Regresar

Designing traffic flow management strategies under uncertainty

Abstract:

This paper proposes a framework for designing and adapting strategic traffic management under uncertainty. A primary function of strategic traffic management is the development of traffic management initiatives to mitigate potential large-scale congestion. However, the associated longer planning horizon-2-24 hours in advance of anticipated problems-means that congestion forecasts are highly uncertain. Furthermore, complex interactions exist between management initiatives and traffic propagation, producing a non-intuitive planning environment. The proposed adaptive planning framework captures these features, enabling quantitative design of traffic management initiatives that balance uncertainty with performance. Specifically, a decision tree is constructed to represent critical deviations in the forecast over the planning horizon. Corresponding decision points provide opportunities for management initiatives to be defined, and a genetic algorithm is employed to optimize the expected performance of the initiatives over the entire decision tree. Thus, this method identifies optimal strategies under forecast uncertainty, capturing the tradeoff between employing initiatives that may not be required and lost opportunities due to inaction. Performance of the adaptive design framework is compared with alternate design approaches, verifying the potential value of this new approach for real time decision making.