Walking Research On Lower Extremity Exoskeleton Rehabilitation Robot Based On Dynamic Motion Primitive With Reinforcement Learning

Exoskeletons can be described as a kind of wearable robots,which become an integrated system combining human body and robotic machines by both human intelligence and strength and endurance of robots.At present,exoskeleton robots are widely used in medical,military and other fields.In this paper,a lower extremity exoskeleton rehabilitation robot is proposed to serve people with a lack of lower extremities.There are total of six active degrees of freedom for hip and knee joints and two passive degrees of freedom for the ankle joints in this robot.It can drive people with lower limb movement capacity to stand and walk at a constant speed,to help them reintegrate into society and have better life.The main research contents of this paper are as follows:(1)For the walking task of the lower extremity exoskeleton rehabilitation robot,this paper proposes a new trajectory learning program based on the reinforcement learning algorithm(RL)and the dynamic movement primitive algorithm(DMP),which aims to provide assistance for the walking of the wearer.In the proposed strategy,a two-level planning is designed.In the first level,the inverted pendulum approximation under the consideration of the locomotion parameters is utilized to guarantee the zero-moment point(ZMP)within the ankle joint of the support leg in the phase of single-leg support,and designed the trajectory of the exoskeleton rehabilitation robot in the mission space,then,we transform the trajectory of the mission space into the trajectory of the joint space.In the second level,the joint trajectories are modeled and learned by DMPs.Meanwhile,the RL is adopted to learn the trajectories for eliminating the effects of uncertainties in joint space.The experiment based on a lower-limb exoskeleton robot demonstrates that the proposed scheme can effectively suppress the disturbances and uncertainties.(2)An adaptive control scheme by incorporated fuzzy control approaches into exoskeleton system is developed to help the leg movement on a desired periodic trajectory and handle periodic uncertainties with known periods.The proposed control approach does not require a muscle model and can be proven to yield asymptotic stability for a nonlinear muscle model and an exoskeleton model in the presence of bounded nonlinear disturbances(e.g.,spasticity,fatigue).(3)Under the guidance of the doctor,four patients with spinal cord injury were selected as volunteers to carry out the experiments.The recovery strategy of the lower extremity exoskeleton rehabilitation robot was proposed,and finally the application was verified through experiments.