1

DEVELOPMENT OF AN FES-BASED ANKLE ORTHOSIS

By

GANESHRAM JAYARAMAN

A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT

OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE

UNIVERSITY OF FLORIDA

2010

2

© 2010 Ganeshram Jayaraman

3

To my parents, M. Jayaraman and Geetha Jayaraman; my sister Ramya; and my

friends and family members, who constantly provided me with motivation, encouragement and joy

4

ACKNOWLEDGMENTS

I would like to express my gratitude to my supervisory committee chair and mentor

Dr. Warren E. Dixon for giving me the opportunity to work on this project and for the

advice and encouragement that he had provided me with during the course of my study

at University of Florida. I would also like to extend my gratitude to my co-advisor Dr.

Chris M. Gregory for helping me with the setting up and conducting the experiments at

the VA hospital and for his valuable inputs at various stages of the project. I also thank

my committee member Dr. Scott A. Banks for lending his knowledge and support.

I thank my colleagues for helping me with the thesis research. I especially thank

Dr. Qiang Wang and Nitin Sharma for sharing their knowledge and for helping me out

when I was conducting my experiments. I also thank members of the Human Motor

Performance Laboratory at the VA hospital for their help in running the Lokomat and the

treadmill.

Finally I would like to thank my parents for their understanding, encouragement

and patience.

5

TABLE OF CONTENTS page

ACKNOWLEDGMENTS .................................................................................................. 4

LIST OF FIGURES .......................................................................................................... 6

ABSTRACT ..................................................................................................................... 8

CHAPTER

1 INTRODUCTION .................................................................................................... 10

2 EXPERIMENTAL SETUP ....................................................................................... 15

2.1 Components Description ................................................................................ 16 2.1.1 Lokomat ............................................................................................... 16 2.1.2 Goniometer .......................................................................................... 16 2.1.3 Signal Conditioning .............................................................................. 16 2.1.4 Event Switch ........................................................................................ 17 2.1.5 Computing System ............................................................................... 17 2.1.6 Stimulation Circuit ................................................................................ 17

2.2 Robotic Orthosis Lokomat .............................................................................. 18 2.2.1 Lokomat Setup ..................................................................................... 18 2.2.2 Linear Drives ........................................................................................ 20 2.2.3 Lokomat Control Setup ......................................................................... 21

2.3 Goniometer Sensor Integration ...................................................................... 21 2.3.1 Angle Measurement ............................................................................. 24 2.3.2 Moving Average Filter Implementation ................................................. 24

2.4 Data Acquisition and Computing System ....................................................... 27 2.5 Event Switch .................................................................................................. 33 2.6 Stimulation Circuit .......................................................................................... 36

3 EXPERIMENTAL METHODS ................................................................................. 37

3.1 Desired Trajectory Setup ................................................................................ 37 3.2 Desired Trajectory Offset ............................................................................... 38 3.3 Experimental Procedure ................................................................................. 41

4 RESULTS AND DISCUSSIONS ............................................................................. 43

5 CONCLUSION ........................................................................................................ 53

REFERENCES .............................................................................................................. 54

BIOGRAPHICAL SKETCH ............................................................................................ 59

6

LIST OF FIGURES

Figure page 2-1 Various components of the experimental setup. ................................................. 15

2-2 Schematic diagram of a Lokomat setup [6]. ....................................................... 19

2-3 A person walking with the help of a Lokomat. .................................................... 19

2-4 Various adjustments made in the Lokomat are indicated by double sided arrows [6]. ........................................................................................................... 20

2-5 Control setup for the Lokomat [6]. ...................................................................... 22

2-6 Biometrics Ltd., goniometer sensors. ................................................................. 23

2-7 Goniometer sensor placement across the ankle joint. ........................................ 23

2-8 Block diagram for the implementation of Moving Average Filter algorithm. ........ 26

2-9 Raw goniometer reading from the ADC port with and without Moving Average Filter at steady state. .......................................................................................... 28

2-10 Calibrated angle reading from the goniometer at steady state plotted over a period of 0.25 sec. The final measuring angle is a combination of the effect of the first and additional 71 point MA Filter on angle measurement for 0.142 sec and after 0.142 sec respectively. ................................................................. 29

2-11 Calibrated angle reading from the goniometer at steady state plotted over a period of 15 seconds. The final measuring angle is a combination of the effect of the first and additional 71 point Moving Average Filter on angle measurement for 0.142 sec and after 0.142 sec respectively............................. 30

2-12 Raw goniometer reading from the ADC port with and without the Moving Average Filter while tracking............................................................................... 31

2-13 Calibrated angle reading from the goniometer while tracking showing the effect of Moving Average Filter. .......................................................................... 32

2-14 Motion lab system event switch. ......................................................................... 33

2-15 Digital input circuit. ............................................................................................. 34

2-16 Digital input reading of toe switch. ...................................................................... 35

2-17 Toe switch state .................................................................................................. 35

7

2-18 Stimulation circuit board ..................................................................................... 36

3-1 Toe switch counter for recording desired trajectory ............................................ 38

3-2 Desired trajectory recorded of the subject walking on treadmill with the support of the Lokomat. ...................................................................................... 39

3-3 Desired trajectory recorded of the subject walking on treadmill without the support of the Lokomat. ...................................................................................... 39

3-4 Coordinate system for measuring the ankle and knee angle .............................. 40

3-5 Actual measurement of the ankle and knee angle .............................................. 40

3-6 Experimental setup with the Lokomat and computing system ............................ 42

4-1 Actual and desired trajectory plot of a person walking on the treadmill without the support of the Lokomat. ................................................................................ 45

4-2 Error plot of a person walking on the treadmill without the support of the Lokomat. ............................................................................................................. 46

4-3 Switch state plot of a person walking on the treadmill without the support of the Lokomat. ....................................................................................................... 47

4-4 Stimulation voltage plot of a person walking on the treadmill without the support of the Lokomat. ...................................................................................... 48

4-5 Actual and desired trajectory plot of a person walking on the treadmill with the support of the Lokomat. ................................................................................ 49

4-6 Error plot of a person walking on the treadmill with the support of the Lokomat. ............................................................................................................. 50

4-7 Switch state of a person walking on the treadmill with the support of the Lokomat. ............................................................................................................. 51

4-8 Stimulation voltage plot of a person walking on the treadmill with the support of the Lokomat. ................................................................................................... 52

8

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science

DEVELOPMENT OF AN FES-BASED ANKLE ORTHOSIS

By

Ganeshram Jayaraman

May 2010

Chair: Warren E. Dixon Co-Chair: Chris M. Gregory Major: Mechanical Engineering

Individuals suffering from stroke and Spinal Cord Injuries (SCI) require

rehabilitation to perform normal gait like motion with more independence. Recently

developed robot assisted devices for rehabilitation includes a device powered with an

actuator at the joints and a simple control program to enable the limbs of a person to

track a predefined trajectory. Muscle training through Functional Electrical Stimulation

(FES) is another rehabilitation tool which involves the application of voltage to the

surface of the skin through electrodes, thereby producing a desired functional muscle

contraction. The first commercially available FES device for walking known as the

