The Internet of Things (IoT) is increasingly adopted across various domains. However, these sensors are prone to ageing, which can lead to failures in IoT systems due to inaccurate data collection. A possible solution to this challenge is IoT rejuvenation, a proactive and cost-effective technique designed to mitigate sensor ageing and ensure long-term data accuracy. This paper investigates the integration of Machine Learning (ML) in IoT rejuvenation, with the goal of developing a model capable of detecting measurement errors resulting from sensor ageing. The core idea is to identify these errors and either correct them or trigger recalibration when possible, ensuring continued accuracy and reliability in sensor performance. Such a solution is suitable for different contexts including smart hospitals, smart malls, smart cities, and so on. The study is conducted in a smart city scenario within a Cloud/Edge continuum environment, where intelligent street pole lamps equipped with low-cost ultrasonic sensors activate lighting upon vehicle detection. To evaluate our approach, we trained and compared various ML models for sensor ageing prediction, which can be deployed either on the Cloud or Edge. Experimental results demonstrate the effectiveness of our approach in accurately detecting and mitigating sensor degradation.
Towards Enabling IoT Rejuvenation Through Machine Learning on Cloud/Edge Continuum: A Study to Fight the Proximity Sensor Ageing on the Edge
Celesti A.;Lonia G.;Quattrocchi A.;Montanini R.;Villari M.;Fazio M.
2025-01-01
Abstract
The Internet of Things (IoT) is increasingly adopted across various domains. However, these sensors are prone to ageing, which can lead to failures in IoT systems due to inaccurate data collection. A possible solution to this challenge is IoT rejuvenation, a proactive and cost-effective technique designed to mitigate sensor ageing and ensure long-term data accuracy. This paper investigates the integration of Machine Learning (ML) in IoT rejuvenation, with the goal of developing a model capable of detecting measurement errors resulting from sensor ageing. The core idea is to identify these errors and either correct them or trigger recalibration when possible, ensuring continued accuracy and reliability in sensor performance. Such a solution is suitable for different contexts including smart hospitals, smart malls, smart cities, and so on. The study is conducted in a smart city scenario within a Cloud/Edge continuum environment, where intelligent street pole lamps equipped with low-cost ultrasonic sensors activate lighting upon vehicle detection. To evaluate our approach, we trained and compared various ML models for sensor ageing prediction, which can be deployed either on the Cloud or Edge. Experimental results demonstrate the effectiveness of our approach in accurately detecting and mitigating sensor degradation.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
