Blockchain and Multi-Agent Learning Empowered Incentive IRS Resource Scheduling for Intelligent Reconfigurable Networks
IEEE/ACM Transactions on Networking
As a promising technology, intelligent reflecting surface (IRS) enables future communications and networks to realize programmable data transmissions. Due to the untrustworthiness of the communication environment and the selfishness of wireless devices, secure and intelligent IRS resource management is still an open issue. In this paper, we aim to implement IRS resource scheduling with properties of security, intelligence, efficiency, and fairness. To realize the above goals, we propose the blockchain and multi-agent learning empowered incentive scheduling system for tamper-proof and undeniable IRS resource management. To overcome the low throughout and intensive computation issues of blockchain, we devise a hybrid framework combining traditional Satoshi-style and directed acyclic graph blockchain for IRS resource scheduling. Due to the storage limitation of wireless devices, an intelligent blockchain storage reduction mechanism is proposed, where a multi-dimensional multi-hierarchy feature-based scheme is designed to determine block storage priority. Based on this storage priority and device states, the selection of storage-reduction devices is formulated as a cooperative multi-agent decision problem. Then, a multi-agent deep reinforcement learning-driven scheme is proposed to determine reduction strategies. To facilitate IRS providers/subscribers participating in the proposed system and maintain the efficiency of resource scheduling, an auction-based incentive mechanism is devised. In this mechanism, we propose the IRS resource allocation scheme and the payment scheme to achieve economic robustness and high efficiency. Finally, security analysis and experiment analysis indicate the feasibility and effectiveness of the proposed IRS resource scheduling in intelligent reconfigurable networks.
Blockchain, Blockchains, Communication system security, Data communication, incentive mechanism, IRS resource management, Job shop scheduling, multi-agent deep reinforcement learning, Resource management, Throughput, Wireless communication
Q. Pan, J. Wu, J. Li, W. Yang and M. Guizani, "Blockchain and Multi-Agent Learning Empowered Incentive IRS Resource Scheduling for Intelligent Reconfigurable Networks," in IEEE/ACM Transactions on Networking, doi: 10.1109/TNET.2023.3309729