Scholarlink

<p>Effective task offloading strategies have become increasingly important for enhancing the processing speed of IoT devices. Different reinforcement-related approaches were implemented in the previous analyses for task offloading but it has significant challenges such as delays in real-time decision-making, increased latency, and unauthorized access to sensitive patient information. To tackle these limitations, this paper introduces a Novel Deep Reinforcement Learning-based Task Offloading (Novel DRL-TO) strategy. Here, a multi-layer edge computing architecture integrates reinforcement learning to optimize task offloading in IoT environments. It consists of the device layer, where IoT devices generate tasks; the edge layer, which processes tasks locally using edge servers; and the cloud layer, where tasks are forwarded when edge resources are insufficient. The edge-cloud broker manages task distribution, and a task scheduler allocates tasks to edge servers based on load and resource availability. To improve offloading efficiency, an actor-critic-based RL agent learns the optimal task allocation strategy by interacting with the environment and adjusting decisions through continuous feedback using a replay buffer. This system significantly reduces latency and enhances resource utilization by offloading computational tasks dynamically between edge and cloud layers, while the RL agent optimizes the task distribution in real-time. The comprehensive researches are conducted for efficient offloading and attained better performances in all measures. The proposed task offloading model provides a better resource utilization rate of 70%, task rejection rate of 68.3%, task completion rate of 93.5%, and throughput of 350 kbps respectively. The trade-off between energy consumption and delay was effectively managed, highlighting the algorithm's suitability for real-time healthcare applications. The proposed solution provides a robust and scalable framework for handling complex healthcare tasks in cloud-edge environments.</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 novel task offloading model for IoT: enhancing resource utilization with actor-critic-based reinforcement learning

  • 期刊论文

  • 发表日期：

    2025年2月17日

查看全文
  • Saranya G,
  • Kumaran K,
  • Vivekanandan M

摘要

Effective task offloading strategies have become increasingly important for enhancing the processing speed of IoT devices. Different reinforcement-related approaches were implemented in the previous analyses for task offloading but it has significant challenges such as delays in real-time decision-making, increased latency, and unauthorized access to sensitive patient information. To tackle these limitations, this paper introduces a Novel Deep Reinforcement Learning-based Task Offloading (Novel DRL-TO) strategy. Here, a multi-layer edge computing architecture integrates reinforcement learning to optimize task offloading in IoT environments. It consists of the device layer, where IoT devices generate tasks; the edge layer, which processes tasks locally using edge servers; and the cloud layer, where tasks are forwarded when edge resources are insufficient. The edge-cloud broker manages task distribution, and a task scheduler allocates tasks to edge servers based on load and resource availability. To improve offloading efficiency, an actor-critic-based RL agent learns the optimal task allocation strategy by interacting with the environment and adjusting decisions through continuous feedback using a replay buffer. This system significantly reduces latency and enhances resource utilization by offloading computational tasks dynamically between edge and cloud layers, while the RL agent optimizes the task distribution in real-time. The comprehensive researches are conducted for efficient offloading and attained better performances in all measures. The proposed task offloading model provides a better resource utilization rate of 70%, task rejection rate of 68.3%, task completion rate of 93.5%, and throughput of 350 kbps respectively. The trade-off between energy consumption and delay was effectively managed, highlighting the algorithm's suitability for real-time healthcare applications. The proposed solution provides a robust and scalable framework for handling complex healthcare tasks in cloud-edge environments.