<scp>NEXUS</scp> ‐Scheduler: A Context‐Aware Task Scheduling and Resource Optimization Framework for Real‐Time <scp>IoT</scp> Applications in Fog–Cloud Environments

ABSTRACT The rapid expansion of Internet of Things (IoT) ecosystems and the increasing demand for real‐time services impose significant challenges on task scheduling and resource optimization across fog–cloud infrastructures. Existing schedulers often struggle with fluctuating workloads, heterogeneous devices, and mobility, resulting in high latency, inefficient resource usage, and frequent service‐level agreement (SLA) violations. To address these limitations, this paper introduces NEXUS‐Scheduler, a unified, event‐triggered, and context‐aware scheduling framework designed for dynamic IoT–fog–cloud environments. The framework integrates five coordinated components: (i) sensor‐triggered task generation, which activates tasks only when operational thresholds such as temperature &gt; 35°C, CPU &gt; 85%, or AQI &lt; 50 are breached; (ii) Nexus‐WNet, a lightweight GRU‐based multivariate predictor for short‐horizon workload forecasting; (iii) Nexus‐RCA, a reinforcement learning–based allocator that selects optimal execution tiers (edge, fog, cloud) based on predicted load, mobility, and resource conditions; (iv) Nexus‐ERA, an elastic resource manager that dynamically provisions CPU, memory, and bandwidth according to task priority and SLA requirements; and (v) TELEO, a decentralized self‐optimization layer that continuously refines scheduling strategies using confidence‐weighted feedback and lifelong reflective memory. Comprehensive experiments on a SmartCity testbed demonstrate that NEXUS‐Scheduler reduces latency by 41%, minimizes energy consumption by 32%, improves resource utilization by 26%, and enhances SLA compliance by 12% compared to state‐of‐the‐art heuristic, metaheuristic, and deep learning schedulers. These results highlight the effectiveness of NEXUS‐Scheduler as a scalable, adaptive, and efficient solution for real‐time IoT–fog–cloud deployments.