Latest revision as of 14:00, 26 January 2021

Abstract

Extensive urban and metropolitan development over the past decades has resulted in the need to better move goods and people over the nation's motorways. However, physical limitations in motorway infrastructure and inefficient use of existing infrastructure have caused considerable economic loss through the late delivery of goods and loss of productive labor, as well as environmental damage, national security concerns, and financial burden resulting from fuel wasted during congestion. Improving utilization of existing physical assets is a crucial and relatively inexpensive means to achieving those ends. The proliferation of inexpensive mobile devices over the past decade has fundamentally changed traffic management, both from the perspective of motorists, as well as planners and logisticians. Motorists now have rapid access to realtime traffic information, and can thus easily re-plan their journeys while already underway. Likewise, organizations that operate these services have access to motorists' whereabouts and itineraries, and are in a position to shape and manage traffic by providing instructions to individual drivers. This research addresses an exploratory agent-based approach towards modeling a road network in which agents 1. have varied routing behaviors and 2. may or may not have access to realtime rerouting capabilities. The objective is to explore what effect modifying driver portfolios will have on the ``price of anarchy" associated with a commuter network. This information can be used by urban planners to determine which driver and information-access portfolios result in the least vehicle delay and best use of motorway resources, thus informing policies and incentives that can alleviate network congestion. Results on a Northern Virginia test case demonstrate that active rerouting has overall modestly positive effects on network latency for small and moderately-sized populations; however, in larger populations, active rerouting can lead to competitive pressures between agents, resulting in overall degradation of performance. It is suggested that future work implement competing managerial agent classes, as well as test over additional geospatial datasets.

Original document

The different versions of the original document can be found in:

• http://dx.doi.org/10.18130/v3kx18

• https://libraetd.lib.virginia.edu/downloads/1j92g7669?filename=agm-thesis-signed.pdf