提高机器人位置、速度传感器分辨力的方法

用光栅构成的机器人速度、位置传感器测量系统常用硬件进行信号细分,存在细分数不高,硬件复杂等问题。为此,研究了一种神经网络细分方法,并研究了一种用于细分的直接映射小脑模型神经网络。实验和实用证明,研究的小脑神经网络具有学习精度高、速度快,算法简单等特点;只要用很少的训练样本即可达到很高的细分精度,使分辨力得到很大提高,简化了硬件设计,提高了系统的可靠性。

Method of improving resolution for position sensor and speed sensor in robots

*"It is difficult and lower with accuracy that signal of sensors measuring system for controlling position and speed of robots which are composed of gratings is subdivided by hardware.Wherefore,a method of subdividing grating signals is studyed using a neural networks and a input direct mapping low dimensional cerebellar model neural networks is proposed.The results of experiment and practical application show that it's accuracy of learning low dimensional nonlinear functions is very high,while the algorithm is very simple,learning speed is very fast,and the subdividing accuracy is very high as long as to use several training samples.So,not only the hardware designing is greatly simplified,but also the Resolution and reliablity of system are remarkably improved.