intelligent control system for an active afo with the...

Chapter 4

Intelligent Control System for

an Active AFO with the

Application of Fuzzy Logic

4.1 INTRODUCTION

Fuzzy logic controller (FLC) is a set of linguistic control rules related by the dual

concepts of fuzzy implication and the compositional rule of inference. In essence, then the

FLC provides an algorithm which can convert the linguistic control strategy based on expert

knowledge in to an automatic control strategy. Experince shows that FLC yields results

superior to those obtained by conventional control algorithms [49]. In particular, the

methodology of the FLC appears very useful when the processes are too complex for analysis

by conventional quantitative techniques or available sources of information are interpreted

qualitatively, inexactly, or uncertainly. Thus fuzzy logic control may be viewed as a step

toward a rapprochment between conventional precise mathmatical control and human like

decision making.

Since humans use bipedal walking, mobility and stability are exclusive terms. In this

sense, humans need a trade-off of the two terms to achieve a task given or a goal desired. In

order to do this, humans require sensory systems to gather information about surrounding

environments, and control strategies utilizing information. The visual, vestibular, and

proprioceptive systems are used when humans gather this information. From the perspective

of locus of control, the pertinent interactions occur among the central nervous system, the

Intelligent control system for an active AFO with the application of Fuzzy logic

Page 40

peripheral nervous system, and the musculoskeletal system. In physiological system, control

strategy employs feedback control and feed-forward control [50].

Feedback control is referred to as closed-loop, because the outputs return and

influence the inputs. The role of feedback information is to provide the substrate with the

detection and correction of movement errors. Thus, if feedback information is readily

available then desired movements can be achieved more efficiently. However, in humans,

there are two major problems in the feedback system: transmission delays leading to low

feedback gain and processing overload in the neural system. Feed-forward control strategy

can be used to compensate feedback control strategy by avoiding problems with transmission

delays and also in reducing the amount of processing required. One of key features of feed-

forward control strategy is a movement model of the future which allows planning for future

events. The model is based on knowledge of the results of the dynamics involved in a

movement, and also considers system constraints and future goals. This knowledge is often

referred to as an internal model, or internal representation. In FLC predefined rule base with

an expert system is used, also called as internal model to minimize the computational delays.

This chapter discuss about control design for active ankle foot orthosis with the

application of fuzzy logic. The control logic is developed based on the property of symmetry

in the foot movements in a gait cycle. The gait patterns are monitored through National

Instrument’s data acquisition system which is interfaced with a gyro sensor. The sensor data

is then fed as an input to the FLC which generates the control signal for the actuator in real

time.

4.2 GAIT ANALYSIS USING GYROSCOPE

Current gold standard devices for spatial gait analysis include electronic pressure

mats, inertial systems and motion analysis systems, which are extremely costly and may only

be used in specialized environments with highly trained personnel. Recently, gyroscope-

based gait analysis has become increasingly popular in research as well as clinical practice

[51]. This is due to many factors: gyroscopes are inexpensive, portable, battery powered and

are compatible with wireless data transmission. Additionally, the output of a gyroscope has

also been shown to depend purely on the orientation of the sensor’s axis of rotation rather

than its physical location. Accelerometers share many of these features; however their output


Page 41

is also affected by gravity, and is known to vary with sensor location as well as axis of

rotation [51-53].

4.2.1 Gyroscope

A gyroscope is a device that can measure angular velocity. The LPY530AL is a low-

power two-axis micro-machined gyroscope able to measure angular rate along pitch and yaw

axes. The LPY530AL has a full scale of ±300 °/s and is capable of detecting rates with a -3

dB bandwidth up to 140 Hz. The gyroscope is the combination of one actuator and one

accelerometer integrated in a single micro machined structure. It includes a sensing element

composed by single driving mass, kept in continuous oscillating movement and able to react

when an angular rate is applied based on the Coriolis principle. The Coriolis affect is a

deflection of moving objects when they are viewed in a rotating reference frame. In a

reference frame with clockwise rotation, the deflection is to the left of the motion of the

object; in one with counter-clockwise rotation, the deflection is to the right. The Coriolis

force is proportional to both the angular velocity of the rotating object and the velocity of the

object moving towards or away from the axis of rotation. A CMOS IC provides the measured

angular rate to the external world through an analog output voltage, allowing high level of

integration and production trimming to better match sensing element characteristics. Figure

4.1shows the LPY530AL dual axis gyroscope module pin diagram and the axis of rotation.

