A reinforcing learning driven game theory algorithm for the multi-objective frequency assignment problem
Fatma Laidoui, Fatima Benbouzid-Si Tayeb, Malika Bessedik · International Journal of Computers and Applications · 2025
Over the past two decades, reinforcement learning (RL) has garnered significant attention for optimizing diverse applications, particularly in wireless networks. Despite extensive research, a notable gap remains in addressing the multi-objective Frequency Assignment Problem (FAP) in cellular networks. This study proposes a novel approach integrating Q-Learning with game theory to tackle multi-objective FAP. The proposed framework treats interference and separation costs as competing objectives, modeled through a two-player game where each player optimizes one objective. Each agent employs an embedded Q-Learning algorithm to assign frequencies to TRXs through environmental interactions. The game progresses over multiple episodes, with evaluated actions generating frequency assignment plans that balance the two objectives. Experimental validation on real-world instances from two US cities, Denver and Seattle, demonstrates promising results using Hypervolume and Nash equilibrium metrics, confirming the approach's efficacy.