Research On Parameters Identification And Control Strategy On Lower Extremity Exoskeleton

An exoskeleton is a type of wearable robot with a wide range of applications in the fields of military,medical rehabilitation and civil power assistance.This paper presents research on sensing systems and control algorithms based on aspects of the Serial Elastic Actuator(SEA),dynamic model,Sensitivity Amplification Control(SAC)and online reinforcement learning parameter optimation.In order to enhance the wearing comfort and improve the bionic characteristics of the robot,this paper discusses improvements made on the base of traditional joints of the exoskeleton.The elastic element applied between the joint drive motor and load is to reduce impact and restore motion energy when walking.The mathematical model is established at the outset,followed by determination of the proper stiffness of the elastic element by analysing certain dynamic characteristics of the elastic joint.Finally,the control algorithm,proposed in the paper and paremeters identified through the experiment,are verified on the exoskeleton platform.In order to simplify the sensing system to the greatest possible extent,this paper selects the SAC as the motion control algorithm.It does not require any human-robot interaction force sensors to evaluate the intention of human motion,but instead adopts higher requirements for accuracy of parameters in the kinetic equation.In this paper,the kinetic equation is derived from the Lagrange equation.Factors such as the joint friction torque and motor rotor inertia are some of the factors taken into consideration in order to ensure accuracy of the model.The link quality and centroid positions are identified through experimentation.The dynamic paratemers calculated through the experiment are used in the SAC.Upon successully completing the fixed sensitivity amplification coefficient control experiment,an online learning optimisation sensitiviy coefficient algorithm is designed to further optimise the robot servo effect.While the SAC is adopted in the lower layer,the DMP trajectory planning is combined with reinforcement learning and used in the upper level to optimise the sensitivity coefficient online.The DMP algorithm is used to study the human joint motion track and provide a reference trajectory.The control algorithm is simulated in MATLAB to verify its stability.Finally the Exoskeleton platform is set up.The sensitivity coefficient optimation algorithm and the accuracy of the parameters identified through experiment are verified.