Figure 4.1 LPY530AL Module and axis diagram

4.2.2 The Data acquisition module

The NI USB-6221 shown in Figure 4.2 is a USB high-performance M Series

multifunction DAQ module optimized for superior accuracy at fast sampling rates. The

module has 16 analog input channels (16 bit, 250kS/s), 2 analog output channels (16 bit, 833

http://en.wikipedia.org/wiki/Rotating_reference_frame


2

kS/s) and 24 digital I/O channels (8 clocked). NI-DAQmx driver and measurement services

software provides easy-to-use configuration and programming interfaces with features such

as the DAQ assistant to help reduce development time. In the experimental set up, the DAQ

acquisition mode is configured at continuous samples to continuously monitor the gait pattern

of the foot while walking. The number of samples to read is set at 80 and the sampling rate

being 40 Hz. This ensures Nyquist’s rule is satisfied for the input signal whose frequency

varies within 10 Hz. The buffer value is set at 80 so that during continuous data acquisition

the internal buffer does not overflow.

Figure.4.2 NI-USB-6221 DAQ card

4.2.3 Extracting useful information from Gyroscope

The angular velocity of the foot during the gait cycle can be extracted from the Vout

of gyroscope using the relation [53]:

ysensitivitrefVoutVOmega (4.1)

where,

Omega – Angular velocity of the foot.

Vout – Output voltage of the gyroscope sensor.

Vref – Output voltage of Gyroscope at standstill position.

Sensitivity – Smallest value of angular velocity which can be detected using gyroscope; this

is 0.83 mv/degree/sec.


Page 43

The angular displacement of the foot from flat foot position is determined by finding

the area under the curve of the Angular velocity versus time plot. This is obtained by

continuously integrating the angular velocity values with respect to time from zero till current

time t. The values of both angular velocity and angular displacement are plotted with respect

to time.

4.2.4 Gait pattern analysis using gyroscope

The output voltage of gyroscope at standstill position is taken as the Vref. The Vref is

generally in the range 1.26 to 1.262 V. During dorsiflexion, the gyroscope gives voltages less

than the steady state value as it is rotated in clock-wise direction. While during planter-

flexion, it senses voltages greater than steady state voltage due to anticlockwise movement.

Angular velocity and angle are measured using the sensor data. Figure 4.3 shows

angular plot in standstil position. The plots of angular velocity and angular displacement give

slight variations about zero indicating noise. The angular displacement increase marginally

with time as the noise in the plot of angular velocity gets integrated over time. However it can

be observed that the angular displacement varies within a small value of 0.4 degrees, hence

accurately documenting the stand still position.

Figure 4.3 Standstill position

As the gyroscope attached to the foot rotates in the anti-clockwise direction about the

z-axis the magnitude of voltage increases from the steady state value (Vref) of 1.262V

reaching a maximum at the heel off position, and as the foot rotates in clockwise direction

returning to the flat foot condition, the voltage decreases back to Vref. The plot of angle

versus time follows the same pattern as voltage showing a complete plantar-flexion cycle.

Figure 4.4 shows angular plot for plantar-flexion movement from stand still and back to the

same position. Similarly when the gyroscope attached to the foot rotates in the clockwise


Page 44

direction about the z-axis the magnitude of voltage decreases from the steady state value

(Vref) of 1.262V reaching a minimum at the toe off position, and as the foot rotates in anti-

clockwise direction returning to the flat foot condition, the voltage increases back to Vref. The

plot of angle versus time follows the same pattern as voltage showing a complete dorsiflexion

cycle. Figure 4.5 shows angular plot for dorsiflexion movement from stand still and back to

the same position.

