Adaptive Impedance Control For A Robotic Exoskeleton

With the rapid development of robot kinematics,control theory,image processing technology,biomedical technology,robotic exoskeleton has attracted the attention of many researchers.In the recent decades,exoskeleton robot,in combination with clinical medicine,could enhance the human's muscle force and help people carry heavy loads for long.As a novel rehabilitation medical mean,exoskeleton robot holds significant potential in rehabilitation training,intelligent control and other fields.In this paper,for the developed exoskeleton robot system,we propose the adaptive impedance controllers to realize the control under the secure,stable and reliable conditions.Based on the related researches,this paper proposes two kinds of adaptive impedance controllers,and realized the two controllers in the developed exoskeleton robot platform.One is adaptive impedance control based on surface electromyography(s EMG)signals.By collecting the sEMG signal,we could explore deeper mechanisms of neuron control,such as impedance regulation mechanism.First,we constructed sEMG-Driven Musculoskeletal Model and got the human's impedance information through this model.Then transformed the human's impedance information to the robot to design the desired impedance model.Considered the unknown deadzone input nonlinearity and the absence of the precise knowledge of the robot's dynamics,we employed the radial basis function neural network(RBFNN)to approximate the deadzone effect and robot's dynamics.For the existence of unmeasured states,a high-gain observer was employed to mesure the unmeasured states for output feedback control design.The other is adaptive impedance double-loop control based on reinforcement learning.The double-loop control framework decouples the control loop into a robot-oriented inner loop control and a task-oriented outer loop control.Reinforcement Learning algorithm is a kind of machine learning algorithm that different form supervised learning.It doesn't need a priori knowledge,and improves self-learning and on-line learning ability through constant trial-and-error interaction with the environment.We learned the optimal robot-human impedance model through reinforcement learning algorithm.Then we employed the model reference adaptive control method,design the impedance controller.In addition,considered the state constraint problem,a novel barrier Lyapunov function(BLF)was employed to guarantee the performance of the closed-loop system.Finally,the validity and reliability of the two methods are verified through relevant experiments.