基于iTransformer模型的工业机器人关节迟滞建模研究  

Modeling of Joint Hysteresis in Industrial Robots Based on iTransformer Models

作　　者：党选举[1] 沈远桂 DANG Xuanju;SHEN Yuangui(School of Electronic Engineering and Automation,Guilin University of Electronic Technology,Guilin 541004,China)

机构地区：[1]桂林电子科技大学电子工程与自动化学院,桂林541004

出　　处：《组合机床与自动化加工技术》2025年第4期177-182,共6页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金项目(62263004);广西重点研发计划项目(桂科AB23075102)。

摘　　要：针对工业机器人的柔性关节表现的复杂迟滞特性问题,提出了一种基于改进的iTransformer(inverted transformer)模型的非线性迟滞建模方法,用于描述柔性关节的输出力矩与扭转角之间的复杂迟滞特性。采用具有输入反转嵌入层的iTransformer模型,体现时间前后信号关联性,其自注意力机制,描述了迟滞的记忆特性。为了提高迟滞模型的精度,从结构及输入前端信息两个方面对iTransformer模型进行改进:在结构上,将LSTM网络与iTransformer模型并联,以增强迟滞模型的动态特性;在输入端,设计了一个自相关模块。将输入扭转角向量进行哈达玛积运算,构成新输入向量,并与所设计类迟滞算子结合,设计一个iTransformer模型前端的自相关模块,充分体现了输入信号的关联性。与iTransformer、LSTM及PI迟滞模型相比较,实验结果表明,所提出迟滞模型,对工业机器人的复杂迟滞特性的描述,具有更高的精度和泛化能力。Aiming at the problem of complex hysteresis characteristics exhibited by flexible joints of industrial robots,a nonlinear hysteresis modeling method based on an improved iTransformer(inverted transformer)model is proposed for describing the complex hysteresis characteristics between the output torque and torsion angle of flexible joints.The iTransformer model with an input inversion embedding layer is used to embody the signal correlation before and after time,and its self-attention mechanism,which describes the memory properties of hysteresis.In order to improve the accuracy of the hysteresis model,the iTransformer model is improved in terms of both structure and input front-end information:structurally,the LSTM network is connected in parallel with the iTransformer model in order to enhance the dynamic characteristics of the hysteresis model,and at the input end,a self-correlation module is designed to perform Hadamard product operation on the input torsion angle vector,and combined with the designed class of hysteresis operators to design a self a correlation module at the front end of the iTransformer model,which fully reflects the correlation of the input signals.Compared with the iTransformer,LSTM and PI hysteresis model,the experimental results show that the proposed hysteresis model,which describes the complex hysteresis characteristics of industrial robots,has higher accuracy and generalization ability.

关 键 词：柔性关节 复杂迟滞特性 iTransformer模型 类迟滞算子 自相关模块 

分 类 号：TH165[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]