Page 340 - Handbook of Biomechatronics

Upper Extremity Rehabilitation Robots: A Survey 333

control scenarios. In this section, we focus on recent advancements in the
control strategies for upper extremity rehabilitation robots with different
mechanical designs, including single- and multirobot manipulanda, mobile
exoskeletons, and semiexoskeletons.

7.1 BCI-Based Strategies for Control and Rehabilitation

Methods for recording electrical (e.g., EEG) or magnetic fields (e.g., func-
tional magnetic resonance imaging (fMRI) and functional near-infrared
spectroscopy (fNIRS)) are used to monitor brain activities. Studies have
shown that the intention to perform a specific physical activity generates
consistent EEG patterns in BCI (Liu et al., 2012; Xu et al., 2014). BCI
may recover brain plasticity and motor function by means of focused atten-
tion on and guidance of activation patterns of brain signals (Daly and
Huggins, 2015; Yao et al., 2017). This feature motivates the application
of BCI in rehabilitation robotics. Recent advancements in real-time signal
processing, identification of new brain signal patterns, widespread accep-
tance of BCI, and less-satisfactory intense rehabilitation methods have
increased the interest in BCI deployment.
BCI-based rehabilitation studies (Ang and Guan, 2015, 2017; Ang et al.,
2015) at the Nanyang Technological University (Singapore) have led to
well-established results in the use of BCI for rehabilitation robots. In Ang
et al. (2015), they used the “MIT-MANUS” (single-robot manipulandum)
with their proposed EEG-based motor imagery BCI (BCI-MANUS
therapy) and compared the rehabilitation results with MANUS therapy.
In the MANUS therapy, poststroke subjects performed self-paced vol-
untary reaching movements. The robot assisted the subject if there were
no detectable movements from them after a 2-second interval. Prior to
the BCI-MANUS therapy, the robot was calibrated based on the recorded
EEG signals when the subject was asked to imagine a voluntary reaching
movement while the robot’s end-effector was locked in its position. Then,
in the BCI-MANUS therapy, the subject was asked to imagine voluntary
reaching movements with minimal voluntary movements. Based on the
trained subject-specific motor imagery results, the robot manipulated the
subject’s arm toward the target.
Results of the study showed that the BCI-MANUS therapy is more
effective than the MANUS therapy. Furthermore, despite the reduced num-
ber of repetitions (i.e., less intensity) in the BCI-MANUS, it results in motor
gains similar to more intense robotic therapy. Although BCI-based