|
Abstract:
|
An transportation network is complex and stochastic with high degrees of unpredictability and uncertainty. Modeling of traffic dynamics has been an ongoing theme for decades. In this paper, we put forward a novel approach for modeling large-scale urban traffic dynamics using the Markovian framework and data from Google Maps. In particular, we introduce a statistically sound method of estimating the unknown model parameters and considering incoming and outgoing traffic streams to the transportation network. Maximum likelihood (ML) is used for the estimation. The Markov chain transition matrix can be fully computed after obtaining estimates for traffic links' travel times and turning probabilities at intersections. Based on probabilistic convergence theory, the results from this work can provide accurate estimates for traffic links' travel times. After combining the ML travel time estimates, turning probability estimates based on relevant data, and incoming and outgoing traffic flow modeling, we come up with a Markov chain model to mimic and simulate the dynamic network traffic. We also provide a case study in downtown Baltimore to illustrate the approach and validate the model.
|
