Model-based feedback control of a rehabilitation robot

Abstract

In this paper, a model-based nonlinear feedback controller for the joints of a novel motorized orthosis is presented. Powered by artificial pneumatic muscles, this orthosis has been developed for locomotion training based on the enhancement of neuroplasticity in patients with neurological disorders. The controller scheme draws inspiration from the innate control strategies of human motor control and is based on model descriptions of the mechanical system, the pneumatical system together with the artificial muscles. The feedback, which includes the inverse models of the mechanical setup and the pneumatic muscles, is used along with the obtained sensor signals to calculate the mass of the driven body parts. With the obtained pseudo-mass, regular models adulterated by occurring disturbances can be corrected and the required inner air pressures of the artificial muscles for realization of a given trajectory can be calculated. The use of a nonlinear pressure controller derived by backstepping combined with an inverse valve model of the proportional-valves guarantees a stabilization of the desired pressures in the muscles under different working conditions and the ability to track the given trajectory with sufficient accuracy.