基于广义回归神经网络的视觉球形机器人建模  

Visual spherical robot modeling based on generalized regression neural network

作　　者：翟光耀 章政 郭昱琛 黄卫华 翟民 ZHAI Guangyao;ZHANG Zheng;GUO Yuchen;HUANG Weihua;ZHAI Min(Institute of Robotics and Intelligent Systems,Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]武汉科技大学机器人与智能系统研究院,湖北武汉430081

出　　处：《传感器与微系统》2024年第6期15-19,共5页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(61773298)。

摘　　要：由于球形机器人具有复杂的机械结构和特殊的运动方式,导致其动力学模型具有非线性、多变量、强耦合、参数不确定等复杂因素,因此难以建立精确的数学模型。针对上述问题,设计了一种改进广义回归神经网络(GRNN)对其进行建模。首先,获取基于机理模型的球形机器人实测数据;然后,基于实测数据训练出改进GRNN模型并分析其预测效果;最后,分别基于改进GRNN和机理模型,设计球形机器人的控制器进行自平衡实验,前者比后者受到干扰时的波动幅度更小、调节时间短了近1 s。实验结果证明了所设计建模方法的可行性和有效性。Due to complex mechanical structure and special motion mode of the spherical robot,its dynamic model is characterized by nonlinear,multivariable,strong coupling,parameter uncertainty and other complex factors,so it is difficult to establish an accurate mathematical model.Aiming at the above problems,an improved generalized regression neural network(GRNN)is designed for modeling.Firstly,the measured data of the spherical robot based on the mechanism model are obtained.Then,an improved GRNN model is trained based on the measured data and its prediction effect is analyzed.Finally,the controller of the spherical robot is designed based on the improved GRNN and the mechanism model respectively for self-balancing experiments.The fluctuation amplitude of the former is smaller and the adjustment time is shorter than that of the latter by nearly 1 s.Experimental results show that the proposed modeling method is feasible and effective.

关 键 词：球形机器人 视觉装置 动力学建模 灰狼优化算法 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置] TP212[自动化与计算机技术—控制科学与工程]