COMPARATIVE ANALYSIS OF EVOLUTIONARY GAME AND MEAN-FIELD GAME APPROACHES FOR SCALABLE NODE SELECTION IN UAV-ASSISTED MEC NETWORKS

Abstract

To further enhance the node selection in unmanned aerial vehicles-assisted mobile edge computing systems, we proposed a comparative analysis between the Mean-Field Game and Evolutionary Game Theory. Because node selection is essential to optimize the allocation of computing load among unmanned aerial vehicles (UAVs) and macro base stations. First, to model the strategic development of users in a finite population, we employed Evolutionary Game Theory. And then, the Mean-Field Game framework is adopted as a scalable alternative to model the user behavior through aggregate mean field interactions to address the constraint of computational complexity and convergence instability of the Evolutionary Game Theory in a large number of users. To reflect the performance of the Mean Field Game approach, both approaches have been formulated and simulated under identical network conditions. Simulation results show that the MFG-based framework achieves lower computational cost, smoother convergence, and enhanced scalability compared to EGT. In particular, this paper demonstrates that MFG reduces computational complexity from O(K²) to O(KN) while maintaining comparable equilibrium performance.

Keywords: UAV-assisted MEC, Node selection, Mean-Field Game, Evolutionary game theory, Scalability, Load Balancing