基于模糊神经网络的传感器可信度实时获取

Acquisition of Sensors' Credence Based on Fuzzy Neural Network

针对传感器在复杂环境中所测信息不完全准确的问题,提出了一种基于专家规则的零阶Sugeno模糊模型神经网络来获取传感器可信度的方法.神经网络经训练样本训练后,可以根据传感器状态和环境信息实时地得到传感器可信度.将该模型学习算法中的最小二乘识别器加以改进,并引入了遗忘因子,可以使该网络实现在线学习,不断更新网络参数.仿真结果表明该模糊神经网络可以有效地获得传感器可信度,且越小则网络在线学习能力越强.

As for the uncertainty of signal obtained by sensors in complex environment, a zero-order Sugeno fuzzy neural network, based on Experts＇ Regulation, is put forward to get the credence of sensors. Trained by the modes for learning, fuzzy neural network can acquire the credence of sensors real-timely according to the states of sensors and environment information. In this network, the algorithm of least square law is improved and an oblivion factor 2 is introduced, so that, the network can do learning on-line and upload parameters continuously. The simulation indicates that this fuzzy neural network can obtain the credence of sensors effectively, and its capability of learning on line is enhanced when 2 gets smaller.