基于神经网络重力前馈补偿的柔性关节机器人分层控制  

Hierarchical Control of Flexible-joint Robot with Gravity Feed-forward Compensation Based on Neural Network

作　　者：李泽国 李国栋 LI Zeguo;LI Guodong(School of Mechatronical Engineering, Beijing Institute of Technology, Beijing 100081, China)

机构地区：[1]北京理工大学机电学院,北京100081

出　　处：《天津科技大学学报》2017年第6期53-58,64,共7页Journal of Tianjin University of Science & Technology

基　　金："十二五"国家科技支撑计划资助项目(2012BAI25B01)

摘　　要：针对柔性关节机器人简化动力学模型,根据连杆侧和电机侧间的扭矩通过弹性单元耦合的特性提出基于分层控制的柔性关节机器人分层控制结构,利用奇异摄动理论给出系统的稳定性证明.由于柔性关节机器人的动力学参数无法准确测量,因此提出应用BP神经网络模型对机器人的重力扭矩进行逼近.通过四自由度柔性关节机器人进行离线实验,实现对神经网络的训练,并验证神经网络模型的有效性.最后应用提出的重力前馈补偿分层控制算法实现对机器人的有效控制,验证该算法的可行性和有效性.A hierarchical control structure is presented in this paper based on elastic torque coupling between the link side dynamic and the motor side dynamic for a simplified model of the flexible-joint robot. The stability of the system wasproved by the singular perturbation principle. A BP neural network model was developed to approximate the gravity torque, because the dynamic parameters of the flexible-joint robot can not be measured accurately. The neural network was trained with theoff-line method and its availability was verified on the flexible-joint robot with 4 degree of freedom. Finally, the proposed hierarchical control with gravity feed-forward compensation was used to realize the effective control of the robot and thus verified the feasibility and availability of the algorithm.

关 键 词：柔性关节机器人 分层控制 奇异摄动 神经网络 重力前馈补偿 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]