Parastep-I system involved open-loop stimulation wherein the stimulation current is

adjusted by the patient using two pushbuttons attached to the left and right handles of a

walking frame or crutches. One of the disadvantages of using FES assisted walking

device is that it is not possible to stimulate the Hip flexors directly and requires voluntary

effort which could be absent in paraplegic patients. The robotic orthosis Lokomat has

four linear actuators to control the two hip joint motions and the two knee motions and

four potentiometers to measure the joint angles. However the Lokomat does not allow

for the control of the ankle-foot complex. A good ankle joint articulation and control is

9

necessary for balancing while standing and walking. A hybrid orthosis device consisting

of a combination of the Lokomat and FES is developed where the ankle-foot complex is

controlled using FES. The advantage of using a hybrid orthosis device is that it greatly

reduces the number of degrees of freedom that contribute to the walking motion, and

hence, can negate the effect of muscle weakness from the remaining degrees of

freedom. The closed-loop proportional integral control method is developed using FES

which combines preset timing and tracking control. The toe event switch is used to

trigger the control program and the goniometer sensor is used as an ankle joint angle

feedback for tracking during walking. The controller was tested on a healthy individual

and produced the required plantar flexion moment during ankle push-off to prevent

individual’s from tripping on their toes.

10

CHAPTER 1 INTRODUCTION

More than 700,000 people suffer a stroke each year in the United States, and

approximately two-thirds of these individuals survive and require rehabilitation [1].

Stroke-related costs in 2005 were nearly $57 billion in the United States. The

Christopher Reeve Foundation estimates that up to 1,275,000 Americans may be living

in the United States with Spinal Cord Injuries (SCI) in 2008, and data from the National

Spinal Cord Injury Statistical Center [2] indicate projections of an annual incidence of 40

cases per million of population [3]. For 61% of the SCI population (under 30 years old)

the lifetime costs of SCI range from $624,441 for incomplete motor function at any

level to $2,801,642 for high quadriplegia [2].

The objective of rehabilitation is to help the survivors to perform normal gait like

motion with more independence. Such objectives can be achieved through robotic

assistance during rehabilitation. Typical robotic assistance would include a device

powered with an actuator at the joints and a simple control program to enable the limbs

of a person to track a predefined trajectory. This assistance can provide a more

quantifiable and repetitive diagnostic and training tool and could be realized through

advancements in measurement techniques to accurately measure the joint angles in a

local coordinate system and new robotic systems able to perform rehabilitation

efficiently and in a safe manner.

Early research in rehabilitation has been carried out using intense step training on

a treadmill using a body weight support and manually assisted movements for repetitive

tasks. The person’s weight is partially unloaded by a body weight suspension system,

and the person is placed on a treadmill with physical therapists helping to move the

11

limbs properly. Recently robot assisted devices have been developed to augment the

manual training methods. Robot devices can improve the quality of training, and assist

in performing repetitive tasks with less supervision. Robotic devices such as the

Lokomat [4-8], the similar AutoAmbulator [9], the Mechanized Gait Trainer [10-11], the

direct drive robot ARTHuR [12-14], and a robotic device for rodent gait training in [15]

are designed to be in constant contact with an individual and guide the individual’s legs

through a preprogrammed gait. Specifically, technologies such as [16-35] have been

developed as powered orthosis or exoskeletons for individuals with more walking

capabilities.

Muscle training through Functional Electrical Stimulation (FES) is another

rehabilitation tool applied to individuals suffering from stroke and spinal injury. Typically,

FES methods involve the application of voltage to the surface of the skin through

electrodes, thereby producing a desired functional muscle contraction. In addition to

moving a person’s limb with robotic assistance or therapist, FES methods can enable

muscle training. Although FES methods can serve as an adjunct for gait training, the

stimulation duration is reduced due to muscle fatigue. Early work by Kralj et al. [36-37]

in FES involved open-loop stimulation wherein the stimulation current is adjusted by the

patient using two pushbuttons attached to the left and right handles of a walking frame

or crutches. This technique was later utilized in developing the first commercially

available FES device for walking known as the Parastep-I system [38]. More advanced

systems have been developed in [39] and [40] focusing on automatic FES control for

assisted walking. The energy requirement by FES assisted walking has been studied in

[41]. In [42], it was reported that one of the disadvantages of using FES assisted

12

walking device is that it is not possible to stimulate the Hip flexors directly and requires

voluntary effort which could be absent in paraplegic patients. A stimulation control

method had been developed in [43] to stimulate the triceps surae for achieving better

push-offs in stroke subjects. The open loop control method in [43] is based on preset

timing and uses a uniaxial gyroscope to measure angular velocity which is integrated to

determine the shank angle that triggers the controller during swing phase. A PD and an

optimal closed loop control of an ankle joint using FES was developed in [44] and [45]

for balancing the leg during arm free standing. However, closed-loop control of an ankle

joint for tracking during walking requires gait phase detection and a dynamic ankle joint

angle sensor feedback. The closed-loop proportional integral control method developed

in this thesis using FES combines preset timing and tracking control. The toe event

switch is used to trigger the control program and the goniometer sensor is used as an

ankle joint angle feedback for tracking during walking. The robotic orthosis Lokomat

provides an initial testbed to investigate FES augmentation to treadmill training.

The robotic orthosis Lokomat has four linear actuators to control the two hip joint

motions and the two knee motions and four potentiometers to measure the joint angles.

The actuators are controlled by a PD joint motion control program. However the

Lokomat does not allow for the control of the ankle-foot complex. As stated in [46], good

ankle joint articulation and control is necessary for balancing while standing and

walking. Also most of the commercially available ankle-foot orthosis may have rigid

ankle stiffness which may not change with respect to change in a terrain or walking

speed [47]. In this thesis, a hybrid orthosis device consisting of a combination of the

Lokomat and FES is developed where the ankle-foot complex is controlled using FES.

13

In [42] it has been reported that the advantage of using a hybrid orthosis device greatly

reduces the number of degrees of freedom that contribute to the walking motion, and

hence, can negate the effect of muscle weakness from the remaining degrees of

freedom. Successful implementation of the FES ankle control requires developments in

sensor integration, control system design and stimulation parameter variations.

Sensors play a crucial role in developing the closed loop control system. They are

essential to measure the ankle joint angle which is then compared with a reference

trajectory to generate an error signal. Then the controller acts on this error signal to

produce the required control torque to satisfy the control objective. Hence, successful

sensor integration is a must for any closed-loop control system design. The proper

choice of a sensor depends on the type of application it is intended for. The FES ankle

control method has been carried out using an encoder [45] to measure the ankle joint

angle. However such measurements are made in a global coordinate system. Hence in

this project a position measuring goniometer sensor, Biometrics Ltd, is utilized to

successfully measure the joint ankle in the local coordinate system. This sensor is

attached directly to a person’s leg to measure the angle. The lightweight and good

dynamic response of the goniometer makes it suitable for tracking the ankle position

while walking or standing with toe-up.

