Performance reliability of road networks has become a more prevailing concern for engineers and planners involved in network design. To improve network reliability the most, engineers and planners may need to decide which links need improvement and at the same time account for the route choice behavior of network users. Using a simple network example, a new reliability version of Braess's paradox is presented showing that increasing the capacity or decreasing the variability of travel time for some links may actually lead to a less reliable network. A bilevel programming model is proposed through which the reliability paradox can be avoided and an optimal improvement scheme for the existing links may be determined to achieve the most reliable network, subject to a budget constraint. Finally, a numerical example is presented to illustrate the model together with a sensitivity analysis-based algorithm.
Abstract