人-机身体交流的运动控制和实验研究  

Motion Control and Experimentation on Physical Human-Robot Interaction

作　　者：谢光辉[1,2,3] 金敉娜[2] 王光建[1] 桥本稔[3] 杨治平[2] 

机构地区：[1]重庆大学机械传动国家重点实验室,重庆400044 [2]重庆电子工程职业学院,重庆401331 [3]日本信州大学

出　　处：《应用基础与工程科学学报》2014年第5期1018-1029,共12页Journal of Basic Science and Engineering

基　　金：国家留学基金委项目(2007102654);重庆市自然科学基金项目(cstc2012jjA40028);重庆市教委科学技术研究项目(KJ112203)

摘　　要：对人和机器人身体交流运动的协调同步问题,提出了基于逐次迭代矢量场的人-机身体交流运动控制模型;设计了逐次迭代矢量场算法,使该矢量场输出的关节期望位移能与输入的扭矩信号保持同步,且同步程度能实现参数可调;分析了两个不同逐次迭代矢量场的相互作用情况;利用7自由度机器人臂进行握手实验,以验证所提出控制模型的有效性,并通过参数识别和BP-NN神经网络对该机器人臂进行重力补偿,以便实时控制.实验结果表明,该控制模型对于实现人和机器人相互运动的协调同步是有效的.To synchronize motions between robot and human, a control model for physical human- robot interaction was proposed, based on vector field with successive iteration. The algorithm on vector field with successive iteration was designed, so that vector field can realize the synchronization between joint torque information as an input signal and desired trajectory of each robot joint as an output signal, and the strength of the synchronization can be varied by adjusting some internal parameters value in the vector field. The mutual interaction for two different vector fields was analyzed. Based on 7 DOF robot arm, human-robot handshaking experiment was implemented by adopting the proposed control method, and gravity compensation for the robot arm was performed to realize real-time control by using back propagation neural network （BP-NN）learning and parameter identification. The experiment results indicate that the control method is effective to synchronize motions between robot and human.

关 键 词：逐次迭代 参数识别 BP-NN神经网络 重力补偿 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]