However, most of the sensors in general are affected by noise and possible

delays. Both problems can be controlled to certain extent through careful design of

filters to remove the noise. While designing the filters care should be taken that it does

not introduce too much delay in the system. To reduce the noise in the goniometer

readings used in this thesis, a double pass 71 point Moving Average Filter is used. The

14

position measuring goniometer without filtering was not suitable for dynamic

measurement. The details of the developed moving average filter are discussed in

Chapter 2.

FES methods consist of applying a stimulation voltage across a muscle group

through stimulation electrodes to generate a desired muscle force. FES can be

implemented using invasive stimulation electrodes like needle electrodes wherein the

electrodes are inserted into individual muscle. However, efforts in this thesis used

surface electrodes where the stimulation voltage is applied to the surface of the skin.

The force produced in the muscle depends on the firing rate and the recruitment of the

motor units. Current and past research has focused on the effect of stimulation

parameter modulation on the force generated by the muscle. The muscle force can be

controlled by modulating the voltage amplitude, pulse width and frequency. The pulse

width modulation (PWM) and pulse amplitude modulation (PAM) controls the motor

recruitment whereas the pulse frequency modulation controls the firing rate of the motor

units [48-50]. Our Central Nervous system (CNS) uses a combination of motor

recruitment/firing rate to control the different muscles [51].

The objective of this thesis is to develop and test a hybrid orthosis device to

enable a person to generate a sufficient plantar flexion moment for ankle push-off to

prevent individuals from tripping on their toes. A toe switch is used to detect the mid

stance of the gait and also to trigger the stimulation voltage. The goniometer sensor

feedback is used to track a desired trajectory once the toe switch is on. These concepts

along with hardware description, experimental methods and computing system have

been discussed in the following chapters.

15

CHAPTER 2 EXPERIMENTAL SETUP

The FES ankle control was implemented using a PI control algorithm with pulse

amplitude modulation (PAM). The controller was tested on a healthy subject at the

Human Motor Performance Laboratory (HMPL) housed in the Brain Rehabilitation

Research Center (BRRC) located at the Veterans Affairs (VA) Hospital in Gainesville.

The various components used in the experiment are illustrated in Figure 2-1.

Figure 2-1. Various components of the experimental setup.

Ankle Angle Goniometer

Hip Angle

Knee Angle

Lokomat

Toe

Switch

A/D & D/A Board

Host PC Stimulator

Electrodes

16

2.1 Components Description

2.1.1 Lokomat

The Lokomat is a Swiss made robotic orthosis instrument primarily used for

rehabilitation. The Lokomat provides treadmill step training for patients suffering from

stroke and spinal cord injury. The Lokomat is equipped with four linear drives to move

the hip and the knee joints. The joint angles are measured using four potentiometers

and the drives are controlled by a PD position control program while the patient is

partially suspended by means of a harness onto the treadmill platform. The ankle joint is

not controlled by the Lokomat, hence it is held by a passive foot lifter that assists only in

dorsiflexion during stance phase. For this experiment the foot lifter is removed and the

plantar flexor moment of the ankle joint is controlled by FES of the calf muscle.

2.1.2 Goniometer

For the experiments in this thesis, a voltage controller is designed to stimulate the

ankle joint to track a desired trajectory. Any closed-loop control requires sensor

feedback to compare with the reference trajectories to generate an error signal. An

angular position measuring goniometer sensor developed by Biometrics Ltd, is used to

measure knee flexion and the angular position of the ankle. These sensors are attached

directly across the joint using double sided adhesive tape to measure joint angular

movement. Signals from the goniometer require additional instrumentation to process

the measurements.

2.1.3 Signal Conditioning

The goniometer is connected to an angle display unit (ADU) made by Biometrics

Ltd, to further process the signals. The ADU unit amplifies the signals from the

goniometer with a 9v DC battery. The measuring range falls between 0-4.5 V which has

17

to be calibrated to interpret the actual ankle joint angle. A ServoToGo I/O card is used

to process A/D and D/A conversions. The board can perform A/D and D/A operations up

to 8 channels and 32 bits combination of Digital I/O at a sampling frequency of 1000 Hz.

The goniometer signals were affected by noise and hence a double pass 71 point

Moving Average (MA) filter has been utilized to remove the noise. The MA filter is

implemented by software through a code written in C++.

2.1.4 Event Switch

The event switch used in this experiment is a product of Motion Lab System. The

event switch connected to the toe is used as an On-Off switch by connecting to the

digital input of a ServoToGo I/O card. When the toe is on the ground, the switch is on

and triggers the control algorithm and terminates once the switch is off.

2.1.5 Computing System

FES control methods and the sensor data acquisition and filtering are

implemented on a Pentium IV PC running on QNX, a deterministic operating system.

The control algorithm is implemented using Qmotor 3.0, a framework for implementing

control programs which is developed for the QNX operating system and uses Photon for

graphical user interface.

2.1.6 Stimulation Circuit

Control algorithms are encoded in C++ and an executable produces a voltage to

stimulate the calf muscle to move the ankle joint along a desired trajectory. The

stimulation voltage is generated using a custom made stimulation circuit board based

on the generated error signal represented as the difference between the desired

position and the actual ankle position measured by the goniometer. The voltage

controller is implemented through a Pulse Amplitude Modulation (PAM) with a fixed

18

pulse width of 100 µ sec and a fixed frequency of 100 Hz. A voltage to frequency circuit

was implemented to control the frequency of the pulse.

2.2 Robotic Orthosis Lokomat

2.2.1 Lokomat Setup

The Lokomat is a robotic orthosis device which can be fitted to the patient and

strapped around the waist and the legs. The patient is partially suspended by means of

a harness onto the treadmill platform. A schematic diagram of the Lokomat and a

picture of person mounted on the Lokomat are shown in Figure 2-2 and 2-3.

The passive foot lifter (Figure 2-2) assists in dorsiflexion of the ankle joint during

the stance phase. This leads to rigid ankle joint stiffness which may not be desirable

since ankle joint control is required for balancing in walking and standing. In this

experiment the ankle joint is controlled using a FES based proportional integral voltage

controller to stimulate the calf muscle while the knee and hip joints are controlled by the

Lokomat. The ankle joint angle feedback is measured using a goniometer which is

attached across the joint.

The upper body of a person has to be balanced in the vertical direction during

treadmill training and muscle stimulation to prevent slipping. A parallelogram is used to

attach the lokomat to the railings of the treadmill which enables the Lokomat to move

only in the vertical direction. This configuration provides an initial testbed to investigate

FES augmentation to treadmill training, while constraining movements to the sagittal

plane.

Any orthosis has to be flexible in its design since it has to be fitted to various

people with different anatomical features. Figure 2-4 illustrates the various adjustments

19

Figure 2-2. Schematic diagram of a Lokomat setup [6].

Figure 2-3. A person walking with the help of a Lokomat.

20

that can be made with the Lokomat such as width of the hip and position and size of the

leg braces. A small cushion pad is placed between the braces and the skin to prevent

skin soreness.

Figure 2-4. Various adjustments made in the Lokomat are indicated by double sided arrows [6].

