Real-time traffic modeling and estimation with streaming probe data using machine learning

摘要: Traffic information systems play an important role in the world as numerous people rely on road transportation network for their most daily functions. This dissertation proposes a general system architecture processing traffic data and disseminating accurate, timely via internet. It also specifically addresses challenges with estimating arterial conditions using only from GPS probe vehicles. promises to be ubiquitous source of years come transit agencies decrease investment traditional fixed-location sensing infrastructure. The introduces design implementation Mobile Millennium system. A joint project between UC Berkeley, Nokia Navteq, aggregates sources, runs state art estimation forecast algorithms, provides drivers other targets. took over two build result is robust framework any researcher access vast stores quickly easily well test number algorithms. For conditions, this hybrid approach leveraging advances fields machine learning theory (based hydrodynamic theory). foundation model. variety model/algorithm approaches are presented, one ultimately proving superior rest that should carried forward research area continues.