跨层认知无线Mesh网络中多agent学习的功率分配策略  

Multi-agent learning power allocation strategy in cross-layer cognitive wireless Mesh network

作　　者：李跃新[1] 杨岗[1] 

机构地区：[1]湖北大学计算机与信息工程学院,武汉430064

出　　处：《计算机应用研究》2015年第10期3101-3103,3113,共4页Application Research of Computers

基　　金：湖北省国际交流与合作项目(2012IHA0140)

摘　　要：为了改善无线Mesh网络的无线电频谱的利用效率以及提高能源利用效率,提出跨层认知无线Mesh网络中多agent学习的功率分配策略。首先,通过多agent智能学习方法,将网络状态定位为一个有限状态马尔可夫决策过程,在环境状态变化时,通过任务来学习一个优化策略,最大化网络的资源和能量的总期望收益。其次,为了更好地节省网络能量,通过一种奖励机制和非合作博弈模式正式确定功率的分配问题,在每个环境状态下学习到最佳的功率分配策略。最后,实验仿真结果表明,该算法在提高频谱利用效率和能源效率上具有明显的效果。In order to improve the utilization efficiency of radio spectrum for wireless Mesh network and improve energy efficiency,this paper proposed a cross-layer cognitive wireless Mesh network based on multi-agent learning. First,it positioned multi-agent learning power allocation strategy in cross-layer cognitive wireless Mesh networks as a finite-state Markov decision process,changed in the state of the environment,through the task to learn an optimal strategy to maximize network resources and energy of the total expected revenue. Secondly,in order to better save energy networks,it formalized the distribution of power through a none-cooperative game mode and incentive mechanism to learn optimal power allocation strategy in each state of the environment. Finally,the experimental results show that the algorithm has a significant effect in improving the efficiency of spectrum utilization and energy efficiency.

关 键 词：认知无线Mesh网络 多agent智能学习 奖励机制 非合作博弈模式 功率分配策略 

分 类 号：TP393.07[自动化与计算机技术—计算机应用技术]