The rapid proliferation of LEO mega-constellations has intensified spectrum competition, necessitating decentralized allocation under Quality of Service (QoS) constraints. Existing methods often suffer from prohibitive complexity or a lack of formal stability. To tackle this, we model the resource competition as a non-cooperative game. While shown to lack the properties of an exact potential game, we propose the Service-Aware Regularized Potential Game (SARPG) framework. By integrating a strategy migration mechanism with stochastic search dynamics, we provide a formal convergence guarantee to a stable ε-Nash Equilibrium. Finally, experimental validation demonstrates its superior performance across multiple metrics in dynamic handover scenarios.
YANG et al. (Thu,) studied this question.