Rollback Approach for Demand Consistency Checking of Real-Time Traffic Network State Estimation Models

Abstract

The paper presents a real-time traffic network state estimation model with online demand consistency checking and updating capabilities. In contrast to reactive-based methodologies proposed in the literature, the model adopted a time rollback with a corrective actions approach. When an instance of inconsistency between the measured and estimated network state was observed, the model was allowed to roll back in time and promptly resimulated a predefined past period after the appropriate model's parameters were adjusted to minimize the observed inconsistency. A demand correction algorithm was developed and used for demand adjustment for each rollback period. The results of applying the developed model for a test bed network are presented. Results show that the approach improves the model's consistency with real-world observations.