Inferring Road Maps from GPS Traces

Google Tech Talk March 7, 2013 (more info below) Presented by James Biagioni ABSTRACT Driven by the near-ubiquitous availability of GPS sensors in a variety of everyday devices, the past decade has witnessed considerable interest in the automatic inference and construction of road maps from GPS traces. Existing methods for performing this task can be largely divided into three categories: k-means clustering, trace merging, and kernel density estimation. In this talk I discuss the three classes of existing algorithms guided by examples from canonical work in the literature, and illustrate their strengths and weaknesses. In light of these findings I then introduce and explain the operation of our new hybrid map inference algorithm, developed to overcome these existing methods' limitations with GPS-trace coverage disparity and measurement error. Lastly, I discuss some directions for future work in this problem domain. For further details on the material presented in this talk, please refer to the following research papers: James Biagioni and Jakob Eriksson. "Map Inference in the Face of Noise and Disparity." In SIGSPATIAL GIS, pages 79-88. ACM, 2012. Xuemei Liu, James Biagioni, Jakob Eriksson, Yin Wang, George Forman, and Yanmin Zhu. "Mining Large-Scale, Sparse GPS Traces for Map Inference: Comparison of Approaches." In KDD, pages 669-677. ACM, 2012. James Biagioni and Jakob Eriksson. "Inferring Road Maps from GPS Traces: Survey and Comparative Evaluation." In 91st Annual Meeting of the Transportation Research Board, 2012. All of which are freely-available on my website: http://www.cs.uic.edu/Bits/JamesBiagioni Speaker Bio James Biagioni is a PhD candidate in his final year at the University of Illinois at Chicago, where he is advised by Professor Jakob Eriksson in the BITS Networked Systems Laboratory. James' research interests are centered around the problem of inferring interesting and useful phenomena from sensor data, and its applications.