heoretical understanding of traffic flow has gotten a boost in recent years by applying methods from nonlinear dynamics (partial differential equation models ; continuous car-following models [2]), and statistical physics (cellular automata models ). This progress was secured by excruciatingly few real world measurements-too often the existing measurements or their descriptions were too incomplete to test the theory against them, or the quantities which were measured were not measured precisely enough, or they were simply not the ones which would yield insight.
Kerner and Rehborn now look at real world data from a theoretical angle. Having access to comprehensive data sets collected with modern equipment, they find three regimes for traffic flow: (i) free flow, (ii) congested "synchronized" flow, and (iii) jammed traffic. Regimes (i) and (iii) were predicted by existing (single-lane) theories; regime (ii) has been recognized before ("queue discharge" [4]), but it was never shown convincingly why it could not be a measurement artifact (a time average composed of pieces of (i) and pieces of (iii)). Kerner and Rehborn now show that regime (ii) is definitely different: Contrary to both (i) and (iii), consecutive data points in the flow-density plot do not fall in a line; and contrary to (i), the velocities on all lanes are equal (and fairly low). Inside these three regimes, the measurements suggest further differentiation.
The challenge now, obviously, is to explain these complex patterns by models which are (hopefully) simpler than the patterns themselves.