基于HJB方程稳定流形方法的两轮自平衡车最优控制  

Optimal Control for Two-wheeled Self-balanced Vehicles Based on Stable Manifold of HJB Equation

作　　者：岳玉环 YUE Yuhuan(Jiaxing Vocational and Technical College,Jiaxing 314036,China)

机构地区：[1]嘉兴职业技术学院,浙江嘉兴314036

出　　处：《机械制造与自动化》2023年第5期176-180,共5页Machine Building & Automation

基　　金：嘉兴职业技术学院校级课题项目(22407030100)。

摘　　要：基于HJB方程的稳定流形结构,借助深度神经网络的学习能力得到反馈式最优控制器,实现对两轮自平衡车欠驱动系统的稳定平衡控制。通过数值仿真结果验证该最优控制器的控制效果,尤其在系统参数不确定及存在外部扰动的情况下表现出了良好的鲁棒性。该方法避免了传统数值求解HJB方程方法中遇到的“维数诅咒”问题,为非线性系统最优控制的实现提供了一种可行的方法。Based on stable manifold of hamilton-jacobi-bellman equation and with help of the learning ability of deep nerual network,an optimal controller in feedback form is abtained,realizing stable balance control of two-wheeled self-balanced vehicles without driving system.The numerical simulation results prove the favourable effectiveness of the optimal controller,and its good robustness under uncertain system parameters and external disturbances in particular.The method avoids the“Curse of Dimensionality”experiencing in traditional numerical solution of HJB equation,and provides a feasible approach for optimal control of nonlinear systems.

关 键 词：HJB方程 稳定流形 深度学习 最优控制 两轮自平衡车 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]