Nonlinear Control Of Single Pneumatic Artificial Muscle Systems

In recent years,with the rapid development of the robot industry,it brings higher requirements for the safety and the comfort of human-machine interaction.However,rigid robots,driven by traditional actuators(e.g.,motors,cylinders,and hydraulic actuators),generally have redundant structures,large sizes,and poor flexibility,and it is becoming difficult for them to meet the application requirements of human-machine interaction.Hence,pneumatic artificial muscles(PAMs),as a kind of flexible actuators similar to biological skeletal muscles,are attracting more and more attention.With many satisfactory advantages(e.g.,safety,cleanliness,lightweight,high power density,good compliance,etc.),PAMs have great application prospects in many fields,especially in flexible robots,biomimetic robots,medical rehabilitation,and emergency relief.However,PAMs' complex geometric structures and material properties bring many inherent defects,such as strong hysteresis,creep effect,complex friction,and so on.Hence,PAM systems usually suffer from high nonlinearities,uncertainties,and time-varying characteristics,which bring great challenges to dynamics modeling and controller design.In recent years,to handle the modeling and control issues of PAM systems,a lot of methods have been presented in the literature.However,there still exist many issues to be solved,especially the poor environmental adaptability and low intelligence.Hence,to follow the requirements of practical applications,this thesis carries out in-depth research on the nonlinear control of PAM systems.Aiming to improve the precision and intelligence of actual control,strict theoretical analysis and sufficient experimental verification are provided,which have both theoretical and practical significance.The main contributions of this thesis are listed as follows:1)Neuroadaptive control for PAM systems.Due to high nonlinearities and uncertainties,it is difficult to precisely obtain the PAM's dynamic model over the entire range of working pressure.To handle the above issues,in this thesis,based on the three-element model of PAMs,a three-layer neural network(NN)is constructed to compensate the unknown nonlinear term in the dynamic model.Without any pretraining,the adaptive laws with projection operators are designed to make the input/output weight matrix/vector of the NN adjusted online.Furthermore,considering that most of the control strategies of PAM systems do not consider the boundary constraints of tracking errors,a nonlinear robust controller is designed based on a sliding mode surface.By means of the proposed controller,tracking errors can not only converge to zero asymptotically,but also never escape the preset bounds throughout the entire control process.Moreover,the stability of the closed-loop system is proven theoretically by utilizing Lyapunov techniques.Finally,a series of hardware experiments are implemented on a vertical single PAM testbed to validate the effectiveness and robustness of the proposed method.2)High-order disturbance observer-based nonlinear control for PAM systems.Except for nonlinearities and uncertainties,PAM systems are rather sensitive to noises.In the working environment,PAM systems inevitably suffer from the unknown,complex,and persistent external disturbances.Considering that most active disturbance rejection control methods of PAM systems assume that the change rate of the lumped disturbance is negligible,a nonlinear control method with high-order disturbance compensation is proposed in this thesis.First,based on the three-element model,system uncertainties,unmodeled dynamics,and external disturbances are regarded as a lumped disturbance.Assuming that the lumped disturbance has high-order bounded derivatives,a nonlinear high-order disturbance observer is designed to estimate the lumped disturbance and its derivatives.The observation errors are guaranteed to converge to zero in finite time.Furthermore,a nonlinear robust controller is designed to achieve the asymptotic convergence of tracking errors.Moreover,the stability of the closed-loop system is proven theoretically by using Lyapunov techniques.Finally,hardware experiments are implemented on a horizontal single PAM testbed to validate the effectiveness and robustness of the proposed method.3)Repetitive learning control for single PAM-actuated exoskeleton robots with saturation.Considering that most of the control methods of PAM systems do not have the ability of “learning”,a repetitive learning law with saturation is designed in this thesis,which can learn and compensate the unknown periodic function in the dynamic model.Then,a nonlinear robust controller is proposed,which can not only drive tracking errors to converge to zero,but also limit them within their preset bounds during the entire control process.Moreover,the stability of the closed-loop system is proven theoretically by utilizing Lyapunov techniques.Finally,hardware experimental results are carried out to validate the effectiveness and robustness of the proposed repetitive learning method.