基于神经网络辨识的移动机器人航向误差校准方法

Calibration method for heading error of mobile robot based on neural networks identification

分析了E-Core RD1100干涉型光纤陀螺的误差产生机理,提出利用RBF神经网络和遗传算法实现光纤陀螺漂移误差模型的辨识。通过实验获得进化神经网络的训练样本,在RBF神经网络的训练中,提出了基于Elitist竞争机制的遗传进化训练方法。RBF神经网络具有很强的局部逼近能力,而遗传算法具有优良的全局搜索与优化性能,从而能够有效地对陀螺误差的非线性与时变特征进行建模与辨识。实验结果表明:该方法大幅度减少了光纤陀螺的误差,从而提高了移动机器人导航定位的精度。

The basis of error occurrence of E-Core RD1100 interferometric fiber optic gyros （FOG） was analysed. Radial basis function neural network （RBFNN） and genetic algorithm （GA） were adopted to realize the identification of error model for FOG. The training samples were obtained by experiments and the genetic evolutionary method based on elitist rule was presented for neural network training. RBFNN has a good capacity of local approximation, while GA is efficient in global optimization. So this method can be used to realize the nonlinear and time-varying error modeling and identification. The experimental results show that it can reduce the error of FOG to a great extent and enhance the localization precision of mobile robot navigation.