Biomedical Engineering Reference
In-Depth Information
wrist flexion/extension and forearm pronation/supination. PLEMO-P1 (Kikuchi
et al. , 2009), having 2-DoF force-feedback function in its working plane and 1
uncontrollable DoF to adjust the inclination angle of the working plane, used
electrorheological fluid brakes to generate resistive forces during active move-
ments. Exercise Machine for Upper Limb (EMUL) (Koyanagi et al. , 2003), using
electrohrheological fluid actuators and having a similar mechanism as PLEMO-
P1, supported active 3D movements. The robot was said to be safer because of its
specific actuators. ACtive REhabilitation robotic device (ACRE 2) (Doornebosch
et al. , 2007) had 5 DoF and supported 4 modes of 3D movements with impedance
control. GENTLE/S (Coote et al. , 2003), combing a commercial robot (Haptic-
MASTER robot, Moog-FCS B.V., The Netherlands) and virtual reality technology,
implemented admittance control and was able to assist free active 3D movements.
ACT 3 D (Sukal et al. , 2005) was based on the same HapticMASTER robot and had
similar functions. REHAROB (Toth et al. , 2005) combined 2 6-DoF industrial
robots and was implemented with 3D passive movements. As the technology
has advanced, more complicated robots with multiple degrees of freedom were
designed. NeRebot (Fanin et al. , 2003) was a 3 DoF wire-based robot for 3D
passive movements and its successor MeriBot (Rosati et al. , 2005) was a 5 DoF
wire-based robot that was augmented with 2 more mechanical DoF and had a
larger working space. Some of the more recent robots are in the category of the so-
called exoskeleton. ARMin (Nef et al. , 2007), possessing a haptic display and semi-
exoskeleton kinematics with 4 active DoF and 2 passive DoF, could deliver patient-
cooperative arm therapy. Rehabilitation modes with both active and passive
movements in 3D space were implemented and the controller was basically a PD
controller for torque control. Robot assisted UPper Extremity Repetitive Therapy
(RUPER T TM )(He et al. , 2008), using “McKibben” type pneumatic actuators, was
a light-weighted and portable exoskeleton robot. It provided 5 DoF movements
(shoulder elevation (flexion), elbow extension, humeral external rotation, forearm
supination, and wrist/hand extension). The controller consisted of a modified
PID feedback controller with a nonlinear filter. Additionally, three of the joint
controllers had a feedforward controller built upon iterative learning algorithm in
parallel. It is noted that most of these robots, irrespective of their DoF, are aimed
at rehabilitation of proximal joints, i.e., shoulder and elbow joints.
Additional problems in designing controllers for rehabilitation robots are the
uncertainty about the biomechanical properties of subjects' upper limb, the non-
linear dynamics of robots such as the Coulomb friction in the mechanism and the
interaction with another unresolved controller (the human control). In a classical
design process, the control parameters are determined according to the system
model and its parameters. For a robot-assisted rehabilitation program, the subject
is also part of the dynamic system and the dynamic model of the subject is not as
clear and invariable as the mechanical system. Fuzzy control(Wang, 1994; Rugh,
1991; Lee et al. , 1998) is known for coping with nonlinear systems and systems
that have uncertainty in its parameters. Yet, a main problem associated with fuzzy
control is how to guarantee the global stability (Tanaka and Wang, 2001). Many
 
Search WWH ::




Custom Search