Figure 4.4 Plantar-flexion cycle

Figure 4.5 Dorsiflexion cycle

Figure 4.6(a) and 4.6(b) show the complete gait cycle when gyroscope is attached to

the left foot and then to the right foot respectively. In both the cases, gait cycle started off

from a stand still position putting the left foot forward. The gait cycle shows that the left foot

first does a dorsiflexion motion followed by a plantarflexion with the cycle continuing till

standstill. When the left foot is in dorsiflexion phase, the right foot is in the plantar flexion

phase in a normal gait cycle. Hence these graphs are are inverse of each other.


Page 45

(a) (b)

Figure 4.6 Gait cycle with sensor connected to (a) left foot (b) right foot.

The average plot of ankle angle for number of normal subjects is shown in Figure 4.7.

The negative part corresponds to the dorsiflexion phase of the gait cycle and the positive part

corresponds to the plantar-flexion phase.

(a) (b)

Figure 4.7 Average ankle angle plot in a gait cycle (a) left foot (b) right foot

The study has been conducted on two hemiplegia patients (volunteer). They have

abnormal gait patterns due to stroke. Patient 1 is about 43 years old, had a stroke six years

ago and able to walk without having any support. Patient 2 is about 71 years old and had

chronic attack 16 years before; cannot walk without any support and having balancing

problem. The gait data of both the patients were taken and averaged to get normalized data.

Both patients have their right foot normal. The gyro-sensor was placed on their left foot first

and gait data were taken. Similarly gait data were taken by placing the sensor on their right


Page 46

foot. Figure 4.8 (a) and 4.8 (b) show the average angular plot of both patients and comparison

with normal gait when sensor placed on the left foot then with right foot respectively. The

sensor placement for data acquisition for the patients is shown in Figure 4.9. Figures 4.7and

4.8 show that, there is symmetry in gait pattern in normal subjects. The non-symmetric gait

pattern in stroke patients is due to balance deficiencies and difficulty in moving the body over

unstable limb. There is no positive ankle angle and ankle is stiff; not able to rotate in clock

wise direction with respect to shank. The angular velocity is almost 60 to 70 % less compared

to normal due to dragging of the foot.

(a) (b)

Figure.4.8 Average angle plot of normal subject and drop foot gait in complete gait cycle

with left foot started with swing first when sensor connected to (a) left foot (b) right foot.

Figure.4.9 Data acquisition of patient 1 and patient 2

From the normal gait cycle pattern it is inferred that the angular plot is symmetric in

nature i.e left and right foot follows the property of symmetry. This symmetric property is

used to control the actuation of an ankle foot orthosis in an intelligent manner. Using these

gait data an expert system is devised with the help of Fuzzy logic which is discussed in the

next section.


Page 47

4.3 FUZZY LOGIC CONTROL FOR ACTIVE AFO

The Fuzzy logic, unlike conventional logic systems, is able to model inaccurate or

imprecise models. It offers a simpler, quicker and more reliable solution that is clearly

advantageous over conventional control techniques [54, 55]. Conventional control techniques

such as Proportional Integral and Differential (PID) control, nonlinear feedback control,

adaptive control, sliding mode control, Linear quadratic Gaussian control etc. are

advantageous when the values of the controller parameters are known and the control signals

are generated exactly. Also, when the underlying assumptions are satisfied, many of these

methods provide good stability, robustness to model uncertainties and disturbances, and

speed of response. However these control algorithms are “hard” or “inflexible” and cannot

handle “soft” intelligent control which may involve reasoning and inference making using

incomplete, vague, noncrisp, and qualitative information, and learning and self- organization

through past experience and knowledge. Fuzzy control is a type of intelligent control whose

main feature is that, a control knowledge base is available within the controller and control

actions are generated by applying existing conditions or data to the knowledge base, making

use of an inference mechanism. The knowledge base and inference mechanism can handle

noncrisp and incomplete information, and the knowledge itself will improve and evolve

through “learning” and past experience [55].

