Strategies & Challenges
Presented by: Rahele Shafaei
Table of contents
♠ FES Control
♠ an overview of some of the Existing
Strategies for Closed-loop Control
♠ Examples of Closed-Loop FES Controllers for
Regulating Knee Angle
♠ EMG Closed-Loop FES Control
Functional Electrical Stimulation
• FES uses short electrical pulses to generate
contractions in paralyzed muscles that exert
torques about the joint.
biphasic waveform: F=20–40 Hz, A= 0–120 mA,PD=0–300 μs.
• The neuron receives the series of pulses that are
delivered using electrodes:
transcutaneous (placed on the skin surface)
percutaneous (placed within a muscle)
epimysial (placed on the surface of the muscle)
cuff (wrapped around the nerve that innervates
the muscle of interest)
• FES is the most
commonly used Push Button
technique for improving
motor function in SCI
Differences between the production of tension in
neurologically intact and SCI individuals
Synchronous motor-unit stimulation
• Joint angle/torque can be controlled by
modulating the intensity of stimulation
delivered to the flexor and extensor muscles,
which actuate the joint in opposite directions.
• Hyperactive feedback loops in the CNS
CPGs and other spinal reflexes affect closed-loop FES control
because these phenomena effectively act as exogenous control
signals sent to the paralyzed muscles in parallel with the FES
control signal. Since these exogenous control signals can
disrupt the desired joint movement, the FES control system
must compensate for these unintended control signals.
Control in clinical FES systems
Require continuous or repeated user input, which means
that the user must devote his or her full attention to operating
the FES device.
• Finite state (closed-loop)
Execute a preset stimulation sequence in an open-loop
fashion when a specific condition is met.
Example:gait of stroke patients who struggle with drop foot
Typically do not correct for model errors or disturbances.
• Many potential applications require more
sophisticated real-time control of the stimulation as
well as closed-loop compensation for modeling
errors and disturbances.
– balancing during standing
– torso control during sitting, and walking.
• Closed-loop FES systems require less user
interaction, thus facilitating tasks
• The response of muscles to electrical stimulation,
which is nonlinear, time varying, and coupled, is often
accompanied by unpredictable perturbations in
people who have SCI.
• The sensors that are required for feedback can make
closed-loop FES systems cumbersome and time
consuming to attach and remove.
• Closed-loop control strategies have not gained
ground in clinical applications of FES technology.
In joint level control, the challenges are:
Nonlinear and time varying response
Spinal reflexes and perturbations due to
spastic muscle contractions
Controlling a highly coupled system
Many muscles are biarticular
Time delay of 10–50 ms between stimulation
and the onset of a muscle contraction
FES control system
muscle spasms &
Considerations when testing closed-loop
FES control systems
FES controller must be initially tested in isolation from
voluntary muscle contractions.
Subjects with SCI have to be trained before controller
testing approximate the muscles of a chronic FES user, so
provide a more realistic testbed for a clinical FES controller.
Individuals must be neurologically stable before they
are recruited (after at least 12 months)
Standard performance measures must be reported
an overview of some of the
Existing Strategies for Closed-loop Control
A controller for FES-based unsupported
standing in a paraplegic subject
• The objective of the controller is to maintain a hip angle of 0◦.
• The control algorithm consists of a proportional-integral
differential (PID) controller in series with a nonlinear function
that relates PID output to the duration of the stimulation pulses.
• The performance of the controller is evaluated by applying a
disturbance that causes the subject to bend at the hip and
recording how quickly the controller rejects the disturbance.
• The results show that the controller provides a 41% reduction in
RMS error and a 52% reduction in steady-state error compared
to open-loop control.
• J.J. Abbas and H.J. Chizeck,1991
An FES system for unsupported standing that maintains
balance by stimulating the ankle flexor and extensor
muscles to regulate ankle moment
• Uses an H∞ controller to regulate the moment about the ankles.
H∞ control guarantees stability when the nominal models of the
plant and the uncertainties in the system are accurate and is
able to compensate for perturbations that are included in the
• The H∞ controller maintains stability during a series of ankle-
moment tracking and disturbance-rejection tests in
neurologically intact subjects.
• K.J. Hunt, R.P. Jaime, and H. Gollee,2001
• W. Holderbaum, K.J. Hunt, and H. Gollee,2002
A neuro-PID controller for regulating knee angle
• Uses an artificial neural network to map the nonlinear
relationship between the desired knee angle and the required
stimulation parameters and also uses a PID controller in a
negative feedback loop to compensate for tracking errors
caused by disturbances and modeling errors. The neural
network is trained using a conjugate gradient algorithm, and the
PID controller is tuned using the Ziegler-Nichols method.
