Page 343 - Handbook of Biomechatronics

336 Borna Ghannadi et al.

rehabilitation robot. In Ding et al. (2010), a musculoskeletal upper extremity
model (without including muscle dynamics) was used to implement a
model-based assistive controller for a full-body rehabilitation exoskeleton.
At the University of Zurich (Switzerland), model-based arm weight
compensation is used inside the controller for “ARMin V” (sem-
iexoskeleton). The results of this study showed that with active model-based
gravity compensation, the patient’s effort will drop significantly.
The biological control structure of the CNS can be represented by an
nonlinear model predictive control (NMPC) with receding horizon. In
the NMPC, a forward dynamics model is used to generate gross optimal
movements, and feedback information is used for fine-tuning. NMPC is
used in a variety of applications in biomechanics (Mehrabi et al., 2017)
and automotive control (Maitland and McPhee, 2018). Recent progress
in the development of NMPC motivated researchers at the University of
Waterloo (Ontario, Canada) to control a rehabilitation robot using NMPC
with a nonlinear dynamic HRI model (Ghannadi et al., 2017). In this
research, the HRI model was confined within an NMPC of the single-robot
manipulandum (which is designed and developed by the Toronto Rehabil-
itation Institute (TRI) and Quanser Consulting Inc.). The proposed con-
troller used a musculoskeletal model of the upper extremity to predict
human movements and muscle activations (Mehrabi et al., 2017), thereby
providing optimal assistance to the patient. In this study, the controller suc-
cessfully predicts the muscular activations in model-in-the-loop simulations.
Model-based strategies for rehabilitation are more appealing than the
triggered-passive methods since they do not require patient preparation
for sensor attachment. However, the models should be identified within
an acceptable accuracy to ascertain the validity of bio-inspired information.
This accuracy should be achieved with a proper parameter identification
procedure that is done with the use of bio-sensors in pretests with the robot.
Thus, having a systematic approach for pretests and developing powerful
tools for parameter identification is a key element in the success of these
methods.

8 CONCLUSION

In this chapter, a review of upper extremity rehabilitation robots was
presented, considering their mechanical design, type of training, form of
rehabilitation, and control scenarios. Then, recent enhancements in the field
of rehabilitation robotics were introduced.