2013 Poster Sessions : Pathlet Learning: Understanding Shared Structure among Trajectory Collections

Student Name : Chen Chen
Advisor : Leo Guibas
Research Areas: Graphics/HCI
Location-based services are becoming ubiquitous in our daily lives due to advances in mobile devices such as GPS receivers in smart phones, cars, etc. As a result, a large amount of spatial-temporal data is being generated at an unprecedented rate. Among such spatial-temporal data, collections of trajectories, such as GPS tracks of trips collected by in-car GPS receivers, contain especially rich information. Understanding such collection of trajectories can be helpful in understanding how people choose to travel, which can contribute to many useful applications, such as route planning, city planning, traffic management, etc. Unfortunately, trajectory data is often noisy and low in sampling rate, e.g., 1 min/point. Such inherent uncertainty, along with the massive size of the data, pose great challenges for mining useful knowledge from such dataset. In our work, we develop a method called pathlet learning to understand large collection of trajectory data. By exploring shared structure among multiple trajectories, we show that a set of pathlets can be learned from the trajectory dataset. We also show that with these pathlets, most of the trajectories can be explained by concatenation of certain pathlets, which reveals the underlying mobility patterns of the trajectories.

Chen Chen received the B.E. degree in Electrical Engineering from Peking University, Beijing, China, in 2008, and the M.S. degree in Electrical Engineering from Stanford University, Stanford, CA, in 2010. He is currently pursuing the Ph.D. degree in Electrical Engineering at Stanford University. His current research interests include mobility pattern mining from large collection of trajectory data.