An Experimental Urban Case Study with Various Data Sources and a Model for Traffic Estimation.

Area of Science:

• Transportation Engineering
• Data Science
• Urban Planning

Background:

• Reliable traffic state estimation is crucial for effective traffic management strategies.
• Data fusion from heterogeneous sensors (e.g., video, thermal, loop detectors) is necessary due to the infeasibility of uniform sensor deployment.
• Challenges in data fusion include varying sensor specifications, noise levels, and data heterogeneity.

Purpose of the Study:

• To assess the accuracy and robustness of different traffic sensors.
• To develop and evaluate a data fusion methodology for traffic state estimation.
• To compare a baseline model with a fused data model for improved traffic flow and travel time prediction.

Main Methods:

• Organized a video measurement campaign in an urban test area for ground truth data.
• Processed video data manually and using license plate recognition algorithms.
• Integrated data from thermal imaging cameras and Google Distance Matrix for sensor evaluation.
• Developed and compared baseline and fused multiple linear regression (MLR) models.

Main Results:

• Evaluated the accuracy and robustness of various sensors under different traffic conditions.
• Demonstrated that the proposed MLR model, fusing diverse data sources, significantly improves traffic state estimation accuracy.
• The fused model achieved high accuracy compared to the ground truth, validating the methodology's effectiveness.

Conclusions:

• The proposed data fusion methodology is efficient and robust for traffic state estimation.
• Integrating data from multiple sensor types enhances the reliability of traffic flow and travel time predictions.
• This approach provides a valuable tool for optimizing urban traffic management systems.