WSNs中一种基于强化学习的跟踪调度算法  

A Tracking Scheduling Algorithm Based on Reinforcement Learning in WSNs

作　　者：李鑫[1] 刘杨 刘立业 LI Xin;LIU Yang;LIU Liye(Department of Electrical and Electronic Engineering,Shijiazhuang University of Applied Technology,Shijiazhuang 050081,China;The Information and Development Research Center of Shijiazhuang Health Commission,Shijiazhuang 050023,China)

机构地区：[1]石家庄职业技术学院电气与电子工程系,河北石家庄050081 [2]石家庄市卫生健康信息与发展研究中心,河北石家庄050023

出　　处：《无线电工程》2023年第5期1221-1227,共7页Radio Engineering

基　　金：河北省自然科学基金(F2021205004)。

摘　　要：目标跟踪是无线传感器网络(Wireless Sensor Networks, WSNs)中一项应用广泛的技术,旨在估计目标在监控区域内移动时的位置。为了探索网络在跟踪精度和能量效率之间的最佳权衡,结合目标的动态特性,提出了一种基于强化学习(Reinforcement Learning, RL)中Q学习框架的传感器调度算法。通过设计与能量效率和跟踪性能相关的奖励函数,网络中的传感器节点能用最小的能量开销实现对目标的高精度跟踪。仿真结果表明,所提算法相较于传统算法不仅在跟踪精度上实现至少1.1%的增益,并降低同时刻下至少34.1%的节点平均剩余能量值,对于提升目标跟踪的性能有一定指导意义。Target tracking is a technique widely used in Wireless Sensor Networks(WSNs)to estimate the position of a target as it moves within a monitored area.In order to explore the optimal trade-off between tracking accuracy and energy efficiency of the network,a sensor scheduling algorithm based on the Q learning framework in Reinforcement Learning(RL)is proposed considering the dynamic characteristics of the target.By designing reward functions related to energy efficiency and tracking performance,the sensor nodes in the network can achieve high-precision tracking of targets with minimal energy overhead.The simulation results show that,compared with traditional algorithms,the proposed algorithm not only achieves more than 1.1%gain in tracking accuracy,but also reduces the average remaining energy by at least 34.1%,which provides a guiding significance for improving the performance of target tracking.

关 键 词：无线传感网络 强化学习 传感器调度算法 跟踪调度 

分 类 号：TN967.1[电子电信—信号与信息处理]