Summary:
This research presents a dynamic user-optimal (DUO) route choice model for predicting dynamic traffic conditions, intended for off-line Advanced Traffic Management Systems (ATMS) and Advanced Traveler Information Systems (ATIS) evaluation and implementation. This DUO route choice model is formulated as a variational inequality (VI) and can be solved efficiently to convergence by the proposed diagonalization algorithm with discrete time intervals.
The test ·network selected for testing the, proposed dynamic user-optimal (DUO) route choice model is the ADVANCE Network. The ADVANCE Network is located in the northwestern suburbs of Chicago and covers about 300 square miles (800 square kilometers). To generate specific link travel times for the investigated network, the expanded intersection representation is employed. Using this network representation, each turning movement is coded as an individual intersection link, increasing the network scale to approximately three times larger than the conventional network representation. Nearly 10,000 nodes and 23,000 links are defined for the solution procedure. For most links a realistic traffic engineeringbased link travel time function, the Akcelik function, is adopted in this research, in place of the simplistic but widely used BPR (Bureau of Public Roads) function, to estimate delays and travel times for various types of links and intersections. Unexpected capacity reducing events causing nonrecurrent traffic congestion are analyzed with the model. Route choice behavior based on anticipatory and non-anticipatory network conditions are considered in performing the incident analysis, extending the capability of this model to contribute to the evaluation of ATMS and ATIS.
Four global network performance measures and a convergence index are defined to monitor the solution process of the model and assess the dynamic traffic condition over the ADVANCE Network. Results from different locations within the ADVANCE Network under different incident scenarios are analyzed in detail. Although not yet fully validated, this model is able to predict time-dependent traffic characteristics for a large-scale traffic network which are reasonable and internally consistent. This is the largest dynamic route choice solution which has been obtained thus far. Conclusions and recommendations for future research are presented as well.
tr57.pdf