基于赌博机模型的非时隙信道选择机制

A selection mechanism of un-slotted channel based on multi-armed bandit

针对未知信息环境网络中信道资源的选择与分配问题,提出了一种新的信道选择机制。借助于无休止多臂赌博机模型搭建网络系统模型,通过最大期望算法(EMA)实现了未知环境下对非时隙信道使用情况的初步学习,借助Q学习算法实现无休止多臂赌博机模型下的Gittins索引值的求解,同时确定出在一定干扰约束下的最优信道选择策略,最后通过借助拍卖机制实现系统内认知用户之间信道选择的冲突。经仿真实现验证,提出的新信道选择机制能够很好地避免认知用户对主用户的干扰,使系统中的信道得到高效利用,系统通信量得到大幅提高。

A new channel selection mechanism was proposed for the problem that how to select and distribute the channels under the unknown environment.Use the restless multi-armed bandit model to build the network system.Then,learning the usage of the channels preliminary by the expectation-maximization algorithm under the unknown environment,and later,achieve the Gittins index of restless multi-armed bandit by using the Q learning.In the meantime,then,obtained the optimal policy of channels selection under the certain interference constraints.Last,this paper used the multi-bid auction to deal with the collision among the users.Finally,the simulation results demonstrate that,the new mechanism can be good to avoid the interference to the primary user,to make the usage of channels efficiently and to improve the traffic of the system greatly.