基于随机权重混沌粒子群反向神经网络算法的RFID定位算法

为了提高射频识别 (radio frequency identification,RFID)定位系统的准确性与可靠性,并改善反向传播神经网络算法在RFID定位算法中收敛速度慢、精度低、稳定性差的缺点,提出一种基于随机权重的混沌粒子群优化反向传播神经网络定位方法 (random weight chaotic particle swarm optimization back-propagation neural networks,RW-CPSO-BP)。在运用RW-CPSO-BP算法对标签进行轨迹预测过程中,首先对3个阅读器接收到的标签接收信号强度指标值 (received signal strength indicator,RSSI)先用均值滤波进行预处理,然后再进行归一化处理,将处理过的数据分为两组在RW-CPSO-BP模型中进行训练,从而得到误差曲线。经过RW-CPSO-BP优化过权值和阈值的BP模型能较好地定位移动标签,而且克服了陷入局部最优解的问题。仿真结果表明,RW-CPSO-BP算法收敛速度和稳定性明显高于BP网络,误差也较BP网络低,更加适合用于进行复杂环境下的RFID定位。

RFID Positioning Algorithm Based on Random Weight Chaotic Particle Swarm Optimization Back-Propagation Neural Networks

In order to improve the accuracy and reliability of outdoor radio frequency identification (RFID) positioning systems, a back-propagation (BP) neural network positioning method based on random weight (RW) chaotic particle swarm optimization (CPSO) is adopted. Since BP model has the characteristics of slow convergence, low accuracy and poor stability in RFID location algorithm, this paper adopts RW based CPSO-BP (RW-CPSO-BP) algorithm to improve these shortcomings. RW-CPSO-BP model with weight and threshold can locate the moving tag well and overcome the problem of falling into local optimal solution. In the process of using CPSO-BP algorithm to predict the track of tags, the RSSI of tags received by three readers was preprocessed by means of mean filtering, and then normalized. The processed data were divided into two groups for training in the RW-CPSO-BP model, so as to obtain the error curve. Simulation results show that RW-CPSO-BP algorithm based on random weight has a higher convergence speed and stability than BP network, and a lower error than BP network, which is more suitable for RFID positioning in complex environment.