• The neuro-PID controller achieves an RMS error of 5◦.
• G.C. Chang, J.J. Luh, G.D. Liao, J.S. Lai, C.K. Cheng, B.L. Kuo, and T.S. Kuo,1997
Examples of Closed-Loop FES
Controllers for Regulating Knee Angle
Uses the inverse knee model as a compensator
Closed-loop PID controller
Combines the inverse knee model with a PID controller
Uses the inverse knee model to deliver a stimulation signal to both the
plant and the direct model, so that the direct knee model functions as an
observer of the plan.
Comparison (M. Ferrarin, F. Palazzo, R. Riener, and J. Quintern ,2001)
• The RMS errors for each controller when tracking a sinusoid:
4) less than 10◦ after 2 min of adaptation.
• The average lag for the same tracking task:
1) 0.18 s
2) 0.29 s
3) 0.18 s
4) not reported because the lag changes during the adaptation process.
• The feedforward-feedback controller performs best. But the inverse
model is imperfect because it neglects noninvertible model
EMG Closed-Loop FES
• A novel assistive system with the minimum effect on the voluntary
• EMG is adopted as the sensing feedback information to regulate
• A two-stage filter is proposed to process the raw EMG signal.
The first stage removes the artifacts in the raw EMG signal contaminated by
The second stage filter separates the high frequency tremulous EMG from the
low frequency voluntary components.
• The extracted tremor EMG of biceps and triceps will then be used
as control input in the FES controller to stimulate the two muscles
• Physical and drug therapy cannot provide a
• Sensors widely used in the system for tremor
suppression are accelerometers, gyroscopes,
goniometers and force transducers
• In using EMG for FES application, the stimulation
pulses will contaminate the natural EMG eliminate
Muscle response (M-wave)
• Reducing the tremulous motion while preserving the
• So the key problem involved with the tremor
suppression is how to distinguish the tremulous
component from the voluntary motion.
• The tremor EMG will be used to control the FES.
System diagram of EMG
controlled FES for pathological
Surface EMG from biceps and
triceps recorded and filtered, the
tremor EMG are used to control
the stimulation for biceps and
triceps reciprocally, in order to
attenuate the tremulous motion
and minimize the effect on the
1) The raw EMG data is collected on healthy subjects.
2) Without the use of electrical stimulation, the patients
are tasked to perform similar movements.
3) Raw EMG data is measured from patients during
electrical stimulation. Filter algorithms developed in
steps 1 2 are applied to process the data.
4) Controller design, the amplitude of electrical pulse
can be controlled directly by the tremor EMG.
The key work of the whole system is about the EMG
Two-stage filter for EMG signal processing for
pathological tremor suppression via FES
In the 1st stage, artifacts caused by stimulation are filtered and
natural EMG is chieved; in the 2nd stage, tremor EMG is
distinguished from volitional EMG.
Arrangement of the electrodes Experimental setup for EMG
recording under FES
Results for 1sth filter
performance with regard to removal of the artifacts performance with regard to removal of the artifacts
caused by FES during voluntary movement caused by FES during voluntary movement
• CHERYL L. LYNCH and MILOS R. POPOVI: Functional Electrical
Stimulation,CLOSED-LOOP CONTROL OF INDUCED MUSCLE
CONTRACTIONS, IEEE CONTROL SYSTEMS MAGAZINE,April 2008
• Dingguo Zhang and Wei Tech Ang, Reciprocal EMG Controlled FES for
Pathological Tremor Suppression of Forearm, Proceeding of the 29th annual
International Conference of the IEEE EMBS, Cité Internationale, Lyon, France
August 23-26, 2007.
Models of the Response of
Electrically Stimulated Muscle
Physiological models Black box models
• Accurate • Reproduce the input-
• complex output behavior of real
• Specific to a particular muscle
subject • Their structure does not
• Values of some of the necessarily reflect the
anatomical and physiology of muscle.
can be difficult to obtain
Two column bullet points
• Bullets go in here • And also in here
Example of a table
Note: PowerPoint does not allow
have nice default tables – but you
can cut and paste this one #
Sample Graph (3 colours)
1st Qtr 2nd Qtr 3rd Qtr 4th Qtr
• Bullet 1
• Bullet 2
Examples of default styles
• Text and lines are Table
• Hyperlinks like this
• Visited hyperlinks
Text & Title Accent & Followed
Background Shadows Fills Accent
Lines Text Hyperlink Hyperlink
Example Bullet Point Slide
• These templates are for personal use
only and must not be distributed, sold or
displayed on the web by anyone other
than Presentation Helper.
– You can find many more free PowerPoint
templates on the Presentation Helper