The fuzzy control gives the optimal performance to control the dc motor and also

overcome the disadvantages of the conventional control sensitiveness to inertia variation and

sensitiveness to variation of speed with drive system of dc motor [56]. The dc motor position

control was implemented to show that Fuzzy Logic Controller (FLC) responds with less

overshoot and minimum settling time over conventional PID control [57]. Fuzzy logic

systems emulate human decision making more closely than artificial neural network. The

main advantages are that no mathematical modelling is required as in PID since the controller

rules are based on the knowledge of system behaviour and experience of control engineer

[54-62]. This research project deals with the novel design for controlling the actuation

mechanism of an ankle foot orthosis (AFO) using FLC.

A control unit is developed based on fuzzy logic for an active AFO with real time

detection to provide active support for the defective foot using principles of artificial

intelligence. The control system provides a real time control of the defective foot by


Page 48

continuously monitoring the gait patterns. The input signals for the control system are

generated by the sensor system having gyroscope. The FLC developed in Laboratory Virtual

Instruments Engineering Workbench (LabVIEW) is used to provide plantar-flexion and

dorsiflexion movements of the ankle during walking.

4.3.1 Theory and methods

The gyroscope sensor LPY530AL is attached to the subject’s working leg. The sensor

is attached to the middle part of the foot so that both the dorsiflexion and planterflexion

phases are determined completely for accurate data acquisition. The sensor unit is connected

to National Instruments data acquisition module NI 6221. The gait patterns of both the legs

during one complete walking cycle are observed for a normal person and accordingly the

controller is designed to replicate the same pattern for the defective leg.

During the data acquisition phase the gyroscope is first attached to the left foot to

observe the gait pattern during one complete cycle followed by the gyroscope being attached

to the right foot. In both the cases a similar walking pattern is replicated so that the complete

gait pattern followed by both the left and right foot could be detected almost synchronously.

In the experimental set up the left foot is taken as the start of the walking cycle first. From the

values collected it is observed that the left foot first makes a dorsiflexion motion followed by

a plantar flexion whereas at the same time the right foot makes a plantarflexion movement

first followed by a dorsiflexion. A similar pattern is observed for angular velocity values.

The angular displacement of the foot from flat foot position can be determined by finding the

area under the curve of the Angular velocity versus time plot. This is obtained by

continuously integrating the angular velocity values with respect to time from zero to current

time ‘t’ [53]. The values of both angular velocity and angular displacement are plotted for an

average gait cycle. The ankle angle plot for left and right foot are shown in Figure 4.10 and

Figure 4.11 respectively. The average angular velocity plot for number of subjects is shown

in Figure 4.12. The negative part corresponds to the dorsiflexion phase of the gait cycle and

the positive part corresponds to the plantar-flexion phase.


Page 49

Figure 4.10 Angle plot for left foot in a gait cycle

Figure 4.11 Angle plot for right foot in a gait cycle


Page 50

Figure 4.12 Average ankle angular velocity plot in a gait cycle

A general graph is obtained as shown in Figure 4.13, based on the angle and angular

velocity plot for the left and right foot in an average gait cycle. The angular velocity is time

differential of displacement, therefore when the angle displacement reaches steady state, the

velocity goes to zero. With graphical illustration of gait pattern obtained it is clear that the

angle and angular velocity of normal leg are enough to predict the actuation of the defective

leg at any moment. This is further used to make the fuzzy rule base.

4.3.2 Fuzzy rule base

Fuzzy rule base is based on the expert knowledge of the system to be controlled. The

four main methods for finding control rules are [63]:

Expert experience and control engineering knowledge.

Based on an experimental operators control actions.

Based on a fuzzy model of the system.

Based on adaptation or learning.


Page 51

Figure 4.13 a) Angular displacement of right foot for a complete gait cycle

b) Angular velocity of right foot for a complete gait cycle

c) Angular velocity of left foot (actuation to be provided) for a complete gait cycle

TABLE 4.1 THE BROAD RULE BASE.

