分数阶微分D^(λ)型迭代学习算法在机器人控制中应用  

D^(λ) Type Iterative Learning Algorithm and Its Application on Robot Based on Fractional Calculus

作　　者：周爱美 王飞 Zhou Aimei;Wang Fei(College of Intelligent Manufacturing,Zhejiang Guangxia Construction Vocational and Technical University,Dongyang,Zhejiang 322100,China)

机构地区：[1]浙江广厦建设职业技术大学智能制造学院,浙江东阳322100

出　　处：《机电工程技术》2022年第5期233-237,共5页Mechanical & Electrical Engineering Technology

摘　　要：工业机器人作为工业4.0的重点研究对象及重要装备支撑,其相关技术研究特别是控制技术研究具有重要的研究意义。由于工业机器人具有特殊的应用环境和任务性质,因此,迭代学习控制在工业机器人等具有重复运行特性的领域有着非常好的应用前景。鉴于此,针对一般迭代学习算法存在的收敛速度、精确等问题,在D型迭代学习控制算法的基础上推广到分数阶微分D^(λ)型,并对这种算法进行了相应的稳定性证明和仿真验证,通过对比,分数阶微分D^(λ)型迭代学习算法通过选取合适的λ可以相应地提高收敛速度和控制精度,同时,新算法通过数值计算的方式可以较好的诠释系统I/0因果性关系。最后,为了进一步增加这类算法的适用性,将D^(λ)迭代学习控制算法和自适应控制算法相结合,以水平关节工业机器人为例,设计了一种自适应D^(λ)型迭代学习控制算法。As the key research object and the important equipment support of Industry 4.0,the research of the related technology,especially the control technology,is of great significance.Iterative learning control has a unique advantage in the control of nonlinear,unknown model system.With repetition characteristics in industrial robots,CNC machine tools and other areas have a good prospect.As a young subject,iterative learning control in many aspects still need further research and improvement.The traditional D type iterative learning control was extended to fractional D^(λ) type iterative learning algorithm.The new algorithm did not require derivative information of the system output error,and could well reflect the causality between inputs and outputs.Focusing on the linear discrete system,the specific iterative learning law was given.The simulation results show that the proposed iterative learning algorithm has better convergence properties than ordinary type iterative learning algorithm.Finally,in order to further increase the applicability of this kind of algorithm,an adaptive D^(λ) type iterative learning control algorithm was designed by combining the D^(λ) type iterative learning control algorithm with the adaptive control algorithm.

关 键 词：迭代学习控制 分数阶微积分 D^(λ)型迭代学习算法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]