Markovian robust filtering and control applied to rehabilitation robotics

Abstract

In this article, we deal with rehabilitation robot robust control in order to obtain best performance in the human-robot interaction. This class of system is subject to abrupt changes due to exogenous inputs and parameter variations caused, mainly, by the human dynamic behavior. In this context, Markovian jump linear systems offer suitable tools to model and control of this class of interactions. We propose force and impedance controllers based on robust recursive Markovian versions of the standard Kalman filter and the linear quadratic regulator. Different levels of the individual's actuation during the interaction with the robot define the Markov states. The measurement of electromyographic signals is used as jump parameter among Markov states. Also, a serious game is used to generate visual feedback, to promote active user participation, and to guide his movement routine. We compare our proposal with two other Markovian-based approaches and show the effectiveness our method through experiments with an ankle rehabilitation platform.