To identify and reduce noise pollution, the Municipality of Koper keeps a close eye on noise levels throughout the entire city. InnoRenew has installed sound pressure sensors all across the municipality in order to accomplish this. These Linux- and Docker-capable sensors track noise levels in the surrounding environment continuously. The sensors record 5-second audio clips when noise levels surpass 70 dB, which are sent to a cloud instance for categorisation in order to identify the source of the excessive noise. In order to process and categorise these recordings, the current system mostly depends on cloud resources for storing, analyzing, and machine learning based noise classification. However, the main input to the classification model are the frequency components of a recording obtained by computing the Fast Fourier Transform (FFT). However, there are drawbacks to this strategy, such as possible privacy issues and the inefficiency of sending a significant volume of not processed audio data to the cloud.

Added value by Swarmchestrate: Swarmchestrate will significantly enhance the existing noise monitoring system by leveraging the computation power of the deployed sensors. The project aims to distribute the computational tasks of FFT and preliminary signal processing to the edge devices (sensors). By processing the audio signals on the sensors and transmitting only the precomputed data to the cloud, the system will mitigate privacy concerns, reduce data transmission requirements, and utilise the untapped computational resources of the edge devices. This distributed approach will not only enhance the efficiency of the noise monitoring system but also serve as a robust demonstration of Swarmchestrate’s capabilities in orchestrating containerised microservices across the cloud-to-edge continuum. Thus, the improved system will provide accurate real-time noise analysis, helping the Municipality of Koper manage noise pollution more effectively and maintain a quieter environment for its residents.