A Game Theoretic Approach to Task Offloading for Multi-Data-Source Tasks in Mobile Edge Computing

Shuyu Chen,Shiyong Sun,Haopeng Chen,Jinteng Ruan,Ziming Wang
DOI: https://doi.org/10.1109/ISPA-BDCloud-SocialCom-SustainCom52081.2021.00111
2021-01-01
Abstract:Mobile edge computing(MEC) is becoming an appealing paradigm due to its low-latency and low-overhead computing services offered to users. In MEC scenarios, most user data are usually stored on multiple edge nodes for their privacy, large scale, and unstructured nature instead of being centralized stored in the cloud datacenter. Consequently, we focus on the multi-data-source tasks and propose a game-theoretic approach to task offloading for task accomplishment. We formulate the multi-data-source task offloading as a unified task offloading optimization problem. The traditional centralized optimal solutions have the issues of high complexity, data privacy, and poor performance. To solve these issues, we decompose the global optimization problem into sub-optimization problems which each edge node can solve. Furthermore, we construct a non-cooperative game-theoretic offloading framework and prove the existence of Nash equilibrium. We then design a distributed task offloading strategy to find the Nash equilibrium, which combines an optimal subtask partial offloading algorithm for each node. We conduct simulation experiments and demonstrate that our proposed approach has better convergence, lower algorithm overhead, and superior performance in task offloading than other typical algorithms.
What problem does this paper attempt to address?