Robust Multisensor-based Framework for Efficient Road Information Services
Next-generation intelligent transportation systems of future road traffic monitoring will be required to provide reports of road conditions. Monitoring road surface conditions benefits drivers and the community. Road surface anomalies contribute to increased risk of traffic accidents, reduced driver comfort and increased wear of vehicles. Municipalities usually inspect the road surface quality through manual eye inspection that is inefficient, exhausting and consumes resources. Motivated by the growing demand for efficient road information services for connected and automated vehicles, this thesis proposes a comprehensive framework for monitoring road surface conditions. An intelligent and robust road surface monitoring system integrating vehicular and smart devices’ inertial sensors and GNSS receivers within and amongst vehicles is designed and developed in this research. To ensure rich data collections, land vehicles of different types are used in the road tests. A wavelet packet analysis was applied to the sensor measurements to separate the road anomalies’ signatures from both the vehicle motion dynamics and the sensor noise. Various feature extraction techniques were used to identify several types of road anomalies. The constructed features were utilized in building datasets of these road anomalies. The datasets were leveraged to build and assess the performance of a support vector machine classifier to detect and categorize multiple road anomalies of different severity levels. In order to overcome the GNSS receivers' challenges in geo-referencing the monitored anomalies,. In this framework, the vehicles’ onboard motion sensors are integrated with GNSS receivers to provide accurate, continuous, and adaptive resolution geo-referencing to the monitored anomalies. The road anomaly information is then utilized to provide assessment for the average quality of the road segments using a fuzzy inference system. The developed road segment quality database was then used by another fuzzy classifier that can be utilized in route planning to evaluate all possible routes and suggest the one of highest road quality. The future deployment of the system developed in this research will contribute to reducing congestion and accidents along the roads, enabling road labelling based on accessibility and conditions, providing personalized alerts and route recommendations, and improving road safety for all drivers.