Implementation of AI based approach for disturbance estimation of a MIMO nonlinear system

Abstract

Postural control is the ability to maintain equilibrium by keeping or returning the center of gravity (COG) over its base of support (BOS), and it relates to how the body’s position in space controls for stability. The center of gravity (COG) is a point at which all an object’s mass can be concentrated in relation to gravity. The postural control system serves as a feedback control circuit between the brain and the musculoskeletal system. The internal dynamics of a system model are one of the major functional components that the posture control system relies on. So, the modeling of CNS will be represented by an extended high gain observer (EHGO) which is based on a feedback linearization controller. Basically, EHGO works as a disturbance estimator and a soft sensor of the internal dynamics, respectively. Moreover, AI approach contributes to a better knowledge of the postural control and STS mechanism. Second part of this research focus on traditional machine learning approach used to improve robotic and exoskeleton design. By using head positions of different experimental objects, regression model will predict the positions of ankle, knee, and hip joints. Therefore, on head positions defned as input and position of joints are outputs of the model. In this research supervised learning is used because inputs and outputs are defned or known. So, the techniques used under supervised learning are random forest regression, support vector regression (SVM), decision tree regression.