DETERMINING OPTIMAL LOCATIONS FOR VARIABLE MESSAGE SIGNS UNDER STOCHASTIC INCIDENT SCENARIOS

Abstract

This paper introduces a framework for finding an optimal set of locations to install a given a number of VMS (variable message signs) in a traffic network. Of particular interest is the planning problem of strategically locating a limited number of VMS to maximize their expected benefit under a variety of traffic incident situations that are not known a priori. Due to the randomness in demand patterns and incident characteristics combined with the complexity of the associated traffic and behavior dynamics, one seeks a solution that is optimal in the sense of minimizing a long run expectation of the system costs. The framework is formulated as a bilevel stochastic integer programming model. The model is constructed as a nested optimization problem in which there are two levels of decision-making. The upper level represents that of a planning agency wanting to evaluate a potential strategy and the lower level represents that of a traffic network capable of carrying out those strategies. The upper level seeks to minimize the total user travel time taken as an expectation over the space of stochastic incidents. A tabu search algorithm is employed at the upper level to generate potential solutions, which are then evaluated at the lower level by solving the user equilibrium dynamic traffic assignment (DTA) problem via simulation. To demonstrate the applicability of the proposed formulation, a solution algorithm is implemented to determine an optimal set of locations for installing VMS in the Fort Worth network. The results underscore the complex dynamics of traffic flow and tripmakers' travel behavior in the presence of real-time information especially in regard to VMS-induced diversions. Important factors affecting the solution include demand, network structure, VMS compliance rate and incidence characteristics

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 14 citations exist upstream.