摘要： 媒体访问控制(MAC)协议负责协调所有认知用户的空闲信道接入服务，是认知Ad-hoc网络支持服务质量(QoS)的关键技术之一。在二进制指数退避算法基础上，提出一种支持服务区分的多智能体Q学习MAC算法。实时调整传输概率，使系统信道接入服务达到最优，建立传输概率调节的Markov链模型，导出分组的传输概率与协议参数的关系，给出基于服务区分的信道吞吐率模型，建立基于MAC协议参数学习的多智能体Q学习算法。实验结果表明，该算法能满足高优先级业务的QoS，且吞吐率和时延性能优于IEEE 802.11e EDCA机制。

关键词: 服务区分, Q学习, 多智能体, 认知无线电, 媒体访问控制, 传输概率

Abstract: The Media Access Control(MAC) protocol which coordinates idle channel accessing service among cognitive radio users, is one of the key technologies in cognitive Ad-hoc network supporting Quality of Service(QoS). This paper is based on the algorithm of binary exponential backoff, proposes a MAC algorithm with multi-agent Q-learning that supports service differentiation, real-time transmission probability adjusting is introduced to optimum the accessing service of system. Markov chain model is established with transmit probability adjusting, and the relationship between packet transmit probability and protocol parameters are derived. The channel throughput model with service differentiation is given. The multi-agent Q-learning architecture based on MAC protocol parameter is realized. Experimental result shows that this algorithm can satisfy the QoS of high priority service, the throughput rate and delay are better than IEEE 802.11e EDCA mechanism.

Key words: service differentiation, Q-learning, multi-agent, Cognitive Radio(CR), Media Access Control(MAC), transmission probability

中图分类号: 

• TN914