Abstract：The modeling and compensation control of hysteresis in pneumatic muscle(PM) are of great significance to the accurate motion control. The traditional hysteresis modeling methods have series problems such as many parameters to be identified, complex parameter identification process and low identification accuracy. Thus the least squares support vector machine (LS-SVM) approach is proposed to characterize the hysteresis phenomenon of pneumatic muscle. This method maps the original data space to the high-dimensional space by non-linear mapping, thus the non-linear problem of the original system is transformed into a linear problem in the high-dimensional space, the least square method is used to solve the system of linear equations, which improves the speed of solution and convergence accuracy. The mathematical model of displacement/pressure hysteresis on pneumatic muscle was established by using LS-SVM method, on the basis of experiments on hysteresis characteristics of pneumatic muscle. The calculation results show that the mathematical model established by least squares support vector machine has higher modeling accuracy and various error indices such as mean variance and mean error are significantly reduced, which reduce 99.21% and 99.1% respectively compared with the classical PI model. Thus providing an effective means for subsequent hysteresis compensation control of pneumatic muscle.
2020, Vol. 41 
