基于神经网络辨识的遥康复系统变增益控制

Variable gain control for tele-rehabilitation systems based on neural network identification

针对康复训练过程中患者肌肉痉挛会对力反馈遥操作系统稳定性和从机械手速度平滑性产生较大影响的问题,提出了一种新的基于反向传播(BP)神经网络辨识的变增益控制方法。该方法通过 BP 神经网络实时辨识患肢动力学参数的变化并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和起到抑制患者痉挛状态的作用。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。

To solve the problem that during tele-rehabilitation training, rehabilitants＇ muscle spasm can make the tele-rehabilitation system instable and cause its slave＇ s movement unsmoothness, the paper proposes a new method for control of the variable gain of a tele-rehabilitation system based on back propagation （BP） neural network identification. The method real-time identifies the changing of limb＇s dynamics parameters with BP neural networks and performs self-adaptive control of gains, thus not only the system instability is eliminated, but also the slave＇ s movement unsmoothness is apparently lessened. The rehabilitation training efficiency can be largely enhanced and it can have an inhibiting effect on patients＇ spasticity. The analysis and simulation results show that this method can effectively restrain the spasm interference, and is much more stable and smooth than the traditional control methods.