Optimal Mean-Field Stackelberg Game-Inspired Resource Allocation for Scalable Multi-Modal Semantic Communications in 6G Networks

  • Do, Tung Son; 
  • Truong, Thanh Phung; 
  • Ho, Manh Cuong; 
  • Do, Quang Tuan; 
  • Dao, Nhu-Ngoc; 
  • ... Cho, Sungrae; 
  • 외 1명
Citations

WEB OF SCIENCE

0
Citations

SCOPUS

1

초록

Sixth-generation networks are expected to support large-scale multimodal semantic communication services under stringent spectrum constraints. Most current frameworks ignore semantic-level quality heterogeneity across modalities or depend on centralized optimization, which becomes computationally prohibitive when serving massive user populations. To address these limitations, this work proposes a mean-field Stackelberg equilibrium (MFSE) framework that applies the mean-field game theory and Stackelberg pricing to optimize scalable resource allocation and modality-aware semantic quality jointly. Acting as the Stackelberg leader, the base station sets modality-specific prices to maximize the network revenue and semantic welfare. Each user independently optimizes its per-modality bandwidth allocation against the mean-field aggregate demand, subject to minimum quality-of-service requirements. The leader pricing is optimized using a cosine-annealed Adam optimizer with Polyak price averaging and exact analytical gradients derived via the implicit function theorem. This work proves the existence and uniqueness of the MFSE, and the resulting strategy profile forms an approximate Nash equilibrium for the finite-player game. Simulations show that the proposed framework achieves 98.89% of the centralized semantic welfare and, under task-aware steering, attains the highest literal end-to-end task score. Furthermore, the equilibrium computation is O(1) in the user population, while distributed deployment scales as O(NJ) .

키워드

Resource management; Games; Pricing; Modeling; Bandwidth; Bismuth; Business intelligence; Speech; Educational institutions; Optimization; Mean-field game; MFG; multimodal; resource allocation; semantic communication; SemCom; Stackelberg game
제목
Optimal Mean-Field Stackelberg Game-Inspired Resource Allocation for Scalable Multi-Modal Semantic Communications in 6G Networks
저자
Do, Tung Son; Truong, Thanh Phung; Ho, Manh Cuong; Do, Quang Tuan; Dao, Nhu-Ngoc; Kim, Hyunbum; Cho, Sungrae
DOI
10.1109/TCOMM.2026.3729489
발행일
2026
유형
Article
저널명
IEEE Transactions on Communications
권
74
페이지
13640 ~ 13656