International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 11 | Nov 2025
p-ISSN: 2395-0072
www.irjet.net
EDGE AI FOR REAL TIME DECISION MAKING IN IOT NETWORKS Renil Joy Department of Computer Application, St Thomas (Autonomous) College Thrissur 680001, Kerala, India -----------------------------------------------------------------------***-------------------------------------------------------------------The convergence of edge computing and artificial intelligence is fundamentally reshaping how smart devices process data in real-time, addressing critical limitations of traditional cloud architectures. reveals how next-generation Edge AI systems leverage cuttingedge techniques Including compressed neural networks, distributed learning paradigms, and precision-aware quantization to achieve remarkable efficiency gains. These innovations enable processing at the network periphery, delivering 45% faster response times and 30% greater bandwidth conservation in mission-critical domains like remote patient diagnostics and predictive industrial maintenance. The strategic implementation of block chain-enhanced security frameworks further fortifies these systems against emerging threats while preserving data sovereignty. [2]
Abstract - The integration of Edge AI and IoT networks has revolutionized real-time decision-making by enabling low-latency analytics and localized data processing. Advantages of Edge AI, Including reduced latency bandwidth savings, and enhanced privacy, across domains like smart cities and healthcare. Key techniques such as lightweight deep learning models (e.g.: MobileNet, EfficientNet), model compression (pruning, quantization), and reinforcement learning are explored, alongside architectures like Device-Edge-Cloud and Collaborative Edge. Challenges such as resource constraints, heterogeneity, and security are addressed, with future directions focusing on ultra-lightweight models and federated learning. Keywords: Intelligent edge computing, neural ,smart devices, IoT, Ultra-light weight models, Edge AI.
The convergence illuminates the powerful between the distributed intelligence and next-generation IoT security. By co-locating AI processing with data generation points, these systems achieve microsecond-level responsiveness crucial for autonomous navigation and urban digital twins. The hierarchical edge-fog-cloud framework introduces intelligent data routing complemented by military-grade encryption and self-learning threat detection systems. Field trials demonstrate near-perfect operational accuracy while maintaining robust defense against sophisticated cyber attacks. Yet the relentless growth of computational requirements and increasingly complex threat landscapes necessitate ongoing innovation in energy-aware algorithms and dynamic security protocols.[3] These collective findings underscore Edge AI's paradigm-shifting potential in enabling responsive, secure, and scalable intelligent systems [3]
1. INTRODUCTION The explosive growth of Internet of Things (IoT) devices, there has been a corresponding surge in data generation at the network edge. However, traditional cloud-centric models often struggle to meet the requirements of modern IoT applications, particularly in terms of latency, bandwidth efficiency, and data privacy. To overcome these challenges, Edge AI has emerged as a promising solution—bringing artificial intelligence algorithms directly to edge devices for localized processing. [1]
2. LITERATURE REVIEW The IoT revolution has catalyzed unprecedented advancements in edge intelligence, with research demonstrating how contemporary Edge AI ecosystems achieve near-instantaneous decision-
Further the technological innovations propelling Edge AI forward, including compact machine learning models, decentralized federated learning, and the deployment of 5G networks. A structured classification of Edge AI applications, functional capabilities, and underlying technologies is presented, emphasizing its role in connecting cloud infrastructure with IoT devices. The study also outlines unresolved issues, such as the demand for better resource allocation, standardization across platforms, and adaptable system designs. [4] A novel hybrid Edge-Cloud AI framework is introduced, which intelligently distributes computational tasks between edge nodes and cloud servers, optimizing both
making. Through the fusion of intelligent endpoints, adaptive gateways, and optimized inference models, these architectures slash processing delays by an order of magnitude compared to conventional cloud approaches. Collaborative learning techniques maintain strict data confidentiality by enabling knowledge sharing without raw data exchange. However, the path to ubiquitous adoption remains obstructed by hardware limitations, ecosystem fragmentation, and evolving security risks - challenges demanding breakthroughs in ultra-efficient neural architectures and universal interoperability standards. [1]
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 347