Scholarlink

<p>The integration of Internet of Things (IoT) technology with video applications enables real-time monitoring and decision-making across various contexts. These applications utilize interconnected devices equipped with cameras, sensors, and communication capabilities to collect visual information. Despite their revolutionary impact on multiple industries, IoT video applications face significant challenges, such as high latency, energy consumption, and the complexity of managing large volumes of video data. Traditional Deep Learning (DL) models contribute to high communication latency, creating a need for distributed solutions. To address these issues, the Collaborative Edge Distributed Model Training Video (CoEDMTVideo) framework is introduced. This framework integrates collaborative edge computing and Deep Reinforcement Learning (DRL) for distributed model training in IoT video streaming environments. The Video Streaming System (VSS) within CoEDMTVideo utilizes the Optimal Ensemble Deep Q-learning Network (OEDQNet) for distributed model training in the IoT context. Initially, a multi-driven Collaborative Edge Computing (CEC) platform is deployed to enhance computing capacity and reduce network communication overhead. We further introduce a light-weight fuzzy decision layer that biases action selection toward QoS-aware choices, improving stability under dynamic bandwidth and latency conditions. This platform includes numerous IoT terminals (ITs), Collaborative Edge nodes (CENs), and a Central Server (CS). The OEDQNet algorithm improves distributed model training by leveraging an ensemble learning technique for robust and reliable decision-making and by combining ensemble learning with deep Q-learning to optimize system performance in dynamic IoT environments. Simulations using the YouTube dataset demonstrate that the proposed framework enhances video processing efficiency and reduces energy consumption. The simulation results show that the proposed CoEDMTVideo achieves significant improvements, demonstrating its overall effectiveness in IoT video 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

CoEDMTVideo: Collaborative Edge Computing and Optimal Ensemble Deep Q-Learning Network for Distributed Model Training in IoT Video Stream Environments

  • 期刊论文

  • 发表日期：

    2026年4月7日

查看全文
  • Shaharyar Alam Ansari,
  • Mohammad Luqman,
  • Mohd Anas Wajid,
  • Mohammad Saif Wajid,
  • Shahnawaz Ahmad

摘要

The integration of Internet of Things (IoT) technology with video applications enables real-time monitoring and decision-making across various contexts. These applications utilize interconnected devices equipped with cameras, sensors, and communication capabilities to collect visual information. Despite their revolutionary impact on multiple industries, IoT video applications face significant challenges, such as high latency, energy consumption, and the complexity of managing large volumes of video data. Traditional Deep Learning (DL) models contribute to high communication latency, creating a need for distributed solutions. To address these issues, the Collaborative Edge Distributed Model Training Video (CoEDMTVideo) framework is introduced. This framework integrates collaborative edge computing and Deep Reinforcement Learning (DRL) for distributed model training in IoT video streaming environments. The Video Streaming System (VSS) within CoEDMTVideo utilizes the Optimal Ensemble Deep Q-learning Network (OEDQNet) for distributed model training in the IoT context. Initially, a multi-driven Collaborative Edge Computing (CEC) platform is deployed to enhance computing capacity and reduce network communication overhead. We further introduce a light-weight fuzzy decision layer that biases action selection toward QoS-aware choices, improving stability under dynamic bandwidth and latency conditions. This platform includes numerous IoT terminals (ITs), Collaborative Edge nodes (CENs), and a Central Server (CS). The OEDQNet algorithm improves distributed model training by leveraging an ensemble learning technique for robust and reliable decision-making and by combining ensemble learning with deep Q-learning to optimize system performance in dynamic IoT environments. Simulations using the YouTube dataset demonstrate that the proposed framework enhances video processing efficiency and reduces energy consumption. The simulation results show that the proposed CoEDMTVideo achieves significant improvements, demonstrating its overall effectiveness in IoT video applications.