Loading...

Energy-Delay Aware Computation Offloading in Wireless Powered Mobile Edge Computing

Karami Fathabadi, Ahmad | 2025

143 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 58639 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Shah Mansouri, Hamed; Pakravan, Mohammad Reza
  7. Abstract:
  8. Wireless-powered mobile edge computing (WPMEC) is an emerging paradigm that provides computing resources and energy to resource-constrained Internet of Things (IoT) devices to extend their operational lifespan. Despite these advancements, the lack of mechanisms to efficiently leverage edge computing and wireless power transfer limits the full potential of IoT systems. In this study, we consider a battery-operated IoT device capable of harvesting energy and offloading computational tasks to an edge server. The device can either perform computation tasks locally or offload them to the network edge, where an edge processor executes the tasks and returns the results to the device. The primary objective of this study is to develop a method for reducing the number of tasks that fail to meet their deadlines due to resource limitations. To achieve this goal, we formulate an optimization problem aimed at minimizing the duration of each time slot, comprising both energy harvesting and computation time, while maximizing the stored energy in the device's battery. Given the highly dynamic nature of mobile edge computing environments and the stochastic arrival times and sizes of computation tasks, solving this problem using conventional optimization techniques is challenging. To address this challenge, we propose a deep reinforcement learning (DRL)-based algorithm that combines the feature extraction capabilities of deep neural networks with the continuous learning and decision-making improvements of reinforcement learning. Applying this approach to dynamic IoT environments enhances the decision-making processes. Furthermore, we introduce a dynamic weighting coefficient to balance the trade-off between reducing the computation time and increasing the device's battery energy level, utilizing convex optimization to minimize the time slot duration while maximizing the battery energy. We define a reward function to guide the learning process toward energy conservation and computation-delay reduction. Numerical experiments demonstrate that the proposed DRL-based algorithm reduces the number of tasks that miss their deadlines by approximately 30% compared with existing methods
  9. Keywords:
  10. Deep Reinforcement Learning ; Internet of Things ; Computation Offloading ; Mobile Edge Computing ; Wireless-Powered Mobile Edge Computing ; Energy Delay Product

 Digital Object List