外骨骼关节驱动神经网络滑模力控制研究  

Study on Sliding Mode Force Control of Eexoskeleton Joint Drive Neural Network

作　　者：周秦源 卢日荣 赵岩 胡贤哲 邓越平 张磊 ZHOU Qinyuan;LU Rirong;ZHAO Yan;HU Xianzhe;DENG Yueping;ZHANG Lei(School of Mechanical and Electrical Engineering,Central South University of Forestry and Technology,Changsha Hunan 410004,China)

机构地区：[1]中南林业科技大学机电工程学院,湖南长沙410004

出　　处：《机床与液压》2023年第19期78-83,共6页Machine Tool & Hydraulics

基　　金：湖南省重点研发计划(2019NK2022)。

摘　　要：针对外骨骼机器人液压关节驱动系统具有非线性、不确定参数等特性,导致模型建立困难以及负重时具有不确定冲击扰动的问题,基于电液伺服系统特性,建立以弹性负载为外负载的数学模型。为减小负重时冲击扰动项对力控制的影响,引入径向基(RBF)神经网络对干扰项进行补偿,设计一种基于RBF神经网络的滑模力控制策略。通过系统特性进一步验证模型可行性,并进行仿真试验对比。结果表明:与PID控制相比,所设计的控制策略响应时间更短,跟踪误差缩小70.5%;变负载工况下,所设计的控制策略具有更好的跟随能力、更强的鲁棒性能,可以满足外骨骼机器人关节驱动的力控制要求。平台试验进一步验证了仿真结果的有效性与正确性。Aiming at the exoskeleton robot hydraulic joint drive system with nonlinear and uncertain parameters characteristics,which lead to modeling difficult and the uncertain impact disturbance problems with load,based on the characteristics of electro-hydraulic servo system,a mathematical model with elastic load as external load was established.In order to reduce the influence of impact disturbance term on force control,radial basis function(RBF) neural network was introduced to compensate the disturbance term,and a sliding mode force control strategy based on RBF neural network was designed.The feasibility of the model was further verified by the system characteristic,and the simulation test was compared.The results show that compared with PID control,the response time is shorter and the tracking error is reduced by 70.5%;under variable load conditions,the designed control strategy has better following ability and stronger robust performance,which can meet the force control requirements of the exoskeleton robot joint drive.The platform experiment further verifies the validity and correctness of the simulation results.

关 键 词：外骨骼机器人 关节驱动 RBF神经网络 力控制 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]