Scholarlink

To encounter a progressive Internet of Things (IoT) system requirements, edge computing forwards dispensation energy as well as loading nearest to an edge network to reduce the power or energy consumption. Edge computing is progressively famous due to these benefits; however, it poses difficulties in effectively handling the resources. However, most existing approaches eliminate the flexibility and effectiveness impacted through the firm architecture of the industrial control system as well as ‘end-to-end’ edge computing network of IoT. Hence, this research proposes the EdgeAISim architecture named Epsilon Greedy Strategy-based Q Learning (EGS-QL) approach with edge computing is proposed for scheduling the tasks and the resource management. The EGS system balances the exploration as well as exploitation through randomly selecting actions with a small probability while mainly selecting actions with high expected rewards. This make sure the learning process doesn’t deteriorate in suboptimal solutions and adapts efficiently to dynamic data management tasks. The experimental results demonstrates that the proposed EGS-QL approach attains the better total power consumption of 995 W as compared to the existing method of QL with graph neural network (GNN) approach.

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

Epsilon Greedy Strategy-Based Q Learning for Data Management in Edge Computing-Based Internet of Things

  • 章节

  • 发表日期：

    2026年

查看全文
  • Pushpa Mohan,
  • K. S. Nandini Prasad,
  • K. Durga Bhavani,
  • R. Mohan Naik,
  • Alok Kumar Pani

摘要

To encounter a progressive Internet of Things (IoT) system requirements, edge computing forwards dispensation energy as well as loading nearest to an edge network to reduce the power or energy consumption. Edge computing is progressively famous due to these benefits; however, it poses difficulties in effectively handling the resources. However, most existing approaches eliminate the flexibility and effectiveness impacted through the firm architecture of the industrial control system as well as ‘end-to-end’ edge computing network of IoT. Hence, this research proposes the EdgeAISim architecture named Epsilon Greedy Strategy-based Q Learning (EGS-QL) approach with edge computing is proposed for scheduling the tasks and the resource management. The EGS system balances the exploration as well as exploitation through randomly selecting actions with a small probability while mainly selecting actions with high expected rewards. This make sure the learning process doesn’t deteriorate in suboptimal solutions and adapts efficiently to dynamic data management tasks. The experimental results demonstrates that the proposed EGS-QL approach attains the better total power consumption of 995 W as compared to the existing method of QL with graph neural network (GNN) approach.