Deep adaptive control with online identification for industrial robots

作     者：SHEN Tan QIAO XueChun DONG YunLong WANG YuRan ZHANG Wei YUAN Ye 

作者机构：School of Artificial Inteligence and AutomationHuazhong University of Science and TechnologyWuhan 430074China Department of Mechanical and Energy EngineeringSoutherm University of Science and TechnologyShenzhen 518055China 

出 版 物：《Science China(Technological Sciences)》 (中国科学（技术科学英文版）)

年 卷 期：2022年第65卷第11期

页      面：2593-2604页

核心收录：

学科分类：08[工学] 0835[工学-软件工程] 0802[工学-机械工程] 080201[工学-机械制造及其自动化] 

基　　金：supported by the National Natural Science Foundation of China (Grant No. 52188102) 

主　　题：industrial robots sparse Bayesian learning online identification adaptive control 

摘      要：Derivation of control equations from data is a critical problem in numerous scientific and engineering *** inverse dynamic control of robot manipulators in the field of industrial robot research is a key ***,researchers needed to obtain the robot dynamic model through physical modeling methods before developing ***,the robot dynamic model and suitable control methods are often elusive and difficult to tune,particularly when dealing with real dynamical *** this paper,we combine an enhanced online sparse Bayesian learning(OSBL)algorithm and a model reference adaptive control method to obtain a data-driven modeling and control strategy from data containing noise;this strategy can be applied to dynamical *** particular,we use a sparse Bayesian approach,relying only on some prior knowledge of its physics,to extract an accurate mechanistic model from the measured *** parameters are further identified from the modeling error through a deep neural network(DNN).By combining the identification model with a model reference adaptive control approach,a general deep adaptive control(DAC)method is obtained,which can tolerate unmodeled *** adaptive update law is derived from Lyapunov’s stability criterion,which guarantees the asymptotic stability of the ***,the Enhanced OSBL identification method and DAC scheme are applied on a six-degree-of-freedom industrial robot,and the effectiveness of the proposed method is verified.