Study On Terminal Sliding Mode Cntrol Of Manipulators Based On Neural Network And State Observer

Fast response and high precision position tracking control have always been the research hotspots of manipulators' field.Because most of the current control methods stay at the stage of asymptotic convergence of tracking error,it is of great significance to study the finite-time tracking control algorithm.Terminal sliding mode control,due to its finite time convergence characteristics and strong robustness to time-varying parameters,has attracted widespread attention in the field of robotic arm control.However,it also faces some problems that need to be resolved.On the one hand,the terminal sliding mode is subject to chattering and singularity problems.On the other hand,terminal sliding mode control is unable to independently achieve position tracking when the nominal dynamics of manipulators are unknown.In addition,most of the current design of terminal sliding mode controllers require feedback information of joint velocity signals.But considering the malfunction of speed sensors or other reasons,the velocity signals may not be available for feedback,which poses a huge challenge to the controller's design.Considering the above-mentioned terminal sliding mode control problem of manipulators,this paper studies a series of finite-time tracking control strategies for robotic manipulators,based on neural network,state observer,adaptive theory and other techniques.The research work is stated as follows:Firstly,a fast non-singular terminal sliding mode control strategy is proposed for achieving the finite-time position tracking of a robotic arm with respect to external disturbances.This method designs a reduced-order fast terminal sliding surface to improve the convergence rate of the system state which is far away from the equilibrium point,meanwhile introducing saturation suppression to eliminate singularity.Initially,external disturbances are compensated by designing robust control terms with the form of switching functions.Considering that the design of this robust term will cause chattering,an improved robust control term based on the super-twisting algorithm is designed to achieve chattering-free tracking control.Secondly,a new fast non-singular terminal sliding mode control strategy based on neural network with adaptive weight is proposed for the finite-time tracking problem when the nominal dynamics of the manipulator is completely unknown,and the effect of the actuator is considered.This method is based on fast non-singular terminal sliding mode technique to design the control law.The unknown nominal dynamics are approximated by a neural network with adaptive weights,so model-free control is achieved.Thanks to the design of adaptive control gain,this method overcomes the limitation that other mainstream controllers require the prior knowledge of the upper bound of system perturbation.Finally,a novel fast non-singular integral terminal sliding mode control algorithm based on state observer is proposed for the finite-time tracking problem of a uncertain robotic arm without velocity signal feedback.Unlike the existing finite-time observers,the proposed state observer is continuous,and its estimated speed error can converge to zero in finite time even in the face of system uncertainties,because of the introduction of an adaptive neural network method.In order to achieve high-performance position tracking,the observer is combined with a new fast non-singular integral terminal sliding mode technology to design the controller.At the same time,the controller uses a new type of adaptive super-twisting algorithm,which can not only eliminate chattering but also settle the problem that the control parameters are difficult to be selected.