MODEL PREDICTIVE CONTROL OF EXOSKELETON JOINT ANGLES

Abstract

Lower limb exoskeletons are used for rehabilitation purposes, so their precise control is an important issue. Classical controllers are often unable to provide high accuracy and speed, which can lead to errors in exoskeleton operation. Complexities associated with lower limb exoskeleton control systems arise from the dynamic and nonlinear nature of human locomotion. Solving these complexities involves designing modern control systems. One of the modern approaches is Model Predictive Control, which, predicting future deviations of system state variables, performs optimization calculations and selects such input signals, in which case the error will tend to the minimum value. This controller is widely used for controlling nonlinear dynamic systems with constraints. Considering this fact, it was proposed to develop a controller with a predictive model for controlling the angles of the exoskeleton joints. Control with a predictive model assumes that the dynamic equations of the control object are known, so a nonlinear mathematical model of the lower limb exoskeleton was firstly designed, taking into account the mechanical and electromechanical components. An examination of the exoskeleton state variables and control input constraints, which were taken into account during the design of the controller, was performed, and a reduction function was defined, according to which the controller with a predictive model should perform an optimization calculation, thereby enabling the reduction of errors or inaccuracies. Simulations performed in the MATLAB/Simulink environment validate the effectiveness and accuracy of the developed controller, showing small deviations between the desired and actual joint angles during dynamic motions. Future efforts will focus on improving control quality and validating results through empirical studies, thereby strengthening the role of model predictive controllers for the design of exoskeleton control systems.