2.2.2 Linear Drives

The Lokomat consists of four DC drives at the hip and the knee joints which are

driven by a precision ball screw (KGT 1234, Steinmeyer GmbH & Co., Germany) where

the nut on the screw is driven by a DC motor (MaxonTM RE40, Interelectric AG,

Switzerland) using a toothed belt with a Mechanical power of 150 W as indicated in [6].

The drives are designed to move the hip and knee joints along a gait pattern.

21

2.2.3 Lokomat Control Setup

The control setup of the Lokomat is shown in Figure 2-5 which consists of a user

interface, target PC and a controller. The therapist user interface is used to record all

the adjustments made in the Lokomat for a particular person which could be used to

setup the Lokomat quickly for the next training period. The user interface also has an

interface to the Lokomat using LabView software and can be controlled by the therapist.

The four joints comprising the hip and knee joints are controlled separately through

independent closed-loop proportional derivative position control to generate various gait

trajectories. The joint angles are measured using potentiometers and processed using

an analog to digital converter for real time use in the current controller. The position

controller is implemented using a real time system (RealLink TM) which runs on the

target PC. A serial bus (RS 232) is used to communicate between the host and the

target PC. The speed of the Lokomat can be changed at the user interface which is

transferred to the target PC to adjust the gait pattern and to change the speed of the

treadmill.

2.3 Goniometer Sensor Integration

Sensor feedback plays an important role in developing the closed-loop control for

the FES. In general, the sensors are used to measure the system response which is

then compared with a reference input to generate an error signal. Then most control

methods use this feedback to generate a control input based on the error signal with the

objective of reducing the error to zero. In the experiments in this thesis, the ankle joint

angle is used as feedback to develop a proportional integral controller to stimulate the

calf muscle. FES based ankle joint control has been carried out using encoder [45].

However measurements from such sensors are made in the global coordinate system.

22

Figure 2-5. Control setup for the Lokomat [6].

23

In the developed experiment, a goniometer angular position measuring sensor is used

to measure joint angles in the local coordinate system. The goniometer (Figure 2-6) is

attached directly across the ankle joint angle (Figure 2-7) using a double sided adhesive

tape.

Figure 2-6. Biometrics Ltd., goniometer sensors.

Figure 2-7. Goniometer sensor placement across the ankle joint.

24

2.3.1 Angle Measurement

The two end blocks of the goniometer are connected by a protective spring which

houses a thin composite wire inside. The composite wire has a series of strain gauges

attached on its circumference. Hence, when the angle between the two ends change,

the change in strain measured along the length of the wire is equated to the angle. The

output from the strain gauge is very low and instrumentation is needed to obtain a

reasonable output. The goniometer is connected to an angle display unit which is used

to amplify the signals from the goniometer. The primary use of the angle display unit is

to display the angle measured by goniometer using a LCD. However, the angle display

unit is powered by a 9V DC alkaline battery which is used to amplify the goniometer

signal by a gain factor of 90. The measurement range falls between 0.5 V to 4.5 V. The

angle display unit is then connected to an A/D board to further process the signals. The

sensor calibration is performed dynamically by software during the execution of the

control program.

2.3.2 Moving Average Filter Implementation

The sensor feedback is affected by noise in the measurement. The noise can be

reduced to a greater extent through careful design of the filters. While designing filters,

care should be taken that they do not introduce delay in the signal measurement. The

Moving Average Filter is a simple filter to remove random noise and to maintain sharp

step response for signals in the time domain. However the Moving Average Filter is not

ideal for signals in the frequency domain as it cannot effectively separate one frequency

band from the other. The Gaussian, Blackman and Multiple Pass Filters are relatives of

the Moving Average Filter which improves the frequency response to some extent with

increased computation time. In the developed experimental testbed, a double pass 71

25

point Moving Average Filter is used to remove the random noise as well as to improve

the frequency response.

The Moving Average Filter operates by averaging the number of points in the input

signal to produce the output signal. The moving average filter is defined as:

∑−

=

+=1

0][1][

M

jjix

Miy , (2-1)

where x [ ] is the input signal, y [ ] is the output signal and M is the number of

points in the average. The moving average filter is a convolution of the input signal with

a rectangular pulse of area one. The frequency response of the Moving Average Filter is

the Fourier transform of a rectangular pulse as

)sin()sin(][

fMfMfHπ

π= , (2-2)

where f and H[f] denote the frequency and frequency response respectively. In the

subsequently described experiments, the Moving Average Filter is implemented in a

recursive fashion as the algorithm reduced the computation time considerably. Instead

of taking the sum of all 71 points every time, our testing indicated that it is best to setup

a accumulator to store the sum for the first point in the filter and then for the next point

in the filter, the new sample is added and the first sample is subtracted to get the new

sum. By this method the computation time reduces since there are only two

computations per point.

The flow diagram of the moving average algorithm used in the subsequently described

experiment is shown in Figure 2-8. The raw goniometer reading whose sampling

frequency is 1000 Hz is amplified by the 9V DC supply and then passed through a 71

point Moving Average Filter. While performing the filter operation the first 71 samples

26

Figure 2-8. Block diagram for the implementation of Moving Average Filter algorithm.

Goniometer Low Output Reading

Angle Display Unit 9V DC supply

Gain K = 90

Sampling Frequency 1000 Hz

Calibrated angle with one pass 71 point Moving Average Filter

71 point Moving Average Filter

Final Calibrated Angle with two pass 71 point Moving Average Filter

Joint Angle Sensor Feedback

First 71 Samples

10 point Moving Average Filter

71 point Moving Average Filter

27

were too noisy, and hence they were passed to a 10 point Moving Average Filter.

However, even after the application of the filter the calibrated angle had considerable

noise. This is due to the fact that the signal from the angle display unit had a high

sensitivity as calibration required multiplication by a gain of 90. Hence even a very small

change in voltage like a +0.01 V lead to an angle deviation of °9.0 when calibrated. The

uncalibrated goniometer voltage was varying in the range of 210− V and so applying a

higher number of points averaging did not help the cause. Hence the calibrated angle is

passed through another 71 point moving average filter. The filtered angle had a

deviation of °± 1.0 which will be used as a joint angle sensor feedback. The effect of

filtering on the uncalibrated and calibrated goniometer reading (see Figure 2-9-2-13)

has been tested under steady state and transient conditions. The Moving Average

Filters are not the only filters to remove the noise in sensor feedback. However, the

Moving Average Filter performed adequately as a smoothing filter in the time domain.

2.4 Data Acquisition and Computing System

The goniometer signals from the angle display unit are processed with an A/D

converter to be used as a sensor feedback in the closed-loop control system. The A/D

and D/A processing are carried out on a ServoToGo I/O card at a sampling rate of 1000

Hz.The implementation of a PI voltage controller is carried out on a Pentium IV PC

running on QNX. The control algorithm is implemented by QMotor 3.0, a framework for

implementing control program which is developed for QNX operating system and uses

Photon for graphical user interface. The QMotor allows the implementation of advanced

control algorithm as C++ programs which improve execution speed.

The components of QMotor include control program library, Hardware servers and

28

Figure 2-9. Raw goniometer reading from the ADC port with and without Moving Average Filter at steady state.

