Noise pollution monitoring at pedestrian level by autonomous vehicles in urban areas

This study presents a mobile noise monitoring system designed to assess pedestrian-level environmental noise using an autonomous ground vehicle (Yape) equipped with a calibrated sound level sensor. Unlike fixed monitoring stations or simulation-based models, our approach enables dynamic, high-resolution data collection along entire urban areas. Key challenges included ensuring that recorded noise accurately reflected pedestrian exposure and isolating the vehicle's own acoustic emissions from ambient noise. To address this, we collected training data by positioning a calibrated reference sensor at predetermined distances from the Yape and developed a machine learning model to estimate true environmental noise levels. The model achieved a high predictive accuracy (R2 = 0.94, RMSE = 1.3 dBA), demonstrating reliable separation of Yape-generated noise from environmental signals. The system was tested across three distinct streets in Genoa, Italy, and captured a wide range of urban acoustic profiles. This method offers a scalable solution for urban noise mapping and has potential applications in smart city planning, pedestrian comfort analysis, and real-time environmental monitoring.