王祺尧,
冯辉,,
罗灵兵
1.复旦大学信息科学与工程学院 ??上海 ??200433
2.复旦大学智慧网络与系统研究中心 ??上海 ??200433
基金项目:国家自然科学基金(61501124),上海市公安局科学技术发展基金(2017012)
详细信息
作者简介:胡波:男,1968 年生,教授,研究方向为数字信号处理、数字通信和系统设计
王祺尧:男,1993 年生,硕士生,研究方向为传感器网络、强化学习、序贯决策等研究
冯辉:男,1980 年生,副教授,研究方向为分布式信号处理理论与应用
罗灵兵:男,1992 年生,硕士生,研究方向为图像处理
通讯作者:冯辉 hfeng@fudan.edu.cn
中图分类号:TP393; TP391计量
文章访问数:1896
HTML全文浏览量:433
PDF下载量:58
被引次数:0
出版历程
收稿日期:2017-12-06
修回日期:2018-05-04
网络出版日期:2018-07-12
刊出日期:2018-09-01
Adaptive Sensor Scheduling Algorithm for Target Tracking in Wireless Sensor Networks
Bo HU,Qiyao WANG,
Hui FENG,,
Lingbing LUO
1. School of Information Science and Technology, Fudan University, Shanghai 200433
2. Research Center of Smart Networks and Systems, Fudan University, Shanghai 200433
Funds:The National Natural Science Foundation of China (61501124), The Public Security Bureau Science and Technology Development Foundation of Shanghai (2017012)
摘要
摘要:在无线传感器网络目标跟踪的过程中进行节点调度,可以综合考虑跟踪误差和能量消耗,延长传感器网络的使用寿命。为了综合考虑节点调度的短期和长远损失,该文将问题建模为部分可观测马尔科夫决策过程(POMDP)以得到更优的调度策略,并提出一种近似求解算法C-QMDP。该算法利用马尔科夫链蒙特卡洛方法(MCMC)推导连续状态空间的置信状态的转移,并计算瞬时代价。使用状态离散化方法,基于马尔科夫决策过程(MDP)值迭代求解未来代价的近似值。仿真结果表明,相比现有POMDP近似算法,该文算法既可以降低跟踪过程中的累积损失,又可以将大量运算进行离线计算,减小了在线决策时的计算量。
关键词:无线传感器网络/
目标跟踪/
节点调度/
部分可观测马尔可夫决策过程
Abstract:In the process of target tracking, the sensor scheduling algorithm can achieve the tradeoff between the tracking error and the energy consumption so as to extend the service life of the sensor network. The issue can be modeled as a Partially Observable Markov Decision Process (POMDP), which takes both short- and long- term losses of sensor scheduling into account and makes a better decision. A C-QMDP approximation algorithm suitable for continuous state space is proposed. The Markov Chain Monte Carlo (MCMC) method is used to derive the transfer function of belief state and calculate the instantaneous cost. The state discretization method is used to solve the approximation of future cost based on Markov Decision Process (MDP) iteration. Simulation results show that compared to the existing POMDP approximation algorithms, the proposed algorithm can reduce the cumulative losses and computation load in the tracking process by offline computation.
Key words:Wireless Sensor Networks (WSN)/
Target tracking/
Sensor scheduling/
Partially Observable Markov Decision Process (POMDP)
PDF全文下载地址:
https://jeit.ac.cn/article/exportPdf?id=cae17cbe-5cfb-47f9-8153-a7a49aa7abb1