A Bayesian Game Theoretic Approach to Task Offloading in Edge and Cloud Computing
Anna V. Guglielmi, Marco Levorato, Leonardo Badia · 2018
This paper addresses the management of computational offloading in a three-tier hierarchical architecture comprising mobile devices, edge computing, and cloud computing. Although edge and cloud devices have lower data processing time than mobile devices, the simultaneous transmission of heavy data streams may overload the local wireless network, resulting in an overall larger delay. Herein, a game theoretic framework is proposed for the distributed decision making in a scenario where mobile users share the same network resource and do not have a priori information in the wireless links. We evaluate at first the rational gameplay of the nodes in a scenario with complete knowledge, and we compare it with a scenario with incomplete information modeled as a Bayesian game. In particular, we consider network positions, and therefore channel gain and distance-related parameters, to be uniformly distributed within a given range, and the nodes only have this knowledge available as a prior. The analysis demonstrates that rationality (implying selfish behavior) of the mobile users does not necessarily lead to a more efficient allocation and actually the scenario of incomplete information leads to a socially better outcome, thereby suggesting an interesting guideline for the design of computational offloading strategies in realistic scenarios.