IF

Angular

velocity

Positive

AND

Angle

Positive

THEN

Actuation

Negative

Zero Positive Zero

Negative Positive Positive

Negative Negative Positive

Zero Negative Zero

Positive Negative Negative


Page 52

Experimental operators are used in this implementation of the system, used for

controlling the AFO. With reference to the Figure 4.13 six broad rules define the output

actuation pattern. The first part of the rule which uses input variables and membership

functions is called the antecedent part. The resulting conclusion which describes the output

variable and membership functions is called the consequent part. Considering inputs are

angle and angular velocity and output is the actuation in terms of angular velocity. Table 4.1

gives the broad rule base. The output actuation has been approximated based on these six

rules. The amount of angular velocity is not constant during the gait cycle. To get a more

accurate actuation curve, the angular velocity has been divided into further membership

values, each having different magnitude (experimental results have given an optimal output

for seventeen divisions). The fuzzy rule base is as given in Table 4.2. The division is made

like N8, meaning highest value of angular velocity and N7, a scale of magnitude less than

N8, and N7,N6,N5,N4,N3,N2,N1, continues. Hence N1 having smallest value of negative

angular velocity, also close to zero. In the positive angular velocity too, the same division is

helpful. So P8, P7, P6, P5, P4, P3, P2, P1 divide the positive velocity into eight magnitude-

wise different membership function. Hence, P8 has largest magnitude of positive angular

velocity and P1 has smallest value of positive angular velocity. Zero angular velocity is

represented by Z membership function. And to get a smoother change angle, it is divided into

five different membership functions. Positive angle is divided into two membership

functions, MP and LP, where MP is more positive and LP is less positive. Similarly, negative

angle is divided into MN and LN. And zero angles are represented by the membership

function Z. In all these membership functions positive part corresponds to plantar flexion and

negative part corresponds to dorsiflexion movement.

The output actuation is in terms of angular velocity. So it is also divided into different

membership functions. Positive actuation are represented by eight membership functions,

namely p8, p7, p6, p5, p4, p3, p2, and p1, with p1 being the smallest and p8 being largest.

Similarly, negative actuation are represented by eight membership functions, namely n8, n7,

n6, n5, n4, n3, n2, and n1, with n1 being the smallest and n8 being largest.


Page 53

TABLE 4.2 FUZZY RULE BASE

IF

ANGULAR

VELOCITY

AND

ANGLE

THEN

ACTUATION

P8 Z N8

P7 Lp N7

P6 Lp N6

P5 Lp N5

P4 Mp N4

P3 Mp N3

P2 Mp N2

P1 Mp n1

Z Mp Z

N1 Mp p1

N2 Mp P2

N3 Mp P3

N4 Mp P4

N5 Lp P5

N6 Lp P6

N7 Lp P7

N8 Z P8

N7 Ln P7

N6 Ln P6

N5 Ln P5

N4 Mn P4

N3 Mn P3

N2 Mn P2

N1 Mn P1

Z Mn Z

P1 Mn n1

P2 Mn n2

P3 Mn n3

P4 Mn n4

P5 Ln n5

P6 Ln n6

P7 Ln n7

P8 Z n8


Page 54

Figure 4.14 Membership functions of input Angular Velocity

The membership functions of fuzzy variable Angular velocity are depicted in the

Figure 4.14. The range is set at -300 degrees/seconds to +300 degrees/seconds to

accommodate any type of walking pattern like slow, medium and fast.

The membership functions of fuzzy variable Angle are depicted in the Figure 4.15.

The range is set at -50 degree to +50 degrees to accommodate any type of walking pattern

like slow, medium and fast. We have divided angular displacement into five membership

functions MP, LP, Z, LN, MN. The range of angular displacement is -50 degree to +50

degrees, as the real time angle falls in this range. Triangular membership function is used so

that the membership value (u) is evenly distributed over the entire range reducing the width

of quantization steps of the output actuation fuzzy set.

4.15 Membership functions of Angle


Page 55

Figure 4.16 Membership functions of output actuation

The membership functions of fuzzy variable actuation are depicted in the Figure 4.16.

The range is set at -250 degrees/seconds to +300 degrees/seconds to accommodate any type

of walking pattern like slow, medium and fast. It has been observed from the data acquisition

results that the angular velocity in dorsiflexion phase has a smaller peak than that in the

plantarflexion phase; hence the negative part has a larger range compared to the positive part

for the actuation provided by controller. It has been divided into seventeen membership

functions namely n8, n7, n6, n5, n4, n3, n2, n1, z, p1, p2, p3, p4, p5, p6, p7,p8.

