摘要
针对机器人多指手自身的特点,通过分析人手的抓取特性,对其可能具有的抓取模式进行分类。考虑被抓取物体的几何特征和任务要求,采用基函数为高斯核函数的RBF神经网络来表示被抓物体的样本特征和抓取模式之间的复杂非线性映射。将抓取模式分为10类,对于新的被抓物体,利用训练好的神经网络自动生成抓取模式,并利用VC++/OpenGL建立了可视化仿真平台,进行了抓取模式分类仿真实验,结果表明对于新的物体,机器人可以选择适当的抓取模式进行抓取。
According to the characteristics of robot hands and human hand grasping, the probable grasp modes of robot hands are classified. Taking the sample features of the object to be grasped and requirements for the task into consideration, we use the radial basis function（RBF） neural network with the Gaussian kernel function as its base function to represent the complex nonlinear mapping relationship between the grasp mode and the sample features of the object to be grasped. We classify the grasp mode into 10 categories and utilize the trained neural network to automatically generate the grasp mode for grasping the object. A visualized simulation platform is established on the platform of VC ＋ ＋/OpenGL to simulate the grasp mode classification. The simulation results show that the robot hand can select an appropriate grasp mode to grasp an object.
出处
《机械科学与技术》
CSCD
北大核心
2007年第7期889-892,共4页
Mechanical Science and Technology for Aerospace Engineering
基金
江苏省应用基础研究课题(BJ98057)
南京航空航天大学科研创新基金项目(CX200407)资助