29

Figure 2-10. Calibrated angle reading from the goniometer at steady state plotted over a period of 0.25 sec. The final measuring angle is a combination of the effect of the first and additional 71 point MA Filter on angle measurement for 0.142 sec and after 0.142 sec respectively.

30

Figure 2-11. Calibrated angle reading from the goniometer at steady state plotted over a period of 15 seconds. The final measuring angle is a combination of the effect of the first and additional 71 point Moving Average Filter on angle measurement for 0.142 sec and after 0.142 sec respectively.

31

Figure 2-12. Raw goniometer reading from the ADC port with and without the Moving Average Filter while tracking.

32

Figure 2-13. Calibrated angle reading from the goniometer while tracking showing the effect of Moving Average Filter.

33

clients and a Graphical User Interface (GUI). The control program library contains

control program class which is used to implement the control algorithms. The functions

included in the control programs class allow control initiations, control calculation and

termination of control. The hardware server is used to connect and access the

ServoToGo I/O board. The clients are used to facilitate communication between the

control programs and the ServoToGo I/O card. The graphical user interface (GUI)

provides a user interface to the control programs. It consists of a main window, log

variable window, control window, and plot window. The main window is used to load the

control program and to set the duration of the control execution. The log variable

window is used to log certain variables such as error signal and control signals. The

control window contains variable that can be used to tune the controllers like control

gains and control frequency. The plot window is used to plot the log variables in real

time and these plots can be exported with Matlab capabilities for further data analysis.

2.5 Event Switch

The event switch sensor used in this project is a product of Motion Lab System.

The sensor is small and compact and can be easily attached to the toe. The sensor has

a two pin lead which can be connected to the event switch cable and then onto the

signal processing board. Figure 2-14 shows a standard event switch sensor.

Figure 2-14. Motion lab system event switch.

34

The event switch sensor is used as an on-off switch by connecting to the digital

input of ServoToGo I/O card using a simple circuit as shown in Figure 2-15.

Figure 2-15. Digital input circuit.

As it can be seen from Figure 2-15, the event switch is connected to a 5v power

supply and a resistor. When load is applied on the event sensor the switch closes and

voltage flows through the circuit and the digital input reads 0, and when load is removed

the switch opens and the digital input reads 1. Since the ServoToGo houses 32 bits

combination of Digital I/O spread over four ports of 8 bits each, it had been difficult to

read a single bit. So in this experiment the event switch was connected to Port D and

the digital input was read as a whole byte after converting from the binary to decimal. In

the case of a toe switch when the toe is off the ground, the switch state is open and the

byte reading from Port D was ‘11111111’ which when converted to decimal read ‘255’.

When the toe strikes the ground the switch state is closed and the byte reading from

Port D was ‘11111101’ which when converted to decimal read ‘253’. Figure 2-16 and 2-

17 shows a typical event switch response during a gait cycle.

35

0 1 2 3 4 5 6 7 8 9 10 11252

252.5

253

253.5

254

254.5

255

255.5

256Toe Switch Reading from Digital Input

Time in Secs

Byt

e R

eadi

ng (B

inar

y to

Dec

imal

) OFF

ON

Figure 2-16. Digital input reading of toe switch.

0 1 2 3 4 5 6 7 8 9 10 11-1

-0.5

0

0.5

1

1.5

2Toe Switch State

Time in Secs

Sw

itch

Sta

te

ON

OFF

Figure 2-17. Toe switch state

36

2.6 Stimulation Circuit

The voltage needed for FES of the calf muscle is calculated by the PI control

algorithm implemented in C++. The calculated voltage and a reference voltage needed

to generate a frequency of 800 Hz are written on to the two D/A channel along with the

power supply voltage. The data from the D/A channel are used as input to design a

custom made stimulation circuit. The functions of the circuit are to produce a stimulation

voltage and a desired frequency using a voltage to frequency scheme. The stimulation

circuit board used for this experiment is shown in Figure 2.18. The voltage controller is

implemented through a Pulse Amplitude Modulation (PAM) with a variable amplitude

positive square wave with a fixed pulse width of 100 µ sec and a fixed frequency of 100

Hz.

Figure 2-18. Stimulation circuit board

37

CHAPTER 3 EXPERIMENTAL METHODS

This chapter describes the various steps needed to achieve the thesis objective.

3.1 Desired Trajectory Setup

The objective of this experiment is to design a voltage controller to track a desired

angular trajectory of the ankle joint from mid stance to toe-off position. Developing the

desired trajectory involves two steps; gait phase detection, and actual recording of the

trajectory. The gait phase is detected by means of an event switch sensor as described

in the previous chapter. In this experiment a toe switch is taped at the bottom of the

subject’s shoe. The switch state is set to ‘ON’ when the toe strikes the ground and ‘OFF’

when the toe leaves the ground. The desired trajectory can be recorded every time the

switch state is ‘ON’; however, it will be difficult to develop a controller to control a

varying trajectory within a short period for every gait cycle. For this experiment, the

longest duration trajectory for any time the switch is ‘ON’ is chosen as the desired

trajectory for developing the closed-loop control. This is achieved by the use of a

counter in the software program which increments from zero when the switch is ‘ON’

and then set to zero again when switch is ‘OFF’. Once the trial experiment is completed

the counter plot in Figure 3-1 tells the number of times the toe switch is ‘ON’ and also

the duration for which the switch is ‘ON’ for each gait cycle.

The actual trajectory is recorded from the goniometer sensor feedback which is

written into a data file during the trial experiment whenever the switch is ‘ON’. Once the

longest counter duration is chosen from the counter plot (Figure 3-1) then it is a simple

process of locating that data in the data file and copying it into the new data file which

will be used as the reference trajectory for developing the closed-loop control algorithm.

38

0 5 10 15 20 25

0

200

400

600

800

1000

1200

Toe Switch Counter

Time in Secs

Sw

itch

Cou

nter

ONOFF

Figure 3-1. Toe switch counter for recording desired trajectory

The trajectory recording experiment was carried out for two circumstances; the

subject walked on the treadmill with the support of the orthosis Lokomat, and the

subject walked on the treadmill without the support of the Lokomat. The two different

trajectories obtained from the experiment are shown in Figure 3-2 and Figure 3-3.

3.2 Desired Trajectory Offset

In this experiment there are two systems that have to be executed simultaneously;

the Lokomat and treadmill system and the voltage controller system. Each system is

controlled by a different person, which leads to an offset in measuring the ankle joint

angle every time. Also, the ankle joint angle feedback from the goniometer is described

between the foot and the shank of the leg. The local coordinate system used for

measuring the ankle and knee joint angle in the sagittal plane using goniometer is

described in Figure 3-4.

39

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8-8

-6

-4

-2

0

2

4

6

8

10

12

14Desired Trajectory

Time in Secs

Ankl

e An

gle

in D

egs

Figure 3-2. Desired trajectory recorded of the subject walking on treadmill with the support of the Lokomat.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7-20

-15

-10

-5

0

5

10Desired Trajectory

Time in Secs

Ankl

e An

gle

in D

egs

Figure 3-3. Desired trajectory recorded of the subject walking on treadmill without the support of the Lokomat.

