Blockchain-Empowered Multi-Domain Resource Trading in TN-NTN 6G Networks: A Hierarchical Multi-Agent DRL Approach

Abegaz Mohammed Seid, Hayla Nahom Abishu, Rajkumar Singh Rathore, Aiman Erbad, Rutvij H. Jhaveri, Jianfeng Lu

Allbwn ymchwil: Cyfraniad at gyfnodolynErthygladolygiad gan gymheiriaid

Crynodeb

The advent of 6G networks promises transformative advances in communication technologies, ushering in an era of unprecedented connectivity, low-latency communication, and ubiquitous computing. However, one of the key challenges in realizing the full potential of 6G lies in efficiently managing and trading resources across diverse domains in terrestrial and non-terrestrial integrated networks (TN-NTN). In this paper, we introduce a novel multi-domain resource trading approach that integrates blockchain technology, hierarchical multi-agent deep reinforcement learning (HMADRL), and multi-leader multi-follower (MLMF) Stackelberg games to address this challenge. HMADRL provides an intelligent and adaptable framework for resource requesters and provider agents to learn optimal resource allocation methods in multiple domains. The MLMF Stackelberg game framework represents the strategic interactions between virtual service providers (SPs), network operators, and computing resource providers to enable them to make dynamic resource allocation and pricing decisions. This hierarchical decision-making strategy ensures a structured bargaining mechanism where leaders optimize their revenue by setting resource pricing and followers strategically decide on their resource-buying strategies. In the proposed scheme, the blockchain includes smart contracts that autonomously execute resource trading agreements, ensuring that transactions are tamper-proof and executed according to predefined rules. The proposed method optimizes the system’s utility by dynamically adjusting the pricing strategy to changing resource demands. The simulation results show that our proposed HMADRL algorithm increases the utility by 19.645%, 33.018%, and 48.791% compared to the modified multi-agent deep deterministic policy gradient (MADDPG), MADDPG, and DDPG algorithms, respectively.
Iaith wreiddiolSaesneg
Tudalennau (o-i)1-1
Nifer y tudalennau1
CyfnodolynIEEE Transactions on Consumer Electronics
Dynodwyr Gwrthrych Digidol (DOIs)
StatwsCyhoeddwyd - 9 Rhag 2024

Dyfynnu hyn