Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 55665 (05)
- University: Sharif University of Technology
- Department: Electrical Engineering
- Advisor(s): Gholampour, Iman; Haj Sadeghi, Khosrou
- Abstract:
- The use of IoT-based technologies is expanding in many areas today. The use of audio and video processing in IoT systems has been used as an alternative to human operators by increasing power and reducing processing costs. Due to the large volume of audio and video data and bandwidth limitations, complete data transfer to cloud processing servers is not cost-effective in terms of efficiency and energy consumption. As a result, the solution that has provided good results is to discharge these device tasks to the available clouds. In other words, the capacity of resources in the environment can be used to optimize the total latency of the system and energy consumption. In this dissertation, we consider a scenario in which tasks from different IoT devices are lined up in a queue, and existing clouds and the final server process a number of these time-sensitive tasks that have different arrival times and sizes. Their arrival is accidental in nature. These tasks can be performed locally on the device, or transferred to one of the near-edge servers or to a cloud data center in our infrastructure. We first formulate the problem of making optimal evacuation decisions that minimize energy and time consumption as a Markov decision process (MDP). We then developed the Advantage Actor Critic (A2C) algorithm, one of the deep modelless learning methods (DRL), to solve this problem. This algorithm optimizes the task dismissal policy by getting feedback from the environment to update and allocate resources. This method has been tested and compared with DRL, Full offload, RoundRobin and stochastic methods for several different modes of IoT device functions and in different cloud capacity situations. The proposed method has significantly improved compared to the other methods studied. On average, 24%, 14.25% and 11% compared to the three methods of roundrobin, stochastic and DRL, respectively, have improved in the amount of total cost (total weight of delay and total energy consumption).
- Keywords:
- Internet of Things ; Smart City ; Task Offloading ; Reinforcement Learning ; Audio Visual Processing ; Deep Reinforcement Learning ; Markov Decision Making
Digital Object List-
محتواي کتاب - view
Bookmark