Abstract

The proliferation of the Internet of Things (IoT) has revolutionized the way devices interact, share data, and respond to real-time stimuli. Embedded IoT systems offer low-power, high-efficiency solutions for a variety of domains such as smart homes, healthcare, agriculture, and industrial automation. Despite advancements, achieving seamless connectivity and real-time intelligence in embedded IoT remains challenging due to limited computational power, energy constraints, and fragmented communication protocols. These limitations hinder performance, scalability, and responsiveness in real-world deployments. This research proposes a hybrid edge-cloud framework utilizing lightweight embedded devices integrated with optimized firmware for real-time processing, adaptive sensor data management, and low-latency communication. The method leverages MQTT protocol for lightweight messaging and integrates TinyML models on microcontrollers for localized intelligence, reducing reliance on centralized cloud services. The proposed system was tested using simulations in MATLAB and real-world deployments using Raspberry Pi 4 and ESP32 devices. Compared with existing models (CoAP-based IoT, MQTT without ML, Edge-Only, and Cloud-Only), the hybrid framework improved latency by 35%, energy efficiency by 27%, and inference speed by 42%, with minimal compromise on accuracy. The results validate the model’s scalability, responsiveness, and real-time intelligence for embedded IoT environments.

Authors

V.S. Anooja¹, M. Krishnakumar², S. Brilly Sangeetha³, Winson Rajaian⁴
Vidya Academy of Science and Technology, India^1,2, IES College of Engineering, India³, University of Technology and Applied Sciences, Sultanate of Oman⁴

Keywords

Embedded IoT, Edge Intelligence, Real-Time Communication, TinyML, MQTT