Loading...

Low-Latency Cloud Gaming Using Task Offloading and Resource Allocation in Mobile Edge Networks

Ghorbanny, Behzad | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 58190 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Bayat Sarmadi, Siavash
  7. Abstract:
  8. The video game industry has turned into one of the most remunerative entertainment industries amongst multimedia applications. The development of cloud-based services in recent years, has also turned cloud gaming in the form of ``gaming as a service'' into one of the most intriguing novel applications. In previous works, it has been shown that the employment of servers closer to the users in providing cloud gaming services could greatly enhance users' quality of experience, reduce network core congestion, and improve quality of service metrics. In this research, task offloading and resource allocation optimization for providing cloud gaming using edge computing has been investigated in newer generations of mobile networks. To the best of our knowledge, few studies have touched task offloading to edge servers optimization for the video gaming applications in the past and even fewer have employed reinforcement learning approaches to this problem. None of these works present a model, close to real world problem constraints and frameworks, especially when it comes to bandwidth and the requirements of quality of experience. In this work, after investigating resource constraints and the framework for providing cloud gaming on edge nodes in newer generations of mobile networks, a fair model, taking into account user mobility and service migration has been proposed to optimize task offloading and resource allocation, with a focus on delay reduction, users' quality of experience enhancement and increasing computation resource utilization. Due to the complexity of the modeled optimization problem, an MDP model, solvable by a reinforcement learning approach, has been proposed and optimized. To this end, the proper reinforcement learning algorithm has been suggested and sound definition of the state space, the action space, and the reward function has been proposed regarding the problem model. Evaluations indicate that under stable conditions, the proposed solution results in roughly 17% reduction in task execution latency, approximately 18% increase in users' quality of experience, roughly 42% better network computation resource utilization, and approximately 5% better fairness in the offloading decision compared to some baseline algorithms and the most relevant previous study
  9. Keywords:
  10. Edge Computing ; Task Offloading ; Resources Allocation ; Reinforcement Learning ; Cloud Gaming ; Mobile Edge Computing

 Digital Object List

 Bookmark

...see more