Measuring inefficiency in dynamic traffic networks: a numerical study

Abstract

In traffic assignment, the network inefficiency (NI) is described as the ratio of total system cost in user equilibrium (UE) solution and that in system optimal (SO) solution. The worst case of the NI is represented by the price of anarchy (POA). An understanding of the range of NI in traffic networks is very useful in the design of new mechanisms to bridge the inefficiency. In this study, we systematically explore the trends in NI for dynamic equilibrium problems. Authors begin their analysis with the single bottleneck (SB) model and show that the NI converges to a certain value when demand increases. Then, we explore general networks with multiple origins and destinations. Authors observe that the NI tapers off to a steady value with increase in demand keeping all else the same. Extensive computational analysis illustrates the results in various traffic networks with different levels of demand. Open questions of interest to the research community will be discussed.

Keywords:

• Network inefficiency
• Price of anarchy
• Dynamic user equilibrium
• Dynamic system optimal
• Single bottleneck
• Cell transmission model