning the data into segments, trip purposes can be identified more precisely. Models obtained on the basis of this approach would be more accurate than those obtained using survey data.

(b) Analyzing the GPS trace data can also provide additional insights regarding the modes of transportation used by people during different parts of the day.

(c) By mining the trace data from vehicles, one can obtain predictive models to estimate performance measures of traffic flow and to identify potential bottlenecks and accidents.

Some challenges that arise in analyzing the trace data are the following:

(a) The volume of data to be stored and analyzed is extremely large. Thus, algorithmic techniques that can efficiently compress, analyze and mine the data are needed.

(b) There may be errors in the trace data (because of a GPS device???s limited accuracy) and some of the data may be missing (because the quality of the signal from the satellite may be poor in certain regions). Thus, the data needs to be carefully cleansed before analyzed.

(c) In certain applications (e.g. monitoring traffic flow), trace data appears as a collection of real-time data streams. Special stream-oriented operations must be carried out on the data to obtain useful information.

The research team includes a faculty member from the Department of Geography & Planning with expertise in various aspects of transportation and three faculty members from the Department of Computer Science with expertise in data mining, machine learning, processing stream data and algorithm design. In addition, some researchers from General Electric Global Research Center (Niskayuna, NY) have expressed willingness to participate in this research. Thus, the research team is well qualified to undertake the proposed research.

This project was cosponsored by the Research and Innovative Technology Administration of the U.S. Department of Transportation through the University Transportation Centers program.