Visualizing Street Pavement Anomalies through Fog Computing V2I Networks and Machine Learning

Area of Science:

• Civil Engineering
• Computer Science
• Data Science

Background:

• Urban infrastructure maintenance often relies on reactive measures like citizen reports or incident responses.
• Current systems lack real-time capabilities for detecting pavement anomalies, leading to inefficient public spending on mobility.

Purpose of the Study:

• To develop and evaluate a real-time system for detecting and classifying road surface anomalies using vehicle-mounted sensors.
• To leverage fog computing and Vehicle-to-Infrastructure (V2I) networks for efficient data processing and anomaly detection.
• To compare the effectiveness of different Machine Learning Algorithms (MLA) for pavement condition analysis.

Main Methods:

• Utilizing accelerometry sensors in instrumented vehicles to capture road roughness data and establish a flat reference.
• Implementing a fog-computing architecture integrated with a V2I network for data transmission and processing.
• Applying supervised Machine Learning Algorithms, specifically Artificial Neural Networks and K-Nearest Neighbors, for anomaly detection and classification.

Main Results:

• The developed system successfully detects and classifies various road problems and abnormal pavement conditions.
• Comparison of MLA performance indicated suitability for analyzing acquired road data.
• The system provides a mechanism for visualizing street quality and mapping areas with significant anomalies.

Conclusions:

• The proposed system offers a proactive approach to urban infrastructure maintenance, enhancing decision-making for public spending on mobility.
• Real-time anomaly detection through vehicle-mounted sensors and fog computing is feasible and effective.
• This technology can significantly improve the management and upkeep of city streets and avenues.