Estimating Traffic Conditions At Metropolitan Scale Using Traffic Flow

1 Traffic has become a major problem in metropolitan areas around the world. It is of a great 2 importance to understand the complex interplay of road networks and traffic conditions. We 3 propose a novel framework to estimate traffic conditions at metropolitan scale using GPS traces. 4 Our approach begins with a coarse inference of network travel times by solving a convex 5 optimization program based on traffic flow theory. Then, we iteratively refine the inferred network 6 conditions and vehicle paths. Last, we perform a nested optimization to estimate traffic conditions 7 on road segments that are not covered by GPS data. The evaluation and comparison of our 8 approach over two state-of-the-art methods show up to 96.57% relative improvements. We have 9 further conducted field tests by coupling road networks of Beijing and San Francisco with 10 real-world GIS data, which involve 128,701 nodes, 148,899 road segments, and over 26 million 11 GPS traces. 12