Automated Fault Tolerant Framework for IoT Composite Services in Remote Patient Monitoring

To reduce hospital overcrowding, remote patient monitoring systems have emerged as an effective solution by enabling continuous observation of patients in home environments. Leveraging the Internet of Things (IoT), these systems collect real-time physiological data using smart wearable devices and connected services. Together, these connected services constitute a composite service, by integrating multiple data sources and functionalities into a unified framework. However, the reliability of such systems is often compromised by IoT service failures, data corruption, or transmission loss, all of which may lead to critical clinical risks. To address this challenge, this paper presents an intelligent and automated fault-tolerant framework for monitoring Type 1 diabetic patients, based on a combination of data-driven and domain-driven approaches. The proposed system integrates two main components. A pretrained Bi-LSTM Autoencoder, designed to reconstruct missing multivariate physiological data by modelling temporal dependencies and cross-variable relationships. A Fuzzy Logic Controller, which analyses the Bi-LSTM Autoencoder output data using medical expert rules to assess insulin dosage and trigger real-time alerts in case additional insulin is potentially required. The proposed framework was tested on a real-world biomedical dataset containing patients’ physiological data. Experimental results demonstrate the system’s effectiveness in reconstructing missing data, and maintaining service continuity, thus ensuring resilience in IoT-based patient monitoring services.