基于卷积神经网络的冗余机械臂运动学逆解求解

Inverse Kinematics Solution of Redundant Manipulator Based on Convolution Neural Networks

针对7自由度冗余机械臂运动学逆解求解时间过长以及通用性差等问题,通过对LeNet模型进行改进,对卷积核、激活函数、损失函数、优化算法、模型结构与初始参数等进行设计与选择,建立了卷积神经网络模型来进行求解。对卷积神经网络模型进行轨迹跟踪实验验证,结果表明:所得到的机械臂关节角度平滑,轨迹跟踪最大误差小于3 mm;所建立的模型是有效的且精度较高;此方法可求解机械臂运动学逆解问题。

The inverse kinematics solution of 7-DOF redundant manipulator is complex,and the problems such as the long solving time or the poor generality need to be solved.In this paper,the LeNet model was improved.Then the convolution kernel,the activation function,the loss function,the optimization algorithm,the model structure and the initial parameters were designed and selected.On this basis,a convolution neural networks model was established to solve the inverse kinematics of 7-DOF manipulator.The convolution neural networks model was verified by trajectory tracking experiment and the maximum error of trajectory tracking was less than 3 mm.The results showed that the model was effective and accurate,and the method could be used to solve the inverse kinematics problem of the manipulator.