Model-Driven Architecture with Feature Profiling for Scalable and Interoperable IoT Industrial Applications: A Real-Time Smart Building Validation Study
Main article
Abstract
The proliferation of heterogeneous IoT devices across industrial environments introduces critical scalability and interoperability challenges that inhibit large-scale deployment of real-time IoT applications. Existing solutions address either protocol interoperability or device scalability in isolation, failing to provide a unified framework for both. This paper proposes MDA-FP, a Model-Driven Architecture framework augmented with a novel Feature Profiling (FP) metamodeling technique that simultaneously addresses IoT scalability and cross-platform interoperability. The framework establishes a six-layer architecture spanning Computation Independent Models (CIM), Platform Independent Models (PIM), and Platform Specific Models (PSM), with automated model transformation rules that generate protocol-specific adapters for MQTT, CoAP, AMQP, and HTTP/REST target platforms. The Feature Profiling mechanism captures device capability signatures—bandwidth, latency tolerance, processing capacity, and power budget—and employs these profiles to guide transformation rule selection and protocol adapter configuration, ensuring generated implementations are optimally matched to device constraints. Evaluation on a smart building power consumption dataset comprising 60,215 instances demonstrates that MDA-FP achieves a mean classification delay of 5.23 ms (72% improvement over MQTT baseline), precision of 92.62%, sensitivity of 92.52%, specificity of 92.22%, MAE of 3.82%, and RMSE of 1.68%—outperforming four competing frameworks across all metrics. Scalability analysis confirms that MDA-FP maintains latency below 10 ms and reliability above 87.5% at 100 concurrent devices, with graceful degradation characteristics superior to protocol-specific baselines. The framework provides a principled pathway for deploying enterprise-scale IoT industrial applications without requiring per-device manual configuration or protocol-specific middleware expertise.
