用于功能性电刺激精密控制的BP神经网络整定PID算法(英文)

BP Neural Network Tuning PID Algorithm for Precision Control of Functional Electrical Stimulation

为解决功能性电刺激(FES)的电流强度精密控制问题,使电刺激作用效果能准确完成预定的功能动作,利用3层误差后向传递(BP)人工神经网络来整定传统的比例微积分(PID)控制器参数.根据下肢膝关节运动角度准确、稳定、实时地反馈控制FES系统刺激电流强度,并通过PID整体控制过程中偏差的均方根(RMS)值及实际运动轨道与预期运动轨道的偏差值评估其控制效果.实验结果表明:与传统的Ziegler-Nichols整定PID算法相比,新方法控制下的FES系统刺激电流强度偏差可以维持在相对更低的范围内,使膝关节运动轨迹与预设目标更好地吻合,从而保证更为稳定的康复训练效果.

To solve the problem of current intensity precision control for functional electrical stimulation（FES）and guarantee its exact stimulation effect to perform preset action,a back-propagation（BP）artificial neural network model with three layers was proposed to tune the parameters of traditional proportional-integral-derivative（PID）controller.Stimulation current intensity of FES system was accurately and stably modulated through a real-time feedback control based on the trajectory of knee joint angle.The control effect was evaluated by the deviation between real and preset trajectory of knee joint angle and the root mean square（RMS）of deviation in PID whole control process.Experimental results show that,in comparison with traditional Ziegler-Nichols tuning PID algorithm,the proposed method gets a lower deviation error of stimulation current intensity for FES system control,which makes the preset trajectory better matching the real trajectory of knee joint angle with more stable rehabilitation training effect.