基于神经网络的机器人无标定视觉伺服系统仿真  

Simulation of an uncalibrated robotic visual servo system based on neural network

作　　者：张家驹[1] 吴洪涛[1] 

机构地区：[1]南京航空航天大学机电学院,江苏南京210016

出　　处：《应用科技》2011年第9期1-5,共5页Applied Science and Technology

基　　金：国家自然科学基金资助项目(50375071)

摘　　要：针对机器人三自由度平动的视觉定位问题,建立全局有效的非线性视觉映射关系模型,并设计人工神经网络(ANN)对其拟合,从而实现无标定视觉伺服.伺服系统的物理结构采用单目视觉和眼在手上(eye-in-hand)构型.机器人系统采用典型工业机器人puma560,并限定末端姿态使其保持平动.基于机器人工具箱(ro-botics toolbox for Matlab)和神经网络工具箱,在Matlab/Simulink环境下建立仿真模型,对系统的开发具有重要的参考意义.仿真结果表明,该系统对静态物体的定位能够达到较快的响应和较高的精度.Aiming at the 3-DOF robot visual orientation problem, we established the model which shows nonlinear visual mapping relation and designed ANN （artificial neural network） to fit it to realize the uncalibrated visual servo. It is a monocular, eye-in-hand model. The camera keeps translational in work space to accomplish positioning. The robot system is a typical industrial robot-puma560, and we limit the freedoms of posture. Based on robotics toolbox for MATLAB and neural network toolbox, we established simulation model in MATLAB/Simulink, which has important reference meaning. The simulation results show that this system for static object localization can achieve a quicker response and higher precision.

关 键 词：视觉定位 人工神经网络 puma560 MATLAB/SIMULINK 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]