Scholarlink

<p>The deployment of massive Internet of Things (IoT) networks has further heightened demand for real-time analytics, autonomous decision-making, and ultra-low-latency processing. Traditional cloud-centric systems suffer from delays in communication, network overload, and a lack of flexibility in dynamic systems such as industrial automation, smart healthcare, and connected mobility. The main problem is to achieve scalable, smart, distributed processing while remaining reliable, efficient, and coordinating across heterogeneous nodes in the IoT. The paper aims to solve this issue by introducing ELARA (Edge-Level Adaptive Reasoning Architecture)- a hybrid intelligent infrastructure that will maximize the placement of computation and accelerate real-time inference. The methodology incorporates: (i) Federated Feature Extraction to reduce transmission of raw data, (ii) Adaptive Task Offloading Engine based on workload-sensitive scheduling, and (iii) Reinforcement-based Resource Orchestration based on predictive scaling, elasticity, and edge-node collaboration. Also, ELARA uses a consensus-based micro-cluster architecture to support fault tolerance and distributed learning without a centralized cloud. Experimental results indicate that ELARA delivers 39–52 ms end-to-end latency, saves up to 48% Bandwidth, 31% energy, and 850–932 samples/sec inferences, which is better than current models, including CEA, DSS, and DFJSP. It also provides higher task completion rates (93–98%), higher offloading accuracy (89–93%), and much faster recovery from failures. To sum up, ELARA creates a scalable framework for intelligent, low-latency, and robust edge-based IoT systems that can be used in emerging autonomous applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A low-latency intelligent edge computing architecture for real-time decision-making in large-scale IoT networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月18日

查看全文
  • Yanming Jia

摘要

The deployment of massive Internet of Things (IoT) networks has further heightened demand for real-time analytics, autonomous decision-making, and ultra-low-latency processing. Traditional cloud-centric systems suffer from delays in communication, network overload, and a lack of flexibility in dynamic systems such as industrial automation, smart healthcare, and connected mobility. The main problem is to achieve scalable, smart, distributed processing while remaining reliable, efficient, and coordinating across heterogeneous nodes in the IoT. The paper aims to solve this issue by introducing ELARA (Edge-Level Adaptive Reasoning Architecture)- a hybrid intelligent infrastructure that will maximize the placement of computation and accelerate real-time inference. The methodology incorporates: (i) Federated Feature Extraction to reduce transmission of raw data, (ii) Adaptive Task Offloading Engine based on workload-sensitive scheduling, and (iii) Reinforcement-based Resource Orchestration based on predictive scaling, elasticity, and edge-node collaboration. Also, ELARA uses a consensus-based micro-cluster architecture to support fault tolerance and distributed learning without a centralized cloud. Experimental results indicate that ELARA delivers 39–52 ms end-to-end latency, saves up to 48% Bandwidth, 31% energy, and 850–932 samples/sec inferences, which is better than current models, including CEA, DSS, and DFJSP. It also provides higher task completion rates (93–98%), higher offloading accuracy (89–93%), and much faster recovery from failures. To sum up, ELARA creates a scalable framework for intelligent, low-latency, and robust edge-based IoT systems that can be used in emerging autonomous applications.