40

Figure 3-4. Coordinate system for measuring the ankle and knee angle

Figure 3-5. Actual measurement of the ankle and knee angle

Figure 3-5 shows how the ankle and knee angles are measured using the above

coordinate system with the X-axis corresponding to the foot and the Z-axis to the shank.

Plantar flexion is when the angle between the X and Z-axis increases and dorsiflexion is

vice versa. Although the foot was always on the ground during the experiment, the

orientation of the shank varied. The subject tends to orient the shank differently at the

41

start of every experiment which leads to an offset in the ankle angle reading from the

goniometer. The ranges of ankle angle remained the same but the maximum and

minimum angle varied. The desired trajectory was measured only once so this offset

reflected in the computed error during every gait cycle and lead to saturation of the

controller after some time. This problem was corrected by software wherein the code

captures the first sample of the ankle angle from the goniometer when the switch is ‘ON’

and compares it with the first sample of the desired trajectory. The difference is then

added to every point in the desired trajectory to shift the curve to match the actual

trajectory. This strategy proved extremely useful in conducting experiments without

having to worry about matching the shank orientation every time.

3.3 Experimental Procedure

The experiments were conducted on a healthy person at the Human Motor

Performance Laboratory (HMPL) housed within the Brain Rehabilitation Research

Center (BRRC) located at the Veterans Affairs (VA) hospital in Gainesville, Fl. In this

project, two sets of experiments were performed. In the first experiment the voltage

controller was tested with the subject walking on the treadmill without the support of the

orthosis Lokomat and with the support of the Lokomat in the second set of experiments.

In both experiments, a set of procedures were followed to obtain the final result. At first

the goniometer sensors and the event switches were attached to the subject and the

leads were connected to the signal conditioning board ServoToGo. Then for testing with

Lokomat (see Figure 3-6), additional braces were attached to the subject hip and knee

and then suspended over the treadmill platform by means of a harness. Both sets of

experiments require the subject to walk on the treadmill without stimulation initially to

record the desired trajectory. Once the desired trajectory is recorded the subject is then

42

allowed to stand without walking on the treadmill platform and stimulation voltage is

applied gradually in increments to determine the maximum bound for the voltage. Then

the surface electrodes are attached across the subject’s calf muscle and the control

program is executed which delivers the required voltage based on a proportional

integral control method every time the switch is ‘ON’ to track the ankle angle along a

desired trajectory.

Figure 3-6. Experimental setup with the Lokomat and computing system

43

CHAPTER 4 RESULTS AND DISCUSSIONS

In this project two sets of experiment were conducted where the subject’s ankle

undergoes stimulation while walking on the treadmill without the support of the Lokomat

for the first set of experiments and with the support of Lokomat for the second set of

experiments. The plots for the first set of experiments (see Figure 4-1-4-4) are obtained

with

k p = 2.0, ki = 0.5,

and the maximum voltage is set at 25 volts. The plots for the second set of experiments

(see Figure 4-5-4-8) are obtained with

k p = 5.0, ki = 4.0,

and the maximum voltage is set at 27 volts. The initial voltage and the frequency of the

controller for both the experiments are set as 10 volts and 25 Hz.

The following section presents the results of the two experiments with the

following plots 1) actual and desired trajectory 2) error 3) switch state 4) stimulation

voltage.

From the error plots in Figure 4-2 and 4-6, one can observe that although the error

does not vary much between the two experiments, there is an increase in the error at

some points in Figure 4-2. Efforts in [42] report that the advantage of using a hybrid

orthosis device greatly reduces the number of degrees of freedom that contribute to the

walking motion. A person spends energy in moving the hip and knee joint while the

ankle undergoes muscle stimulation when walking on the treadmill without the Lokomat.

During some point of the experiment the muscle fatigue leads to a slow step during that

44

particular gait cycle. This increases the error to a very high value and leads to saturation

of the controller (Figure 4-4). Variations in step size and duration over each gait cycle

were reported when the subject walked on the treadmill without the Lokomat. This lead

to difficulty in tuning the gains as the variations occurred over a small window of time.

Although the plots (Figure 4-5-4-8) from the subject walking with the Lokomat look

convincing, the person was capable of walking normally. During the experiment the

subject was asked not to perform plantar flexion at some point to test whether the

controller is able to produce the required plantar flexor moment. Although the subject

reported that the controller produced a plantar flexor moment it is difficult to point out in

the plots. The main objective of this project it to provide the required plantar flexor

moment during ankle push-off to prevent individual’s from tripping on their toes. Hence,

another experiment was performed with the person standing on the treadmill without

walking and the control program was executed. The human subject was asked to bend

the shank forward causing dorsiflexion while the switch was ‘ON’ mimicking the motion

of a person collapsing. However the controller produced enough stimulation voltage to

generate the required plantar flexion moment to lift the leg causing plantar flexion.

45

0 2 4 6 8 10 12 14 16 18 20-20

-15

-10

-5

0

5

10

15

20

25

30

35Trajectory plot for walking without Lokomat

Time in Secs

Ankle

Angle

in D

egs

Actual TrajectoryDesired Trajectory

Figure 4-1. Actual and desired trajectory plot of a person walking on the treadmill without the support of the Lokomat.

46

0 5 10 15-15

-10

-5

0

5

10

15

20

25Error plot for walking without Lokomat

Time in Secs

Ankle

Angle

in D

egs

Figure 4-2. Error plot of a person walking on the treadmill without the support of the Lokomat.

47

0 5 10 15-0.2

0

0.2

0.4

0.6

0.8

1

Switch State for walking without Lokomat

Time in Secs

Figure 4-3. Switch state plot of a person walking on the treadmill without the support of the Lokomat.

48

0 5 10 15-40

-30

-20

-10

0

10

20

30

40

50

Stimulation Voltage plot for walking without Lokomat

Time in Secs

Voltage in v

olts

PI VoltageOutput Voltage

Figure 4-4. Stimulation voltage plot of a person walking on the treadmill without the support of the Lokomat.

49

0 2 4 6 8 10 12 14 16 18 20-20

-15

-10

-5

0

5

10

15

20

25

30Trajectory plot for walking with Lokomat

Time in Secs

Ankle

Angle

in D

egs

Actual TrajectoryDesired Trajectory

Figure 4-5. Actual and desired trajectory plot of a person walking on the treadmill with the support of the Lokomat.

50

0 2 4 6 8 10 12 14 16 18 20-20

-15

-10

-5

0

5

10Error plot for walking with Lokomat

Time in Secs

Ankle

Angle

in D

egs

Figure 4-6. Error plot of a person walking on the treadmill with the support of the Lokomat.

51

0 2 4 6 8 10 12 14 16 18 20-0.2

0

0.2

0.4

0.6

0.8

1

Switch State for walking with Lokomat

Time in Secs

Figure 4-7. Switch state of a person walking on the treadmill with the support of the Lokomat.

52

0 2 4 6 8 10 12 14 16 18 20-120

-100

-80

-60

-40

-20

0

20

40

Stimulation Voltage plot for walking with Lokomat

Time in Secs

Voltage in v

olts

PI VoltageOutput Voltage

Figure 4-8. Stimulation voltage plot of a person walking on the treadmill with the support of the Lokomat.

