Nonlinear Compensation Control Method Of Harmonic Drive Joint Based On Command Filtering Backstepping

With the rapid development of science and technology,robots has been widely used in industry,service industry,medical and health,aerospace and other fields.Because of its compact structure,high operation accuracy and large transmission ratio,harmonic reducer occupies a huge market share in robot joints.In the actual working environment,the robot joint with harmonic reducer as the transmission core has many complex nonlinear characteristics,such as flexibility,friction,external disturb and so on.Due to the limitation of physical conditions,the control system can only produce limited amplitude control input.When the control input exceeds the upper and lower limits of the actuator output,it will produce saturation and reduce the tracking performance of the system.The joint controller usually needs the angle and velocity information to feed back to the closed-loop control system,but the joint velocity state is sometimes polluted by noise or not easy to be measured.Therefore,in order to meet the requirements of accuracy and reliability in the work of robot,this paper takes the harmonic drive joint composed of harmonic reducer and permanent magnet synchronous motor as the research object,and considers the factors of flexibility,friction,disturb,input constraints and unmeasurable speed of the system,and studies the nonlinear control strategies of the harmonic drive joint.The main contents of this paper are as follows(1)Aiming at the problems of friction,external disturb and modeling error of harmonic drive joint,a command filtering backstepping control method based on extended disturb observer is designed.The extended disturb observer is used to obtain the effective observation values of friction,external disturb and modeling error.The command filtering backstepping controller solves the problem of " explosion of complexity " in traditional backstepping method,and makes up for the deficiency of filtering error in dynamic surface control method.The Lyapunov function of the system is constructed to analyze the stability of the controller and ensure the convergence of all States and errors of the closed-loop system.The results show that the designed extended disturb observer can estimate high-order disturb with high observation accuracy.The controller with disturb compensation achieves good tracking accuracy and control performance under the action of friction,disturb torque and modeling error.(2)Aiming at the problem of control input constraints of harmonic drive joint,an command filtering adaptive backstepping control method of input saturation based on RBF neural network is proposed.The controller uses saturation function and compensation mechanism to constrain the amplitude of control input,and uses RBF neural network to approach unknown disturbances such as friction,external disturbance and modeling error online.The simulation results show that the amplitude of the control input is always within the specified range under the saturation compensation mechanism,and the RBF neural network can accurately estimate the external step disturb and modeling error.The proposed control algorithm can not only achieve high-precision tracking of target trajectory,but also resist the adverse effects of external step disturb torque and modeling error on the control system.(3)Considering joint speed immeasurability and system control input constraints,a nonlinear command filtering adaptive backstepping control method based on linear extended observer and RBF neural network is designed.The term related to joint velocity in dynamic equation of harmonic reducer is regarded as a part of modeling error,and friction,external interference and modeling error are regarded as compound interference.The linear extended observer is used to observe the compound disturbance and speed at the load end and input end of the harmonic reducer,and the RBF neural network is used to approximate the disturbance and modeling error in the permanent magnet synchronous motor model.The results show that the proposed method has good trajectory tracking performance with input constraints and absence of speed measurement.The steady-state tracking accuracy of the controller will not be reduced by the step external disturb and the change of system model parameters.