Identification and Analysis of Human-Exoskeleton Coupling Parameters in Lower Extremities.

This paper proposed linear and non-linear models for predicting human-exoskeleton coupling forces to enhance the studies of human-exoskeleton coupling dynamics. Then the parameters of these models were identified with a newly designed platform and the help of ten adult male and ten adult female volunteers (Age: years, Height: mm, Weight: kg). Comparing the coupling force error predicted by the models with experimental measurements, one obtained a more accurate and robust prediction of the coupling forces with the non-linear model. Moreover, statistical analysis of the experimental data was performed to reveal the correlation between the coupling parameters and coupling positions and looseness. Finally, backpropagation (BP) neural network and Gaussian Process Regression (GPR) were used to predict the human-exoskeleton coupling parameters. The significance of each input parameter to the human-exoskeleton coupling parameters was assessed by analyzing the sensitivity of GPR performance to its inputs. The novelty and contribution are the establishment of the non-linear coupling model, the design of the coupling experimental platform and a regression model which provides a possibility to obtain human-exoskeleton without experimental measurement and identification. Based on this work, one can optimize control algorithm and design comfortable human-exoskeleton interaction.