The Rule-base is made by observing the complete gait pattern and considering angular

velocity and angle of the working leg as input and accordingly providing the output actuation

which is the angular velocity required for the defective leg. We can change the Antecedent

connective to AND (product), AND (minimum), OR (product), OR (minimum) to fit our

rule-base. Here all rules are AND (minimum). The degree of support can also be modified to

more or less than unity implying that the rule according to the rules weight being more or

less. Corresponding to the seventeen membership functions of input angular velocity and

output actuation, whether it corresponds to five membership functions of input angle, 33 rules

have been formulated to calculate the output actuation required.

The final step is the defuzzification of the aggregated output fuzzy set in to a crisp

output. The ‘center of area’ method is used for this purpose as given in equation 4.2.

( ) ∫ ( )

∫( ) (4.2)


Page 56

where xi is a running point in the continuous universe of the output variable and ( ) is the

membership value in the membership function. The function calculates the output as the

weighted average of the elements in the output set.

4.3.3 Fuzzy test system

The core of a standard fuzzy controller is the fuzzy rule base. In the absence of expert

knowledge, determining the appropriate rules for a given system is a significant and time

consuming challenge. Once the rule-base is formed, the FLC output can be verified by “Test

system”. The test system gives the value of actuation for each of the input values of angular

velocity and angle. It also shows the rule invoked for each set of input values. The resultant

fuzzy test system depicted as a 3D plot in Figure 4.17. The test system maps the two input

measurements to corresponding fuzzy membership functions. Then according to the fuzzy

inference maps the fuzzy variable to the crisp value. Thus the 3-D figure shows the crisp

input measurements to crisp output stimulation.

The fuzzy inference engine evaluates all the rules that are activated as a result of the

fuzzified inputs and aggregates all the corresponding consequent parts. Finally a crisp control

output is obtained after the defuzzification process. The conclusion of the rule is called fuzzy

implication.

Figure 4.17 Fuzzy test system.


Page 57

TABLE 4.3 THE INPUT-OUTPUT RELATIONSHIP

Ang.(I) MP LP Z LN MN

Ang.vel Actuation ( O)

P8 - - n8 - -

P7 - n7 - n7 -

P6 - n6 - n6 -

P5 - n5 - n5 -

P4 n4 - - - n4

P3 n3 - - - n3

P2 n2 - - - n2

P1 n1 - - - n1

Z Z - Z -- Z

N1 p1 - - - p1

N2 p2 - - - p2

N3 p3 - - - p3

N4 p4 - - - p4

N5 - p5 - p5 -

N6 - p6 - p6 -

N7 - p7 - p7 -

N8 - - p8 - -

Table 4.3 shows the input –output relationship. It is similar to the rule base given,

mapping the output with respect to magnitude and sign of input angular velocity and angle.

When the values are in between the two membership functions of the same antecedent, the

antecedent is going to consider the minimum of the two to compute the output value.

4.3.4 Result analysis

The Figures 4.18 to 4.20 give different instances of the fuzzy controller showing the

angular displacement and angular velocity of both the correct foot and the corrective

actuation generated through the FLC. Each gauge has two needles to show the input (I) from

the correct foot and the output (O) actuation. The waveform shows the PWM generated to

provide the actuation to the motor. Figure 4.18 shows the instantaneous values of angle and


Page 58

angular velocity are close to zero showing that the person is standstill. The actuation provided

and the duty cycle is zero.

(a)

(b)

Figure 4.18 (a) Standstill position (b) Enlarged view of input-output needles.

In Figure 4.19 the input needle in the gauges for angle and angular velocity are at a

negative position showing the dorsiflexion phase of the correct foot while the output needle is

at a positive value indicating that the actuation given by the controller replicates a plantar-

flexion motion for the defective foot. The instantaneous values of angle and angular velocity

are negative. It can also be noticed that the actuation provided to the defective foot is greater

in magnitude than the angular velocity of the correct foot to simulate the pattern observed

during analysis of acquired data. The PWM has positive 10V during the on period and 0V

during off period of the duty cycle providing positive actuation through the motor.


