Data-driven Transportation Operation and Management

The traffic management problem can be categorized into off-line traffic management and real-time traffic management. The off-line traffic management models are used to generate and evaluate long-term policies and regulations for a regional network. For example, the effects of congestion pricing in a city can be virtually evaluated using the off-line traffic management model and the fine-tuned management model is capable of optimizing the pricing strategies automatically. The off-line traffic management models are commonly used in the following areas: infrastructure retrofit/improvement plan, bus route design, signal design and facility design.

Real-time traffic management is a systematic way to control the traffic under planned and unplanned traffic events. The real-time traffic management system takes real-time traffic data feeds as input for self-correction, and controls the traffic flows through signals, ramp metering, dynamic message signs (DMS), information provision and so on. Among different real-time traffic management systems, real-time dynamic traffic assignment (real-time DTA) is one of the cutting-edge models to estimate, optimize, predict region-wide traffic conditions. Real-time DTA enables us to infer overall network performance from partial observations of the network conditions. It can also accommodate multiple data sources for network prediction and control.

The traffic management can be decomposed into three steps:

  1. Sensing: multi-source data can be obtained from the physical infrastructures using sensing technologies.
  2. Learning: a data-driven mobility model can be learned from the data and the mobility model can be used to estimate and predict the dynamic evolutions of the mobility system.
  3. Optimizing: with the fine-tuned network models, optimal management strategies can be generated to achieve better system performance. The whole framework requires a comprehensive network model encapsulating the physical mobility system, traffic information, network models and management strategies, while there have been few related studies to date.
An overview of the data-driven transportation management framework.