摘要
研究了定向天线自组织网的邻居发现问题,提出了一种基于Q-Learning的邻居发现算法,并利用大型网络仿真软件OPNET对定向天线自组织网进行建模仿真。基于Q-Learning的邻居发现算法在不知道任何邻居节点先验信息的情况下,通过Q-Learning机制确定每次邻居扫描的收/发模式,并根据当次扫描结果确定回报值,学习扫描中的经验,能够达到提高邻居发现效率的目的。利用三级建模机制搭建了基于OPNET的仿真模型,并利用天线模型编辑器对定向天线进行了建模。基于OPNET的仿真结果表明,提出的算法相比于基于扫描的算法能够提高邻居发现的效率。
Neighbor discovery problem in directional antenna Ad Hoc networks is studied. A Q-Learning neighbor discovery algorithm is proposed, and a simulation model for directional antenna Ad Hoc networks based on OPNET software is setup. This algorithm makes a decision of transmitting and receiving mode for every neighbor scanning without the information of neighbor position through the Q-Learning mechanism. The reward value is determined according to the current scanning result, and the experiences are learned in order to improve the neighbor discovery efficiency. The simulation model based on OPNET software is constructed using three-layer modeling mechanism. Then the directional antenna is modeled by the antenna modeling editor. The simulation result based on OPNET shows that the proposed algorithm has a better efficiency of neighbor discovery compared with the scan based algorithm.
作者
李默
赵亮
Li Mo;Zhao Liang(Department of Electronic Technology, China Maritime Police Academy, Ningbo 315801, China;Division of Informationization, East Sea Fleet, Ningbo 316000, China)
出处
《系统仿真学报》
CAS
CSCD
北大核心
2018年第5期1707-1714,共8页
Journal of System Simulation
基金
国家自然科学基金(61300203)
宁波市自然科学基金(2016A610008)
