Mobile road weather sensor calibration by sensor fusion and linear mixed models

Area of Science:

• Environmental Science
• Sensor Technology
• Data Science

Background:

• Mobile road weather sensors offer high-frequency data but face reliability and accuracy challenges.
• Stationary road weather stations (RWS) provide reliable but sparse, low-frequency data.
• Discrepancies exist between mobile and stationary sensor observations, influenced by road weather conditions.

Purpose of the Study:

• To calibrate mobile road weather sensor observations using a linear mixed model.
• To improve the accuracy and reliability of mobile road weather sensor data.
• To enable direct integration of calibration into sensors for real-time application.

Main Methods:

• Utilized a linear mixed model for sensor calibration.
• Fused approximately 20,000 pairs of mobile and RWS observations from collocated, simultaneous measurements.
• Employed a rendezvous model for sensor calibration.

Main Results:

• The linear mixed model calibration significantly reduced the Mean Squared Error (MSE) between mobile and RWS sensor observations by nearly half.
• Statistical analysis confirmed road weather conditions impact mobile and stationary sensor data differences.
• The developed calibration method is computationally lightweight.

Conclusions:

• Calibration of mobile road weather sensors using a linear mixed model effectively enhances data accuracy.
• The method addresses the reliability gap between mobile and stationary sensors.
• Lightweight calibration is suitable for direct embedding in mobile sensors, improving real-time road weather monitoring.