A Simplified Theory of Urban Congestion

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

This award explores the premise that traffic congestion levels on large cities may be well described with a theory that requires only three aggregate network parameters: average red and green times for traffic lights and average block size. This conjecture has been validated by the research team for a single city where data are available; verifying it with additional empirical data is therefore one of the main motivations for the proposed research. If successful, this research will have profound implications in our discipline and for the society at large. Current planning models might produce suboptimal results, because they rely on detailed origin-destination tables that may be impossible to estimate reliably, and traffic dynamics that must be simplified such that they do not adequately capture the characteristics of congestion. The proposed theory overcomes these deficiencies by employing a macroscopic approach whose results are consistent with microscopic traffic dynamics. The proposed theory enables large-scale field implementations of congestion-saving control strategies for big metropolitan areas. These strategies are robust, and independent of both origin-destination tables and of the details of the road network. This project will enhance the educational and research infrastructure by making available large amounts of empirical data and software tools for their analysis and for simulation. This project will seek to validate and expand the hypothesis that the Macroscopic Fundamental Diagram of an urban network, which gives average travel time as a function of average density, depends mainly on the proposed three parameters. To test this hypothesis, the research will employ empirical data from different cities around the world. The analytical framework from prior NSF-funded research will be expanded to understand the conditions where this theory is applicable. This will shed light on the effects in macroscopic network performance of traffic signal control, route choice models and congestion pricing. Stochastic processes techniques will be employed to generate bounds to approximate network capacity reductions due to short blocks, which has proved difficult in the past. The project will develop the concept of macroscopic dynamic traffic assignment by formulating the assignment problem and its solution through a partition of the network into a collection of subnetworks, each described by a Macroscopic Fundamental Diagram. Closed form solutions will be sought for simple configurations, which will be the building blocks to develop numerical algorithms to tackle general networks. This will allow the computation of congestion control strategies, traffic state estimation and forecast over large urban networks and in real-time, all of which is intractable with current methods.