| Real-time traffic flow management has recently emerged as one of the promising approaches to alleviating congestion. This approach uses real-time and predicted traffic information to develop routing strategies that attempt to optimize the performance of the highway network. A survey of existing approaches to real-time traffic management indicates that they suffer from a number of limitations. In an attempt to overcome such limitations, the current study develops an artificial intelligence (AI)-based architecture for a real-time traffic routing decision support system (DSS), which embraces two emerging AI paradigms, namely case-based reasoning (CBR) and stochastic search algorithms. This architecture promises to allow the routing DSS to: (a) process information in real-time; (b) learn from experience; (c) handle the uncertainty associated with predicting traffic conditions and driver behavior; (d) balance the trade-off between accuracy and efficiency; and (e) deal with missing and incomplete data problems.; To illustrate the feasibility and effectiveness of the proposed architecture, the study develops and evaluates a prototype CBR routing system, built after the proposed architecture, for a real-world network in Hampton Roads, Virginia. Cases for the system's "seed" case-base are generated using a heuristic dynamic traffic assignment (DTA) model, which combines a macroscopic, deterministic, dynamic modeling framework with a stochastic search algorithm. A case selection framework is developed that provides for adequate coverage of the range of problems the system is expected to face, while keeping the size of the case-base manageable. An adaptation module is designed, combining features of simple adaptation with more elaborate adaptation using a domain model. A procedure that allows the system to learn from experience is also proposed. Using a set of new randomly generated cases, the performance of the prototype system is evaluated by comparing its solutions to those of the DTA model. The evaluation results demonstrate the feasibility as well as the effectiveness of the CBR approach. The prototype system was capable of running within the time constraints imposed by the problem, and produced high quality solutions using case-bases of reasonable size. |
