工业机器人智能运动学模型  

Industry Robot Intelligent Kinematic Model

作　　者：丁度坤[1] 张铁[2] 谢存禧[2] 

机构地区：[1]东莞职业技术学院,广东东莞523808 [2]华南理工大学机械与汽车工程学院,广东广州510640

出　　处：《机械设计与制造》2017年第8期242-244,共3页Machinery Design & Manufacture

基　　金：广东省数控一代项目(2013B011301003);广东省高等学校优秀青年教师培养计划项目(YQ2015232);东莞市产学研合作项目(2014509102211);东莞职业技术学院政校行企合作开展科研与服务项目(ZXHQ2014d003);东莞职业技术学院院级基金项目(2015a03)

摘　　要：针对机器人运动学正、逆解推导过程复杂,计算量大的情况,提出了一种基于神经网络的机器人运动学正、逆解计算新方法。首先搭建了6自由度工业机器人实验平台,操纵机器人沿某一轨迹运动,记录下机器人在采样时刻的姿态角、坐标及关节角,获取实验数据。在此基础上,设计了一个三层神经网络,输入所采集到的数据进行训练,构建了机器人运动正反解神经网络模型。文章最后对所构建的模型进行验证,验证结果表明,由运动学模型所得到的预测值与实际的测量值误差小,模型具有较高的准确度。To solve the problem that the robot＇s forward and inverse kinematics deducing process are very complicated and the calculation burden is very great, a new mothod for robot＇s forward and inverse kinentatics solution based on the neural network is proposed in it. Firstly, a 6-DOF industry robot platform was set up. Then the robot was manipulated along certain trajectory. The attitude angles, the coordinates and the joints＇ angles of robot were saved at sample time, so the experiment datas can be obtained. On this basis, a neural network contained three layers is setup, then the experiment datas are input to train the network. Therefore the neural network models for robot＇s forward and inverse kinematics are setup. In the end of it, several experiments are eaught out to test the models. The results showed that the perdietion values calculated by the models are fit to the real measured values, and the models have high accuracy.

关 键 词：工业机器人 机器人运动学 运动学正逆解 神经网络 

分 类 号：TH16[机械工程—机械制造及自动化] TP242[自动化与计算机技术—检测技术与自动化装置]