GPS Data and Machine Learning Tools, a Practical and Cost-Effective Combination for Estimating Light Vehicle...

Area of Science:

• Environmental Science
• Automotive Engineering
• Data Science

Background:

• Vehicle emissions significantly impact air quality and climate change.
• Accurate emissions data is crucial for regulatory compliance and environmental policy.
• Traditional emissions testing methods can be resource-intensive and may not fully represent real-world driving.

Purpose of the Study:

• To develop and validate a novel methodology for estimating real-world vehicle emissions using GPS and machine learning.
• To assess the emissions (CO2, CO, NOx, HC) of the most sold light vehicle categories (sedans, SUVs, pickups) in Ecuador.
• To compare the performance of the developed model against existing emissions estimation models.

Main Methods:

• Acquired driving data from best-selling vehicles in Ecuador using GPS data loggers and Portable Emissions Measurement Systems (PEMS).
• Employed K-means clustering and classification trees to estimate gears used during driving.
• Utilized random forest techniques to determine driving variable importance and Artificial Neural Networks (ANNs) to estimate emissions.
• Validated ANNs using data from a second standardized route and extensive random driving data (300-324 km per vehicle type).

Main Results:

• Developed ANNs accurately estimated CO2, CO, NOx, and HC emissions.
• The model demonstrated comparable results to the IVE model and an OBD-based model.
• The methodology proved effective without requiring lengthy PEMS-equipped test drives.
• The model showed robustness across diverse traffic conditions due to extensive random driving data training and validation.

Conclusions:

• The proposed GPS and machine learning methodology offers a reliable and efficient alternative for estimating real-world vehicle emissions.
• This approach reduces the need for extensive on-road testing with PEMS equipment.
• The findings provide valuable insights for understanding and mitigating emissions from popular light vehicle segments.