DRESS: Diffusion Model-Based Reward Shaping Scheme for Intelligent Networks
IEEE Transactions on Signal Processing · 2025 · 73 · 4285-4300
DOI: 10.1109/tsp.2025.3623239 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Signal Processing, 2025. Verfasst von Feiran You und 4 weiteren Autor:innen. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 51 Literaturangaben. Gefördert von: Research Grants Council of the Hong Kong Special Administrative Region, China, NSFC/RGC CRS, Areas of Excellence scheme.
Abstract (Originalsprache)
Network optimization remains fundamental in wireless communications, with Artificial Intelligence (AI)-based solutions gaining widespread adoption. As Sixth-Generation (6G) communication networks pursue full-scenario coverage, optimization in complex extreme environments presents unprecedented challenges. The dynamic nature of these environments, combined with physical constraints, makes it difficult for AI solutions such as Deep Reinforcement Learning (DRL) to obtain effective reward feedback for the training process. However, many existing DRL-based network optimization studies overlook this challenge through idealized environment settings. Inspired by the powerful capabilities of Generative AI (GenAI), especially diffusion models, in capturing complex latent distributions, we introduce a novel Diffusion Model-based Reward Shaping Scheme (DRESS) to achieve robust network optimization. By conditioning on observed environmental states and executed actions, DRESS leverages diffusion models’ multi-step denoising process to refine latent representations progressively, generating meaningful auxiliary reward signals that capture patterns of network systems. Moreover, DRESS is designed for seamless integration with any DRL framework, allowing DRESS-aided DRL (DRESSed-DRL) to enable stable and efficient DRL training even under extreme network environments. Experimental results demonstrate that DRESSed-DRL achieves about $1.5{\rm{x}}$ times faster convergence than its original version in sparse-reward wireless environments and significant performance improvements in multiple general DRL benchmark environments compared to baseline methods. The code of DRESS is available at https://github.com/NICE-HKU/DRESS.
Förderung
- Research Grants Council of the Hong Kong Special Administrative Region, China
- NSFC/RGC CRS
- Areas of Excellence scheme
- Collaborative Research Fund
- Shenzhen-Hong Kong-Macau Technology Research Programme
- National Natural Science Foundation of China
Zitieren
APA
You, F., Du, H., Hou, X., Ren, Y., & Huang, K. (2025). DRESS: Diffusion Model-Based Reward Shaping Scheme for Intelligent Networks. IEEE Transactions on Signal Processing, 73, 4285-4300. https://doi.org/10.1109/tsp.2025.3623239
Vancouver
You F, Du H, Hou X, Ren Y, Huang K. DRESS: Diffusion Model-Based Reward Shaping Scheme for Intelligent Networks. IEEE Transactions on Signal Processing. 2025;73:4285-4300. doi:10.1109/tsp.2025.3623239
BibTeX
@article{You2025,
author = {You, Feiran and Du, Hongyang and Hou, Xiangwang and Ren, Yong and Huang, Kaibin},
title = {DRESS: Diffusion Model-Based Reward Shaping Scheme for Intelligent Networks},
journal = {IEEE Transactions on Signal Processing},
year = {2025},
volume = {73},
pages = {4285--4300},
doi = {10.1109/tsp.2025.3623239}
}
EUVIMED