基于神经网络辨识的变增益控制在遥康复中的应用

The application of a variable gain control to tele-rehabilitation based on neural network identification

在患者进行遥康复训练时,肌肉痉挛会对遥康复训练系统稳定性及从机械手速度平滑性产生较大的影响。针对这些问题提出了一种新的基于BP神经网络辨识的变增益控制方法。通过BP神经网络实时辨识患肢动力学参数并根据其变化进行自适应调整控制增益,不仅消除了因患者肌肉痉挛带来的不稳定性,而且减少了其对系统运动平滑性的影响,可提高康复训练效果。分析和仿真试验结果表明,该控制方法与传统的控制方法相比,可有效地抑制患者因肌肉痉挛带来的干扰并具有较好的稳定性和平滑性。

Rehabilitant＇s muscle spasm can make a tele-rehabilitation training system instable and cause its slave＇s movement to be unsmooth when a rehabilitant receives a tele-rehabilitation training. To solve these problems, this paper presents a new variable gain control technique based on the BP neural netowrk identification. The BP neural network can perform its real-time identification of limb＇s dynamic parameters and self-adapting control gains according to their changes, thus eliminating the system inslability and lessening the unsmoothness of the slave＇s movement as well as improving the rehabilitation training efficiency. Analysis and simulation results have proved this method can retrain the spasm interference effectivly and is much stabler and smoother than the conventional control.