基于GRNN的人机交互下遥操作力预测方法

Teleoperation force perception method in human-computer interaction based on GRNN

为了解决人机交互作业时操作人员不能准确感知末端接触力信息的问题,提出了一种基于GRNN信息融合的方法.基于主动端的力反馈手控器和从动端的遥操作机器人,搭建实验操作平台,以人为中心构建整个系统,同步采集手控器的姿态信号、手臂肌电信号以及末端机器人的速度、加速度和接触力信息训练GRNN,并将GRNN得到的预测力与真实力进行比较.结果显示,采用纸盒和泡沫板2种不同材料进行穿刺实验的均方误差值分别为0. 24和0. 16,泡沫板进行穿刺和切割2种不同作业得到的均方误差值分别为0. 16和0. 13,从而证明了所提方法的有效性.

In order to solve the problem that the operators can not accurately sense the contact force information in the teleoperation,an information fusion method based on GRNN(generalized regression neural network)was proposed.The teleoperation platform was built based on the force feedback hand controller in the active side and the teleoperation robot in the slave side.The operator in the whole system was considered as the center.The gesture signal of the hand controller,the electromyographic signal in the arm,the velocity,the acceleration and contact force information of the end robot were synchronously collected to train the GRNN.The prediction force by the GRNN was compared with the real contact force.The results show that the mean square errors in the puncture experiments with two different materials,the paper box and the foam board,are 0.24 and 0.16,respectively.The mean square errors in the puncture and cutting experiments with the foam board are 0.16 and 0.13,respectively.These prove the effectiveness of the proposed method.