53

CHAPTER 5 CONCLUSION

This project resulted in the design and validation of a system that can be used to

trigger the stimulation of the human ankle joint and to track a desired trajectory

efficiently in a safe manner. The toe switch was used to trigger the control algorithm and

the goniometer position feedback was used for tracking. The results of the experiment

reported in Chapter 4 were conducted on a healthy person who was capable of walking

normally. One of the limitations of this project is due to the use of a discontinuous

desired trajectory. Polynomial curve fitting or extrapolation of the discontinuous desired

trajectory can be used to obtain the desired velocity and its higher derivatives.

The next step is to conduct the experiment on a person with an affected gait and

test the performance of the controller. The results from this experiment could lead to a

better understanding of the human-Lokomat interaction, and in the design of an

advanced control system. The person will be interacting with the Lokomat which could

contribute up to some extent in providing the plantar flexor moment along with the FES.

In general, the human-Lokomat interaction is considered as a disturbance. However,

the estimation of the human-Lokomat interaction as described in [35] can lead to

information about the contribution of FES in the progress of rehabilitation. The force

produced by the muscle contraction shares a nonlinear relationship with the stimulation

voltage. In this project this relationship is assumed as linear. In the future additional

controllers will be designed and implemented.

54

REFERENCES

[1] Post-Stroke Rehabilitation Fact Sheet, National Institute of Neurological Disorders And Stroke National Institutes of Health, October 17, 2006.

[2] National Spinal Cord Injury Statistical Center, Spinal Cord Injury Facts and Figures at a Glance, Birmingham, Ala.: National SCI Statistical Center, University of Alabama at Birmingham, 2000.

[3] B. K. Go, M. J. DeVivo, and J. S. Richards, “The Epidemiology of Spinal Cord Injury,” in Spinal Cord Injury: Clinical Outcomes from the Model Systems, eds. S. L. Stover, J. A. DeLisa and G. G. Whiteneck, Aspen, Gaithersburg, 1995.

[4] J. Hidler and A. Wall, “Alterations in Muscle Activation Patterns during Robotic-Assisted Walking,” Clinical Biomechanics, vol. 20, pp. 184-193, 2005.

[5] S. Jezernik, G. Columbo, T. Keller, H. Frueh, and M Morari, “Robotic Orthosis Lokomat: A Rehabilitation and Research Tool,” Neuromodulation, vol. 6, no. 2, pp. 108-115, Apr. 2003.

[6] G. Colombo, M. Joerg, R. Schreier, and V. Dietz, “Treadmill Training of Paraplegic Patients Using a Robotic Orthosis,” Journal of Rehabilitation Research and Development, vol. 37, no. 6, pp. 693-700, 2000.

[7] T. G. Hornby, D. H. Zemon, and D. Campbell, “Robotic-Assisted, Body-Weight-Supported Treadmill Training in Individuals Following Motor Incomplete Spinal Cord Injury,” Phys. Ther., vol. 85, pp. 52-66, 2005.

[8] G. Colombo, “Treadmill Training with the Robotic Orthosis ‘Lokomat’: New Technical Features and Results from Multi-Center Trial in Chronic Spinal Cord Injury,” International Journal of Rehabilitation Research, vol. 27, pp. 92-93, 2004.

[9] HealthSouth Corporation, “AutoAmbulator,” Accessed March 2009. Available at: http://www.healthsouth.com/what_we_do/inpatient_rehabilitation/rehabilitation_technology/autoAmbulator.asp

[10] S. Hesse, et al. “An Electromechanical Gait Trainer for Restoration of Gait in Hemiparetic Stroke Patients: Preliminary Results,” Neurorehabilitation and Neural Repair, vol. 15, pp. 39-50, 2001.

[11] S. Hesse and D. Uhlenbrock, “A Mechanized Gait Trainer for Restoration of Gait,” Journal of Rehabilitation Res. And Dev., vol. 37, no 6., pp. 701-708, 2000.

[12] J. L. Emken, J. H. Wynne, S. J. Harkema, and D. J. Reinkensmeyer, “A Robotic Device for Manipulating Human Stepping,” IEEE Transactions on Robotics, vol. 22, no. 1, pp. 185 – 189, Feb. 2006.

55

[13] D. Reinkensmeyer, J. Wynne, and S. Harkema, “A Robotic Tool for Studying Locomotor Adaptation and Rehabilitation,” in Proc. IEEE EMBS/BMES, 2002, pp. 2353-2354.

[14] J. L. Emken and D. J. Reinkensmeyer, “Robot-Enhanced Motor Learning: Accelerating Internal Model Formation during Locomotion by Transient Dynamic Amplification,” IEEE Trans. on Neural Systems and Rehabilitation Engineering, vol. 13, no. 1, pp. 33-39, Mar. 2005.

[15] J. A. Nessler, et al., “A Robotic Device for Studying Rodent Locomotion After Spinal Cord Injury,” IEEE Trans. on Neural Systems and Rehabilitation Engineering, vol. 13, no. 4, pp. 497 – 506, Dec. 2005.

[16] G. Sawicki, K. Gordon, and D. Ferris, “Powered Lower Limb Orthoses: Applications in Motor Adaptation and Rehabilitation,” in Proc. IEEE Int. Conf. on Rehabilitation Robotics, Chicago, IL, 2005, pp. 206-211.

[17] M. Vukobratovic, D. Hristic, and Z. Stojiljkovic, “Development of Active Anthropomorphic Exoskeletons,” Medical and Biological Eng., vol. 12, pp. 66-80, 1974.

[18] M. Vukobratovic, B. Borovac, D. Surla, and D. Stokic, “Biped Locomotion: Dynamics, Stability, Control and Application,”, vol. 7, Springer Verlag, 1990.

[19] A. Seireg and J. G. Grundman, “Design of a Multitask Exoskeletal Walking Device for Paraplegics,” in Biomechanics of Medical Devices, D. Ghista, Ed. Marcel Dekker, 1981, pp. 569-639.

[20] B. J. Ruthenberg, N. Wasylewski, and J. Beard, “An Experimental Device for Investigating the Force and Power Requirements of a Powered Gait Orthosis,” Journal of Rehabilitation Research and Development, vol. 34, pp. 203-213, 1997.

[21] J. A. Blaya and H. Herr, “Adaptive Control of a Variable –Impedance Ankle-Foot Orthosis to Assist Drop-Foot Gait,” IEEE Trans. of Neural Systems, and Rehabilitation Eng., vol. 12, pp. 24-31, 2004.

[22] D. P. Ferris, J. Czerniecki, and B. Hannaford, “An Ankle-Foot Orthosis Powered by Artificial Muscles,” Annual Meeting of the American Society of Biomechanics, San Diego, CA, 2001.

[23] G. Sawicki and D. P. Ferris, “ A knee-ankle-foot Orthosis (KAFO) Powered by Artificial Pneumatic Muscles,” International Society of Biomechanics, Dunedin, New Zealand, 2003.

[24] M. A. Townsend and R. J. Lepofsky, “Powered Walking Machine Prosthesis for Paraplegics,” Med. Biol. Eng., vol. 14, no. 4, pp. 436-444, 1976.

