Wearable Whole Body Exoskeleton Control Based On Adaptive Learning

As the number of patients with hemiplegia increases,there is an urgent need for some rehabilitation medical devices to provide supplementary treatment for patients.At present,the combination of robotics and rehabilitation medical engineering is a research hotspot at home and abroad.The wearable exoskeleton robot is a typical product of this combination,and a device that can interact with hemiplegia patients and provide supplementary training.In order to make exoskeleton robots provide better human-robot interaction and auxiliary training,more and more researchers have studied the related technologies,and have made a rapid development.At the same time,the study of exoskeleton robots is also faced with various problems,such as the unknown dynamics model,the existence of disturbances,and the optimization of humanrobot interaction.To solve these problems,we combine adaptive control with other intelligent control technologies and propose the corresponding controller based on Lyapunov theory.In order to solve the problem of model uncertainty,we employ adaptive control and fuzzy logic system.In order to solve the periodic disturbance,we introduce the iterative learning control method.In order to solve the problem of human-robot interaction,we introduce the impedance control method.The main research contents of this paper can be summarized as:1)The design of control system for the whole body exoskeleton robot.2)Analysis of kinematics and dynamics model for the exoskeleton robot.3)The dual-loop control structure was used to design the adaptive impedance control strategy of the upper exoskeleton robot.In the outer loop,the problem of finding optimal parameters is transformed into the LQR problem.Then,in the absence of knowledge of the robot model and the human arm model,the impedance model parameters are obtained online through the integral reinforcement learning.In the inner loop,an adaptive controller based on BLF is designed to solve the physical constraints,and also achieve a satisfactory tracking performance.4)Gait planning for the lower exoskeleton robot by cubic spline interpolation.5)In the design of control strategy for lower exoskeleton robot,fuzzy control is used to approximate the uncertainties in the system.At the same time,for the time-varying periodic disturbances,it is compensated by iterative learning.