Queue-Aware Deep Reinforcement Learning for Heterogeneous Edge--Fog--Cloud Healthcare Monitoring

In order to facilitate real-time healthcare monitoring and emergency response, physiological data generated by wearable devices and Internet of Things (IoT) sensors must be processed as quickly as possible. However, Edge–Fog–Cloud healthcare infrastructures have serious problems with latency, energy consumption, and resource usage due to rising traffic volumes and dynamic workload conditions. The inability of traditional static and threshold-based offloading techniques to adjust to shifting network conditions frequently leads to congestion, delayed alarms, and ineffective resource allocation. This research suggests a Queue-Aware Edge–Fog–Cloud framework that combines two intelligent routing techniques to overcome these issues. The first approach, QueueAware (QA), uses latency-aware heuristics and lightweight queue use to dynamically route healthcare jobs. The second strategy, DDQNQueueAware (DDQNQA), combines queue-awareness with Double Deep Q-Learning to optimize discrete routing decisions based on queue conditions, latency, energy consumption, packet size, and system congestion. Numerous workload levels and queue-capacity combinations were used in extensive simulations. With maximum observed reductions of up to 40% in mean alert latency, up to 40% in p95/p99 latency, up to 64% in energy usage, and up to 46% in makespan relative to the evaluated baselines, DDQNQA achieved lower mean and tail latency in the majority of tested cases across the evaluated workload levels and queue-capacity configurations. Additionally, DDQNQA maintained fault-tolerance performance and competitive trust. These results show the possibility of integrating queue awareness and Double Deep Q-Learning for latency-sensitive healthcare routing under the simulated Edge-Fog-Cloud situations taken into consideration in this work.