56

[25] D. Ferris, G. Sawicki, and A. Domingo, “Powered Lower Limb Orthoses for Gait Rehabilitation,” Top. Spinal Cord Inj. Rehabilitation, vol. 11, no. 2, pp. 34-49, 2005.

[26] G. Belforte, L. Gastalid, and M. Sorli, “Pneumatic Active Gait Orthosis,” Mechatronics., vol. 11, no. 3, pp. 301-323, 2001.

[27] H. Herr and A. Wilkenfeld, “User-adaptive Control of a Magnetorheological Prosthetic Knee,” Indust. Robot., vol. 30, no. 1, pp. 42-55, 2003.

[28] Y. Sankai, “Robot Suite HAL (Hybrid Assistive Limb),” Accessed April, 2009. Available at: http://sanlab.kz.tsukuba.ac.jp/english/r_hal.php

[29] H. Kawamote and Y. Sankai, “Power Assist Method Based on Phase Sequence Drive by Interaction Between Human and Robot Suit,” IEEE International Work-shop on Robot and Human Interactive Communication, 2004.

[30] K. Kasaoka and Y. Sankai, “Predictive Control Estimating Operator’s Intention for Stepping-up Motion by Exo-skeleton Type Power Assist System HAL,” in Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, 2001.

[31] J. Boyd. (2005,Apr.), “Bionic Suit Offers Wearer Super Strength,” New Scientist, (2494), pp. 19.

[32] J. E. Pratt, B. T. Krupp, C. J. Morse, and S. H. Collins, “The RoboKnee: an Exoskeleton for Enhanced Strength and Endurance During Walking,” in Proc. IEEE Int. Conf. on Robotics and Automation, New Orleans, LA, 2004.

[33] H. Kazerooni, “Welcome to the BLEEX Project,” Accessed April 20, 2005. Available at http://bleex.me.berkeley.edu/bleex.htm.

[34] H. Kazerooni, “The Human Power Amplifier Technology at the University of California, Berkeley,” Robot Autonom. Systems., vol. 19, no. 2, pp. 179-187, 1996.

[35] S. Jezernik, G. Colombo, and M. Morari, “Automatic Gait-Pattern Adaptation Algorithms for Rehabilitation with a 4-DOF Robotic Orthosis,” IEEE Trans. Robot Automat., vol. 20, no. 3, pp. 574-582, 2004.

[36] A.Kralj, T.Bajd, R. Turk, J. Krajnik, and H. Benko, “Gait Restoration in Paraplegic Patients: A Feasibility Demonstration Using Multichannel Surface Electrode FES,” J Rehabil R D/Veterans Adm Dept of Med Sur Rehabil R&D Serv, vol. 20, pp. 3–20, 1983.

[37] A. Kralj, T. Bajd, and R. Turk, “Enhancement of Gait Restoration in Spinal Injured Patients by Functional Electrical Stimulation,” Clinical Orthop., vol. 233, pp. 34–43, 1988.

[38] Sigmedics. Inc, “The Parastep I System,” Accessed May 2009. Available at: http://www.sigmedics.com/TheParastep/

57

[39] D. Popovic, M. Radulovic, L. Schwirtlich, and N. Jaukovic, “Automatic vs Hand-Controlled Walking of Paraplegics,” Med. Eng. Phys., vol. 25, pp. 63–73, 2003.

[40] F. Sepulveda, M. H. Granat, and A. Cliquet Jr., “Gait Restoration in a Spinal Cord Injured Subject via Neuromuscular Electrical Stimulation Controlled by an Artificial Neural Network,” Int. Journal of Artif. Organs, vol. 21, pp. 49–62, 1998.

[41] R. Spadone, G. Merati, E. Bertocchi, E. Mevio, A. Veicsteinas, and A. Pedotti et al., “Energy Consumption of Locomotion with Orthosis Versus Parastep Assisted Gait: A Single Case Study,” Spinal Cord, vol. 41, pp. 97–104, 2003.

[42] T. A. Thrasher and M. R. Popovic, “Functional Electrical Stimulation of Walking:Function, Exercise and Rehabilitation,” Annales de re´adaptation et de me´decine physique, vol. 51, pp. 452–460, 2008.

[43] C. C. Monaghan, W. J. B. M. Van Riel, and P. H. Veltink, “Control of Triceps Surae Stimulation Based on Shank Orientation Using a Uniaxial Gyroscope during Gait,” Med. Biol. Eng. Comput., vol. 47, pp. 1181–1188, 2009.

[44] A. H. Vette, K. Masani, J. Y. Kim, and M. R. Popovic, “Closed-Loop Control of FES-Assisted Arm-Free Standing in Individuals with Spinal Cord Injury: A Feasibility Study,” Neuromodulation, vol. 12, pp. 22 – 32, 2009.

[45] K. J. Hunt, M. Munih, N. Donaldson, and F. M. D. Barr, “Optimal Control of Ankle Joint Moment: Toward Unsupported Standing in Paraplegia,” IEEE Trans. on Automatic Control, vol. 43, pp. 819-832, 1988.

[46] E. C. M. Villalpando, “Estimation of Ground Reaction Force and Zero Moment Point on a Powered Ankle-Foot Prosthesis,” M.S. thesis, School of Architecture and Planning, Massachusets Insti. of Tech., Cambridge, M.A, 2006.

[47] S. K. Au, P. Dilworth, and H. Herr, “An Ankle-Foot Emulation System for the Study of Human Walking Biomechanics,” in Proc. IEEE Int. Conf. on Robotics and Automation, 2006, pp. 2939-2945.

[48] P. E. Crago, P. H. Peckham, and G. B. Thrope, “Modulation of Muscle Force by Recruitment During Intramuscular Stimulation,” IEEE Transactions on Biomedical Engineering, vol. 27, pp. 679-684, 1980.

[49] A. Thrasher, G. M. Graham, and M. R. Popovic, “Reducing Muscle Fatigue Due to Functional Electrical Stimulation Using Random Modulation of Stimulation Parameters,” Artificial Organs, vol. 29, pp. 453–458, 2005.

[50] D. J. Tyler and D. M. Durand, “Combined Modulation of Pulse Width and Pulse Amplitude to Enhance Functional Selectivity of Neural Stimulation,” in Proc. IEEE-EMBC and CMBEC, 1995, pp. 1109-1110.

58

[51] C. J. De Luca, “The Use of Surface Electromyography in Biomechanics,” Journal of Applied Biomechanics, vol. 13, pp. 135-163, 1997.

59

BIOGRAPHICAL SKETCH

Ganeshram Jayaraman was born in Chennai, India in 1986. He received his

bachelor’s in mechanical engineering from Sathyabama University, Chennai, India in

May 2007. He received his Master of Science degree in mechanical engineering in the

Department of Mechanical and Aerospace Engineering at the University of Florida

under the supervision of Dr. Warren E. Dixon in 2010. The primary focus of his research

was to develop an FES-based ankle orthosis for rehabilitation of individuals suffering

from stroke and Spinal Cord Injuries (SCI).