AI-enabled Priority and Auction-Based Spectrum Management for 6G

TL;DR

This work tackles spectrum scarcity in 5G/6G by ensuring minimum data-rate guarantees for vertical sector players (VSPs) while maximizing network throughput. It blends a modified Vickrey-Clarke-Groves auction with a deep reinforcement learning optimizer (DDPG) to perform per-frame, frame-based spectrum allocations, where VSPs are bid in a knapsack-like setting and ranked by QoS and bid truthfulness. The approach is formalized via a proportional fairness objective $\\mathbb{U}_f = \\sum_{i}(x_{i,f} r_{i,f} - r_i^{\\min})$ and solved with an actor-critic DRL framework that learns coefficients $c_{i,f}$ through states $s_f$, actions $a_f$, and rewards $R_f$. Simulation results show substantial improvements in spectrum utilization (about $85\%$ vs $35\%$ for greedy) and fairness as the number of spectrum blocks grows, demonstrating the method’s potential for scalable, QoS-aware spectrum management in 5G/6G and local spectrum sharing scenarios.

Abstract

In this paper, we present a quality of service (QoS)-aware priority-based spectrum management scheme to guarantee the minimum required bit rate of vertical sector players (VSPs) in the 5G and beyond generation, including the 6th generation (6G). VSPs are considered as spectrum leasers to optimize the overall spectrum efficiency of the network from the perspective of the mobile network operator (MNO) as the spectrum licensee and auctioneer. We exploit a modified Vickrey-Clarke-Groves (VCG) auction mechanism to allocate the spectrum to them where the QoS and the truthfulness of bidders are considered as two important parameters for prioritization of VSPs. The simulation is done with the help of deep deterministic policy gradient (DDPG) as a deep reinforcement learning (DRL)-based algorithm. Simulation results demonstrate that deploying the DDPG algorithm results in significant advantages. In particular, the efficiency of the proposed spectrum management scheme is about %85 compared to the %35 efficiency in traditional auction methods.

Figures (5)

• Figure 1: Integrating local high-quality networks (VSPs) into the MNO ecosystem in 6G network.
• Figure 2: The winning percentage of each VSP during the auctions.
• Figure 3: Mean reward per number of episodes.
• Figure 4: Spectrum efficiency for three methods.
• Figure 5: Fairness index per Number of VSPs.