AI-Driven Adaptive Task Offloading in Fog–Cloud Architecture for Smart Healthcare Monitoring

Abstract

The rapid evolution of the Internet of Things (IoT) technology makes smart healthcare monitoring systems, in which patient health data is constantly collected with the help of wearable sensors and medical equipment. Nevertheless, transmitting large volumes of high latency medical information directly to remote cloud-computers can often lead to energy usage and communication latency. These problems are addressed by fog computing through decreasing the network congestion, improving system performance, and moving the processing of the data near the IoT devices. This paper presents an AI-based adaptive offloading of tasks infrastructure with a reinforcementbased scheduling system in the healthcare monitoring field. The proposed approach involves a Q-learning algorithm to dynamically decide whether cloud servers or fog nodes should process healthcare tasks. The decision-making is made considering realtime system components such as network latency, node energy availability, CPU requirements, and job deadlines. Experimental testing based on a healthcare IoMT dataset containing 1000 task instances reveals better performance with approximately $\mathbf{1 5}-\mathbf{2 0}$ percent lower latency and 10-15 percent energy consumption.