用于机器人轨迹定位纠正系统的回声状态神经网络研究

Research on Echo State Networks (ESNs) Applied into Robotic Position Estimation Correction System

循环神经网络(RecurrentNeuralNetworks)是人工神经网络(ArtificialNeuralNetworks)中重要的分支,与前馈神经网络(ForwardNeuralNetworks)相比具有更好的时间序列学习能力。但长期以来其学习法一直不能脱离前馈神经网络而自成一体,回声状态神经网络(EchoStateNetworks(ESN))是打破这一局面的全新学习方法。其独特的结构,良好的短期记忆能力,方便的学习方法,不俗的非线性特性是以前循环神经网络所不可比的。本文在介绍了回声状态神经网络之后将其用于四轮机器人的位置测量系统中,有良好的表现。

Recurrent Neural Networks （RNN） is one important branch belonging to Artificial Neural Networks （ANN） with better ability for learning time series in comparison with Forward Neural Networks （FNN）, howbeit no independent learning algorithm based on FNNs＇. The Echo State Networks （ESN） broke through this situation as one novel approach. It appears special construction, excellent short-term memory, facile learning approach, stunning non-linearity that former RNN never done. In this paper, the ESN would be introduced firstly, and then be used into the position estimation system of robot, resulting well.