Fuzzy Control Study For Walking-Assisted Functional Electrical Stimulation Based On Knee Joint Angle

In recent years, the number of paraplegia patients resulting from spinal cord injury (SCI) and the stroke reveals a general trend of markedly increase. It has been a hot spot of technical front attention of bioinformatics, rehabilitation medicine and electro-neurophysiology that use the functional electrical stimulation (FES) to restore and therapy the loss of function of paraplegia. However, open-loop FES system with disadvantages of being disturbed, poor self-adaptability as well as fatigue-causing does not optimize the stimulation intensity and limits its popularization in clinical application.To improve the performance, fuzzy controller (FC) was used here to modulate the current intensity for FES with angle of knee joint as feedback signal according to human kinematics information. FC, as a novel intelligent control technology, is suitable for nonlinear complex control subject just like muscle system comparing with other traditional control technology. How to fix the parameters, such as scale factor, quantization factor and membership function, is the key issue in FC application, especially for the human system that expects even more robust. In this thesis, Ant colony optimization algorithm (ACO) and genetic algorithm (GA), with advantages of global convergence, synchronous computing and parallel optimization, were used to fix the parameters of FC respectively. And also, genetic algorithm- ant colony optimization algorithm (GAACO), which combined the two by assimilating their advantages, was used for optimization of fuzzy controller parameters in the research. A nonlinear autoregressive moving average exogenous (NARMAX) muscle model determined by relationship of stimulation current intensity and knee joint angle in the experiments was built to simulate the controlled system in this research and then used to verify the effectiveness and stability of fuzzy controllers those optimized by ACO fuzzy controller, GA controller and genetic algorithm-ant colony optimization algorithms respectively. The relative error between the modeled knee joint angle and the experiment joint angle is less than 10%. The maximum absolute error in tracking task between the preset trajectory and output results was less than 5 degree; the averaged absolute errors were less then 2 degree, and the maximum standard deviation of that were less then 0.7 degree. The optimization speed of GAACO was faster then each of ACO or GA significantly.This research implied that the kinematics of the knee angle signal can be used as feedback signal to adjust the physical stimulus intensity and the fuzzy controller which optimizes the FES current mode by GAACO is more reliable than the other two methods. So it is proved a promise approach for future application to improve the automatic FES control system and benefits for the design of updated lower limb prosthesis of artificial motor system.