基于前馈神经网络的柔性连续臂建模与仿真

柔性连续臂的设计研究近年来取得重大进展，但其建模研究一直相对滞后。为此，文章对柔性连续臂进行运动学建模研究，提出使用临时姿态矩阵求解逆运动学的方法，解决了逆运动学不易收敛的问题；利用前馈神经网络对基于模态振型函数（MSF）的模型进行拟合，建立一种端到端的正运动学模型和逆运动学模型，在保证模型精度的同时，显著提高了其逆运动学求解效率。该模型可以灵活更改前馈神经网络输入输出神经元的数量，从而将自身迁移到多节模型中。

Modeling and simulation of flexible continuum arm based on feedforward neural network

In recent years, great progress has been made in the design of the flexible continuum arms, but its modeling research lags behind relatively. This paper focuses on the kinematic modeling of the flexible continuum arm, involving the inverse kinematics, which is hard to converge. The problem is solved by using a temporary pose matrix. The feedforward neural network is used to fit the model based on the mode shape function (MSF), and an end-to-end forward kinematics model and an inverse kinematics model are established, which not only ensures the accuracy of the model, but also significantly reduces the time in solving the inverse kinematics. In addition, the model allows a flexible change of the number of the input and output neurons of the feedforward neural network, thus helps to adapt to a multi-section model.