Fuel Prediction and Reduction in Public Transportation by Sensor Monitoring and Bayesian Networks

Area of Science:

• Environmental Science
• Transportation Engineering
• Data Science

Background:

• Public transportation systems generate vast amounts of data.
• Optimizing fuel consumption in urban fleets is crucial for environmental sustainability and operational efficiency.
• Controller Area Network (CAN-bus) systems offer a rich source of real-time vehicle data.

Purpose of the Study:

• To develop a data-driven approach for reducing fuel consumption in urban public transport.
• To identify key variables and causal relationships influencing fuel usage.
• To provide actionable insights for fleet managers and policymakers to improve air quality and public services.

Main Methods:

• Utilizing Controller Area Network (CAN-bus) data from city buses.
• Employing heuristic and exhaustive algorithms to construct Bayesian networks.
• Applying Granger causality for Bayesian network validation and ranking.
• Validating network structures using synthetic datasets with known ground truth.

Main Results:

• Successfully generated Bayesian networks representing relationships between monitored variables.
• Identified potential cause-effect relationships influencing fuel consumption.
• Demonstrated the validity of the proposed Bayesian network approach through Granger causality and synthetic data comparison.
• Achieved high agreement between discovered networks and ground truth relationships.

Conclusions:

• The proposed method effectively leverages CAN-bus data and Bayesian networks for fuel consumption analysis in public transport.
• Granger causality provides a robust validation and ranking mechanism for discovered causal relationships.
• The findings support informed decision-making for reducing fuel consumption, mitigating air pollution, and enhancing urban mobility.