Flow Reconstruction for Data-Driven Traffic Animation

Flow Reconstruction for Data-Driven Traffic Animation

David Wilkie

University of North Carolina

Jason Sewall

Intel Corporation

Ming C. Lin

University of North Carolina

Abstract

'Virtualized traffic' reconstructs and displays continuous traffic flows from discrete spatio-temporal traffic sensor data or procedurally generated control input to enhance a sense of immersion in a dynamic virtual environment. In this paper, we introduce a fast technique to reconstruct traffic flows from in-road sensor measurements or procedurally generated data for interactive 3D visual applications. Our algorithm estimates the full state of the traffic flow from sparse sensor measurements (or procedural input) using a statistical inference method and a continuum traffic model. This estimated state then drives an agent-based traffic simulator to produce a 3D animation of vehicle traffic that statistically matches the original traffic conditions. Unlike existing traffic simulation and animation techniques, our method produces a full 3D rendering of individual vehicles as part of continuous traffic flows given discrete spatio-temporal sensor measurements. Instead of using a color map to indicate traffic conditions, users could visualize and fly over the reconstructed traffic in real time over a large digital cityscape.

Downloads

ACM Transactions on Graphics (TOG), SIGGRAPH 2013
Volume 32 Issue 4, July 2013 Article No. 89
David Wilkie, Jason Sewall, and Ming C. Lin

Pre-print of the paper. The published version may contain changes.
The pre-print appendix for the paper.
The supplementary video. 54 MB.

Contact

The authors can be reached via wilkie@cs.unc.edu, jason.sewall@intel.com, and lin@cs.unc.edu respectively.