| Exoskeleton robot has been widely studied and applied in the field of medical and in-dustrial cross-rehabilitation,and the demand for intelligent and precise rehabilitation has been increasing.The intelligent control theory and method of exoskeleton robot based on human-robot interaction has become a research hotspot in recent years.The exist-ing control methods of exoskeleton robots ignore the dynamic change characteristics of the physical interaction between human,robot and environment,and cannot dynamically adjust the corresponding important control parameters according to the change,which directly affects the wearing comfort of patients in the process of using exoskeleton reha-bilitation training,human-robot movement coordination and rehabilitation training effect.The impedance control method is one of the important methods to realize the flexible in-teraction between human and machine by adjusting the internal relationship between the physical interaction force and the robot motion.However,the current impedance con-trol methods of exoskeleton robots still face some problems and challenges,such as de-pendence on precise system dynamics model,artificial parameter presetting,weak online learning ability and poor environmental adaptability.In view of the above problems,this dissertation carried out research on impedance control methods for lower limb exoskele-ton robots.By establishing a new physical human-machine-loop communion interaction model and designing an adaptive learning impedance control algorithm based on neural networks,the comfort and flexibility of human-machine interaction in the application sce-nario of rehabilitation training of exoskeleton robots were finally improved.The main research contents of this paper are as follows:(1)In view of the problem that the existing adaptive impedance control algorithm of lower limb exoskeleton relies on sensors to measure interaction force,the impedance control method for rigid exoskeleton robot is studied.An adaptive impedance control strategy of lower limb exoskeleton based on interaction force parameter estimation is pro-posed,and an online neural network weight updating law is designed.The impedance parameters can be adjusted adaptively when the interaction force changes,and the accu-racy of parameter estimation is guaranteed.Finally,the stability of the closed loop control system of exoskeleton is proved by Lyapunov theory.(2)The existing impedance model describing human-robot interaction only considers the interaction force between robot and environment,and lacks the interaction impedance model between exoskeleton,human and environment.Aiming at the interaction between rigid exoskeleton and human and environment,a human-exoskeleton-environment inter-action impedance model is constructed in this dissertation,and then the impedance control problem is transformed into an optimal control problem,and an event-triggered critic learn-ing impedance control algorithm is proposed to effectively solve the optimization problem.The proposed event triggering mechanism reduces the communication cost,and the critic neural network enhances the learning ability of the exoskeleton robot system.In this dis-sertation,based on the gradient descent method,an improved auxiliary term is introduced to reduce the requirement of initial admissible control.At the same time,the historical and transient state data are applied to the parameter updating law at the same time,which relaxes the continuous excitation condition.Finally,the stability of the closed-loop sys-tem is proved by Lyapunov theory,and the effectiveness of the algorithm is verified by numerical simulation.(3)For the exoskeleton system during rehabilitation training or assisting human walk-ing,the robot should operate stably within the safety range and satisfy the state constraints such as the range of joint motion and the magnitude of torque.In this dissertation,a state-constrained system is transformed into an equivalent state-unconstrained system by intro-ducing state error transformation technology,and an optimal control algorithm of hierar-chical critic learning is designed,which integrates the high-level trajectory planner and the lower-level critic learning controller.The algorithm can not only realizes the recognition of human motion intention,but also relaxes the limitations of the initial control conditions by introducing auxiliary terms in the update rate of the weights of the critic network.The stability of state-unconstrained systems is analyzed by using Lyapunov stability theory.Fi-nally,the effectiveness of the proposed learning impedance control algorithm is verified by numerical simulation.(4)In view of the problem that the impedance control strategy of the rigid lower limb exoskeleton model cannot be directly applied to the flexible driven exoskeleton model,this dissertation studies an adaptive impedance control strategy of the flexible driven exoskele-ton.Firstly,a flexible actuator model is proposed to adjust the stiffness of the external skeleton and reduce the vibration caused by the harmonic reducer.Then,the impedance model is constructed,the impedance control model is integrated into the flexible exoskele-ton model,the adaptive impedance controller is proposed,and the stability of the flexible driven exoskeleton system is analyzed by Lyapunov stability theory.(5)For the control problem of flexible exoskeleton robot in uncertain environment,a hybrid control method combining impedance-based perturbation observer,dynamic event-triggered control,H_∞control and experience replay technology is proposed.The proposed dynamic event triggering mechanism can flexibly adjust the update time of the controller,weaken the conservatism of the static event triggering mechanism,and save energy con-sumption.A single critic network framework is proposed to estimate cost function,event-triggered optimal control strategy and time-triggered disturbance strategy.In addition,the combination of historical data and transient data relaxes the conditions for continu-ous excitation.The stability of the uncertain flexible exoskeleton system is analyzed by Lyapunov stability theory,and the simulation results show the effectiveness of the hybrid control method in the uncertain environment. |
