基于遗传小波神经网络的机器人腕力传感器动态补偿研究  

Research on the Dynamic Compensation Based on Genetic Wavelet Neural Network for the Robot Wrist Force Sensor

作　　者：俞阿龙[1] 

机构地区：[1]淮阴师范学院电子与电气工程系,江苏淮安223001

出　　处：《电气自动化》2009年第5期14-17,20,共5页Electrical Automation

基　　金：江苏省高等学校自然科学基础研究基金;本校博士基金资助项目(项目号07KJD510027;07HSBS003)

摘　　要：提出一种基于改进遗传算法进化小波神经网络用于机器人腕力传感器动态补偿的新方法,介绍算法原理。该方法利用腕力传感器的动态标定数据,用改进的遗传算法来优化小波神经网络结构和参数,建立腕力传感器的动态补偿模型。结果表明,采用遗传小波神经网络进行腕力传感器动态补偿,能克服BP算法存在易陷入局部极小点的缺点,网络的复杂度、收敛性和泛化能力得到了好的综合,补偿模型建立的速度和精度得到提高。A kind of new dynamic compensation method is presented based on improved genetic algorithm （IGA） and wavelet neural networks （WNN） for a new type robot wrist force sensor and the principle of algorithm is introduced. In this method, the dynamic compensation model of the wrist force sensor can be set up according to measurement data of the dynamic calibration, where the structure and parameters of wavelet neural networks of the dynamic model are optimized by improved genetic algorithm. The results show that the proposed new dynamic compensation method can overcome the shortcomings of BP algorithm, such as easy get into the local minimum points, and the network complexity, the convergence and the generalization ability are good compromised and the training speed and precision of compensation model are increased.

关 键 词：机器人 腕力传感器 动态特性 动态补偿 小波神经网络 遗传算法 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]