摘要
利用马尔可夫决策过程模型对传感器网络重构决策问题进行建模,提出了一种规则推理和强化学习相结合的动态应用重构决策方法.以能量约束和环境自适应性作为学习目标,设计了一个基于Q-学习的重构决策算法,使重构决策能够适应环境条件的变化.仿真结果表明基于强化学习的动态决策可以使传感器节点在运行过程中不断学习其所部署环境中异常事件发生的规律,自适应地调整节点上的应用,达到以较小的能耗获得较准确的监测效果的目标.
Reconfiguration decision making for sensor networks is modeled using Markov Decision Process,and a dynamic decision making scheme for application reconfiguration which combines rule-based reasoning with reinforcement learning is proposed.Aiming at energy constraint and environmental self-adaptation,a novel based Q-learning reconfiguration decision making （QLRDM） algorithm to make the reconfiguration decision adapt to environmental changes is designed.The simulations demonstrate that our dynamic decision making mechanism based on reinforcement learning can make sensor node continually learn the law of abnormal events in monitoring environment,and self-adaptively adjust applications running on sensor node,thus to achieve the accurate monitoring with small cost.
出处
《北京交通大学学报》
CAS
CSCD
北大核心
2010年第3期23-28,共6页
JOURNAL OF BEIJING JIAOTONG UNIVERSITY
关键词
无线传感器网络
应用重构
重构决策
强化学习
wireless sensor networks
application reconfiguration
reconfiguration decision making
reinforcement learning
