Closed-loop Identiffication of a Multivariable Dynamic Knee Rig. Issue 28 (2015)

• Record Type: Journal Article
• Authors: Chevalier, Amélie; Hernandez, Andres; Ionescu, Clara M.; Verstraete, Matthias; De Keyser, Robin
• Is Part Of: IFAC-PapersOnLine. Volume 48:Issue 28(2015)
• Page Start: 116
• Page End: 121
• Publication Date: 2015
• Subjects: Closed-loop identiffication -- Prediction error methods -- ARMA parameter estimation -- Multi-input/multi-output systems -- Control -- Biomedical systems; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2015.12.110
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: This paper presents the closed-loop identiffication of a multivariable dynamic knee rig, which is used to evaluate the performance of newly developed knee prosthesis. Unlike standard dynamic knee rigs, the system used in this research has the capability to impose cyclic movements on the knee joint, making it a multivariable system, i.e. multi-input/multioutput (MIMO) system. To obtain a simpliffied model of the system, interaction between the multivariable outputs is neglected. As the system is unstable in open-loop, an indirect closedloop identiffication method is used to construct a model of the system. Parametric identiffication of the closed-loop system is performed using a speciffic version of the prediction error method: the ARMAX method. For each of the obtained models of the multivariable system the normalized root-mean-square error (NRMSE) value was obtained which quantiffies the goodness of ffit of the model compared to the experimental data. Based on the closed-loop models, a model for the open-loop system was derived and validated against the measured system's response. The results indicate that the identiffied models correspond to the measured signals and validation is obtained. However, improvements to the models are possible when taking into account the interaction between the variables of the MIMO system which will be the focus of future research. Future work also consists of developing an advanced controller based on the identiffied models to control the system's outputs.