Towards an autonomic architecture for real-time traffic management in congested urban transportation networks

The objective of this research effort is to study the real-time traffic management problem for large-scale urban transportation networks. This research is the first attempt to adopt the autonomic control paradigm for real-time traffic management. The autonomic control paradigm is inspired by the human autonomic nervous system that handles complexity and uncertainties. It aims at realizing dynamic systems and applications capable of managing themselves with minimum human intervention. The architecture is envisioned to enable the transportation infrastructure to maintain an array of self-management capabilities such as self-awareness, self-diagnosis, self-anticipation, self-optimization, self-configuration, and self- learning. The proposed architecture adopts a hierarchal dynamic framework in which a set of distributed controllers are assumed. Through adequate communications, controllers are assumed to share information on traffic pattern as well as their control actions in their vicinities. These controllers are dynamically configured to operate either individually or in teams to develop integrated control strategies that best cope with the observed traffic pattern in the network.;Several sub-problems associated with the main objective of this research work are examined. The first problem aims at developing a traffic network partitioning mechanism for distributed traffic management applications. Two heuristics are developed for this problem. The first heuristic adopts a recursive approach in which the network is partitioned along its sparsest cuts. The second heuristic implements a greedy-based network coarsening methodology. The second problem is related to studying demand data sharing among local controllers. Two solution methodologies are developed, which differ in their level of dependency on historical traffic information. Finally, the third problem examines the optimal collaboration strategies among local controllers to provide efficient integrated traffic management schemes. An algorithm is developed to determine the most efficient coalition structures among controllers.;In order to examine the performance of the new architecture, a set of offline simulation-based experiments are conducted using hypothetical and real roadway networks. A traffic-simulation assignment model is used to evaluate the network performance considering recurrent and non-recurrent congestion scenarios. The results illustrate that more efficient traffic management plans could be obtained through allowing collaboration among the individual controllers. Travel time saving that ranges from 6% to 12% is reported especially during non-recurrent traffic congestion situations.