Study On Torque Planning Algorithm Of Lower Limb Assist Exoskeleton For Walking Speed Difference

Power-assisted exoskeletons for the enhancement of human physical performance have broad application prospects in military,medical,and industrial fields.However,the current power-assisted exoskeleton control strategy relies heavily on the wearer's body size and motion state,which cannot meet the adaptability of the human-machine system during exercise,hindering its application promotion and improving the efficiency of human-machine collaboration.Therefore,this thesis proposes a general exoskeleton torque planning control strategy to solve the problem of exoskeleton assist self-adaptation for the change of human motion state caused by environmental objective factors or human subjective factors.A unified torque planning method is established by constructing a torque planning method based on Kernelized Motion Primitives(KMP)and directly modulating and generating the wearer's personalized assist output trajectory according to the state of human motion.The main research contents of this thesis are as follows:1.Aiming at the difference of knee joint torque caused by human body size and motion state,an exoskeleton KMP assist torque model based on personalized knee joint torque curve is proposed.Simplified application of high-dimensional multi-cluster trajectory probability extraction modeling for dimensionality reduction,collect joint torque data for probabilistic coding,and establish a knee joint assist torque model;evaluate its pros and cons based on model output error and covariance,and optimize by adjusting the number of coding cores and learning data distribution;Through test set verifies that the model has good generalization ability,and the absolute error of the average output of the trajectory is below 0.42 Nm.2.Aiming at the difference between the real-time output and the expected torque caused by the change of the human body motion state,a torque planning algorithm based on the dynamic modulation of the local kernelization probability is proposed to realize the matching of the assist torque and the human state.According to the human body attitude information,formulate the modulation rule of the kernelized probability and parameterize it to obtain the desired output;change the kernelized probability distribution of personalized assist torque by updating the reference database and weighting the reference trajectory respectively;compare and analyze the two planning methods for state synchronization,the simulation verification shows that the planning error can reach 1.42%and 1.28% of the expected torque threshold respectively;considering the overall consideration,the trajectory weighting method is selected for the system experiment and the overall algorithm simulation experiment is carried out.The results show that the synchronization calculation error is basically below 0.5 Nm.3.Validate the power assist efficiency of the algorithm through the lower limb power assist exoskeleton platform.The torque planning algorithm was used to assist the exoskeleton and a compared experiment was carried out.detect the wearer's heart rate and EMG signal,analyze the exercise consumption and EMG signal threshold to evaluate the assist efficiency of the algorithm,and the results show that it can save about 10%-15%energy consumption in some states.The torque planning algorithm based on local kernelized probility dynamic modulation can realize the desired trajectory modulation more flexibly,adapt to the change of the wearer's motion state,and improve the power assist efficiency of the exoskeleton,which has a wide range of application value.