Page 59

(a)

(b)

Figure 4.19 (a) Plantar flexion movement for the defective foot (b) Enlarged view of input-

output needles.

From figure 4.20 it is seen that input needle in the gauges for angle and angular

velocity are at a positive position showing the plantar flexion phase of the correct foot. While

the output needle is at a negative value and magnitude of output is less compared to input

indicating that the actuation given by the controller replicates the dorsiflexion motion for the

defective foot.


Page 60

(a)

(b)

Figure 4.20 (a) Dorsiflexion movement for the defective foot.(b) Enlarged view of input-

output needles.

Table.4.4 gives the output data resulted from FLC. The results show that there is a

good amount of actuation provided by the controller. The control parameters such as angle,

angular velocity, direction and speed to drive the actuator are supported with the proposed

controller design. During the data acquisition phase the gyroscope is first attached to the left

foot to observe the gait pattern of one complete cycle followed by the gyroscope being

attached to the right foot. In both the cases a similar walking pattern has been replicated. The

complete gait pattern of both left and right foot is detected almost synchronously. In this

experimental set up the left foot started of the walking cycle. From the values collected it is

observed that the left foot (L.F) first makes a dorsiflexion motion followed by a plantar-


Page 61

flexion whereas the right foot (R.F) makes a plantar-flexion movement first followed by a

dorsiflexion. A similar pattern is observed for angular velocity values also. Hence the

controller is stable giving desired output for the given input.

TABLE 4.4 CONTROLLER RESULTS IN REAL TIME

Sensor

Placement

Input

Ang.Velocity

(degrees

/sec)

Input

Angle

(degrees)

Output

actuation

(degrees/sec)

Duty cycle

(percentage)

L.F -10.55 -15.83 19.38 16

L.F -19.94 -29.91 41.68 60

L.F -15.56 17.15 21.71 18

L.F 14.73 22.10 -16.07 13

L.F 17.60 26.41 -24.39 41

R.F 12.65 18.97 -15.95 13

R.F 30.36 45.67 -42.02 70

R.F 13.89 -20.81 -26.13 15

R.F -18.37 -27.56 35.66 50

4.3.5 Performance analysis

Input-output data pertaining to left and right foot are taken to analyse the performance

and graphs are plotted. The gait data is taken by placing the gyroscope on each foot. The

graph shown in Figure 4.21 depicts the average angular velocity. Figure 4.22 depicts the

desired output as the actuation provided to left foot in terms of angular velocity. The input

considered is the angular velocity along with angle of right foot. Figure 4.23 represents the

desired input –output relation. When the right foot is in the planatarflexion phase after a 180

degree phase difference left foot is in the dorsiflexion phase. Figure 4.24 represents slow

response of the controller. For 20 input samples output sample is only one.


Page 62

Figure 4.21Average angular velocity for the left and right foot

Figure 4.21Average angular velocity for left and right foot

Figure 4.22 Gait data of the left and right foot using two gyrosensors


Page 63

Figure 4.23 Desired input-output plot

Figure 4.24 Fuzzy controller output for the given input.


Page 64

4.4 CONCLUSION

The control design for an active AFO to provide proper ankle movement is developed

using Fuzzy logic. The designed control system consists of a closed loop system with a

sensor unit providing the feedback through continuous monitoring of the gait orientation. The

sensor data is then fed as input to the FLC where the fuzzy set and rule base are loaded.

Depending on the input rule invoked by the fuzzy controller the control signal is generated

for the actuator in real time. The control parameters such as angle, angular velocity, direction

and speed to drive the actuator are supported with the proposed controller design. The

hardware implementation of FLC is beyond the scope of this dissertation.

To acquire the real time gait data 40Hz sampling rate with buffer size 80 is used. The

draw back in this controller simulation is the poor response i.e for every 20 input samples

there is one controlled output (crisp value). The improvement in the response time and

accuracy are considered as future scope of this work.

Design of portable, untethered AFO which can be used for rehabilitation outside the

clinical therapy is an objective of this research work. An embedded control system using

ATmega 328 microcontroller is designed. The embedded system design for an AFO and its

control mechanism are discussed in the next chapter.

intelligent control system for an active afo with the...

Documents