Design And Control Of Variable Stiffness Actuators Assisting Human Body

The current research on assisting exoskeleton equipment is developing rapidly,and it is very valuable in solving the difficulties of the moving in elderly,assisting in highintensity labor,and enhancing military individual combat capability.In general,the exoskeleton is driven by motor directly or the series spring actuator(SEA).Due to the human joints stiffness dynamic changes in different occasions,it is easy to bring shock and frustration to human when it is assisted joints,even disturb the natural gait.This not only imposes stringent requirements on the control strategy design,but also makes it difficult to increase the assisting force under safety considerations.To solve this problem,a large amount of experimental data needs to be used to explore the problem of matching the stiffness of the external actuator and the human joint.As a physical variable stiffness actuator(VSA),the VSA has the characteristics of dynamic stiffness adjustment and high stiffness bandwidth,which is suitable for simulation matching experiments.Therefore,this article designed a portable variable stiffness actuator based on leaf spring,with the control system can automatically identify the human walking phase state,dynamically adjust the internal stiffness of the actuator while assisting the human body.Hence we can effectively perform dynamic stiffness to study the matching problem between human joint and exoskeleton.According to the lower limb motion parameters of human body,this paper first clarifies the process of stiffness matching experiment,and proposes the output torque and the dynamic response standard.Then based on the two-degree-of-freedom movement of the end point of the four-bar linkage mechanism,using the leaf spring as the elastic element,the mechanical principle is established and the main theoretical structural parameters are calculated,at last,the theoretical dynamic model is established and analyzed.After that,this article uses Abaqus to extract the modal of the leaf spring and generate modal neutral file to establish Adams simulation model.On this basis,the torque-stiffness one way loading experiment and the reciprocating loading experiment are carried out on the variable stiffness simulation model.After analyzing the nonlinearity of the actuator in the experiment,establish the stiffness data table and working range interval according to the experimental data.Then the paper starts with the CPG phase recognition algorithm,uses the perturbation theory to expand the CPG unit in the first order,and obtains the phase convergence formula to verify the phase convergence when using the hip joint angle as the input signal at the same frequency with CPG.Under this premise,the second-order expansion of the CPG unit is carried out,the frequency convergence criterion under the circular limit cycle is analyzed,Simulink is used to verify the validity of the criterion,and the criterion is further extended to the complex limit circle.Finally,based on the above criteria,using feedback structure to remove the mean value of input signal,and complete the phase recognition algorithm for the human hip joint,at last the actual human flat walking data is used to verify the convergence.Finally,this paper analyzes the results based on the kinetic equations.First,the common mode angle feedforward torque compensation and PID differential mode angle control structure are designed for the four-bar linkage.Among them,the stiffness mapping function is established according to the interpolation of the stiffness data table,the parameters of each inertial link are established by using Matlab and Adams joint simulation,and the gradient table is calculated according to the stiffness mapping table to complete the variable stiffness torque feedforward compensation.Carry out the simulation experiment under the condition that the output end is fixed.Then,the identified inertial link and the stiffness characteristics of the leaf spring are linearized at current working point,regard the human body motion information as a disturbance signal,combined with the simplified human-exoskeleton interaction model to calculate the model predictive control quadratic planning objective function.After that,using step response to get the control parameters.Finally,the control strategy is verified based on Adams-Simulink co-simulation.