Page 283 - Control Theory in Biomedical Engineering

Exoskeletons in upper limb rehabilitation 255

Chien, 2010). As solution, a time delay estimation (TDE) is proposed
(Youcef-Toumi and Shortlidge, 1991; Youcef-Toumi and Ito, 1990;
Brahmi et al., 2017). With this method, it is sufficient to delay the
output-input of the system only one step to provide a good approximation
of the unknown uncertainties dynamic model of the exoskeleton robot.
Nevertheless, the TDE approach suffers from time delay error (TDR)
caused by the noisy measurements and hard nonlinear function of the
robot model during delay constant, which degrades the approximation
performance.
On the other hand, many other works have used decentralized control
for these types of robotic systems as in Luna et al. (2016) and Ochoa Luna
et al. (2015). A decentralized adaptive control, based on the virtual decom-
position approach, was proposed, where the whole system was decomposed
virtually into several individual subsystems. This decomposition makes the
parameters, adaptation, and control law very easy. As an example of these
works that applied on other type of robots, an adaptive tracking control
design for an uncertain mobile manipulator dynamics based on appropriate
reduced dynamic model was suggested in Aviles et al. (2012). An adaptive
controller based on the backstepping technique (Brahmi et al., 2016) was
implemented to the trajectory tracking of the wheeled mobile manipulator.
Recently, approximation-based control strategies like fuzzy logic and neural
networks have been used to learn the exoskeleton dynamic model (Chen
et al., 2015; Li et al., 2015). However, through these approaches only uni-
formly ultimate boundedness of the tracking errors was achieved. Mean-
while, the estimated weights were not reached to their actual values. This
might reduce convergence speed during weights training operation, which
stops the approximation-based control for real-time implementation.
It is remarkable from a natural human movement (since the human upper
limb is attached with the exoskeleton robot) that the human does not need
accurate information about kinematics and dynamics of the arm (or any
object carried by upper extremity) to reach an object in space. Due to that,
many control strategies have been designed to solve the problem of kine-
matic and dynamic uncertainties (Arimoto, 1999; Cheah, 2006; Yazarel
and Cheah, 2002; Huang and Chien, 2010; Cheah et al., 2005; Hutchinson
et al., 1996). The main innovative point of these controllers is that the adap-
tation of the both kinematic/dynamic uncertainties has been provided,
which allows the exoskeleton robot to perform the human-like motion
and supplies to the control system more flexibility to handle the uncertainties
and parameters variation. However, the aforementioned controllers are