Mobile Edge Computation Offloading Using Game Theory and Reinforcement Learning

Abstract

Due to the ever-increasing popularity of resourcehungry and delay-constrained mobile applications, the computation and storage capabilities of remote cloud has partially migrated towards the mobile edge, giving rise to the concept known as Mobile Edge Computing (MEC). While MEC servers enjoy the close proximity to the end-users to provide services at reduced latency and lower energy costs, they suffer from limitations in computational and radio resources, which calls for fair efficient resource management in the MEC servers. The problem is however challenging due to the ultra-high density, distributed nature, and intrinsic randomness of next generation wireless networks. In this article, we focus on the application of game theory and reinforcement learning for efficient distributed resource management in MEC, in particular, for computation offloading. We briefly review the cutting-edge research and discuss future challenges. Furthermore, we develop a gametheoretical model for energy-efficient distributed edge server activation and study several learning techniques. Numerical results are provided to illustrate the performance of these distributed learning techniques. Also, open research issues in the context of resource management in MEC servers are discussed.

Distributed Intelligence in Wireless Networks

“…RL techniques have been used as promising solutions to tackle this challenge based on the trial-and-error rule, where the RL agent, i.e., the user, can adjust its policy to achieve the best long-term goal according to the future reward feedback from the environment without prior knowledge of system models. In [108], [119], the authors investigated the dynamic computation offloading process and developed RL algorithms to learn the optimal offloading mechanism with the goal of minimizing latency and choosing the energyefficient edge server. Besides, the DRL algorithm has been proven to be more effective for enabling RL to handle large state spaces by leveraging the powerful DNNs to approximate state-action values, which is envisioned to solve complex sequential decision-making problems.…”

Network Threat Detection Using Machine/Deep Learning in SDN-Based Platforms: A Comprehensive Analysis of State-of-the-Art Solutions, Discussion, Challenges, and Future Research Direction

“…Optimal strategies by each player can be selected by an adaptive learning method known as RL. RL uses historical information for the selection of strategies, such as status of network, strategies, and utility of other players [ 67 , 68 , 69 ]. Thus, an effective technique of decision-making is RL-based game theory.…”

Correlation-Based Device Energy-Efficient Dynamic Multi-Task Offloading for Mobile Edge Computing

“…Through continuous testing, it rewards and punishes a series of actions and then modifies its strategy. After such continuous adjustments, the UE can learn which actions should be chosen in order to achieve the best return in certain situations, as in [165]- [167]. In addition, there are Q-learning [9], deep reinforcement learning (DRL) [168]- [170], and other techniques.…”