Travel Time Information on Signalized Arterials

Area of Science:

• Transportation Engineering
• Data Science
• Traffic Management

Background:

• Travel time information is vital for daily commuting, preventing delays and utility loss.
• Existing methods for collecting travel time data using vehicle identification transponders require robust processing techniques.
• Effective travel time information necessitates accurate outlier filtering and reliable travel time prediction.

Purpose of the Study:

• To develop advanced algorithms for outlier filtering and travel time prediction in transportation networks.
• To improve the accuracy and reliability of travel time information systems.
• To quantify the economic benefits of enhanced travel time prediction.

Main Methods:

• Developed a median-based confidence interval algorithm for outlier filtering, tailored for suburban arterials.
• Integrated Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNNs) into an LSTM-CNN model for travel time prediction.
• Utilized vehicle identification data from dedicated short-range communication transponders.

Main Results:

• The proposed outlier filtering algorithm effectively handles travel time data with frequent entry/exit points.
• The LSTM-CNN model accurately captures both long-term trends and local patterns in travel time data.
• A 2.2% reduction in error rates was achieved under congested conditions compared to current practices, with potential annual savings of USD 135,200 at a 4 km study site.

Conclusions:

• The developed algorithms significantly enhance the accuracy of travel time information systems.
• Accurate travel time prediction leads to substantial economic benefits by reducing delays and improving traveler utility.
• The proposed LSTM-CNN approach offers a promising solution for real-time traffic data processing and prediction.