You are on page 1of 6

Proceedings of the 7th IASTED International Conference

Biomedical Engineering (BioMED 2010)


February 17 - 19, 2010 Innsbruck, Austria

INTELLIGENT DEVICE FOR CONTROL OF ACTIVE ANKLE-FOOT


ORTHOSIS

Veneva I,
Institute of Mechanics, Bulgarian Academy of Sciences,
Acad. G. Bonchev Str., Bl.4, Sofia 1113, Bulgaria;
Email: veneva@imbm.bas.bg

ABSTRACT and monitoring system for gait analysis using tactile


The main aim of this research is the development of an sensors and active ankle-foot orthoses during normal level
autonomous adaptive system for actuation, data walking. The device is used to help or rehabilitate persons
acquisition and control of active ankle-foot orthosis. The with control disorders and other weaknesses of ankle foot
system is composed by microcontroller, driver and sensor complex.
system, which application is to actuate and position the
foot orthotic segment, as well as data acquisition,
communication and friendly oriented software for 2. Method
interpretation of the data obtained during the walking. The
research work combines hardware and software design of The active ankle-foot orthoses has two basic components:
the intelligent control device with graphical interface for Electro-Mechanical and Control Unit. The Control Unit
representation and analysis of the data acquired during generates flexion/extension motions and controls the
human motion. Electro-Mechanical Unit (direct drive actuator and ankle-
A laboratory model of the proposed system was foot orthoses with hinge joint) [4, 5, 8]. The used Voice
implemented to demonstrate its autonomy and verify Coil Actuator (VCA) has two build-in mechanical stops,
experimentally its functionality. The proposed control which limit its range of motion to slightly less than 30
device can be used in several applications involving degrees. The actuator and the hinge joint replace the
human motion analysis and control of different types of damaged muscle complex like a tendon in series with a
orthoses used for gait correction. muscle. The system uses a position transducer mounted
on the hinge joint to detect flexion or extension. The
flexion and extension motions are controlled via a two
KEY WORDS tactile sensor arrays incorporated in the foot part of AFO
Prosthetics and Orthotics, Active ankle-foot orthoses, and in the insole of the healthy leg that triggers the foot
Control, Rehabilitation robotics position.

1. Introduction
Active ankle-foot orthoses (AAFOs) are orthotic devices
intended to assist or to restore the motions of the ankle-
foot complex and are based on force-controlled actuator
and biologically inspired micro-control schemes.
There have different ideas of an actively powered orthotic
device using hydraulic and pneumatic devices or DC
motors which have been researched to provide active
assistance to the individuals with paraplegia, stroke or
drop foot [2, 3]. An active ankle-foot orthosis with a
force-controllable series elastic actuator (SEA) was also Figure 1. Ankle-foot orthosis with direct drive actuator
designed by Blaya and Herr [1] capable of controlling
orthotic joint stiffness and damping for plantar and Active ankle-foot orthosis is a system with one degree of
dorsiflexion ankle motions. freedom which foot segment is connected to the shank
segment by a rotational joint. A direct drive actuator is
We propose an autonomous adaptive device for actuation, attached laterally to the AFO. A feedback with
data acquisition and control of active ankle-foot orthosis. Proportional-Integral-Derivative control was used to
The aim of the work is to present an autonomous control

680-080 100
estimate the trajectory of the foot and position the monitoring mode these data are transferred through the
actuated foot segment of AFO when the foot rotates about RS-232 lines to a graphical user interface for visualization
the ankle. Realizing flexion/extension the actuator applies and interpretation.
a torque adequate to the joint position of the human ankle
during level ground walking [6, 7]. To determine system
state, the sensors measured ankle angle, foot position and
gait speed. During the swing phase, where the clearance
of the toe is released, electro-mechanical system must
actively adjust the flexion of the orthosis by actuator
movement and keep this position till the heel strike
appears. Thus the ankle torque has to be modulated from
cycle-to-cycle throughout the duration of a particular gait
phase.

3. System Design Figure 3. Two tactile sensor arrays TR incorporated in the foot
part of AFO and TL in the insole of the healthy leg.
The complete autonomous system consists of four
primary components - sensing, data acquisition, A very important design characteristic is the system
communication and friendly oriented software for power source. One of the main objectives of this system is
interpretation of the data (fig.2). to reach a significant level of autonomy.
Many batteries technologies have been lately developed.
Nowadays Lithium Ion is the battery technology imposed
in market having a good power density.

3.1 Data Acquisition Unit

The data acquisition unit prototype (fig.4, 5) has been


realised using microcontroller ATmega128 (Atmel Co.),
featuring basic hardware peripherals such as analog to
digital converter (ADC), USART for RS232
communication and a timer with Pulse Width Modulation
(PWM) output. The PWM channel is connected to the
driver to control the direction and speed of the motor by
varying the duty cycle of the PWM output. By varying the
current flow thought the coil the speed and torque of the
Figure 2. Autonomous control and monitoring system with motor can be varied. The position control is handled by
active ankle-foot orthoses
electronics according the outputs from a set of two sensor
arrays - tactile sensors TR incorporated in the foot part of
The sensor system has mounted into two basic AFO and TL in the insole of the healthy leg. The sensors
components: insole for the healthy leg and ankle-foot change their outputs throughout the stance and the swing
orthoses. During the walking the acquisition unit gathers phases of walking.
and digitizes the information from the sensors. In

Figure 4. ATmega128 controlling a VCA

101
Additional feedback element - angular position sensor RP
is attached to the moving parts of the motor assemblies to
sense the velocity and position.

Figure 6. Graphical program written in MATLAB for


visualization of human motion data
Figure 5. Data Acquisition module prototype
3.3 Control Algorithm
The sensors work together to detect walking over one
given interval of time and to collect the following The control algorithm is based on the biomechanical
parameters: ankle joint angles, foot (heel and toe) contacts interpretation of the locomotion. During each gait cycle,
and foot velocities. The microcontroller ATmega128 by measuring the total time TL (for the left leg) and TR
collects sensor data and creates a serial bitstream for (for the right leg) when the foot remains in contact with
transfer. The data are collected on four VS buffers, ADC the ground the microcontroller estimates the forward
buffer; PWM duty-cycle buffer (Tab.l). speed and modulates the swing phase flexion and
extension in order to achieve quite normal lower limb
dynamics (fig.7). Thus the joint torque of the actuator was
Tab1. Buffers for data collection automatically modulated to match patient specific gait
Signal PWM ADC TL1 TL2 TR1 TR2 requirements to permit smooth and natural heel strike to
forefoot transition. The tactile sensors and a rotary
Bufer PWMb ADCb VS0b VS1b VS2b VS3b potentiometer measure ankle joint position in real-time
and send signals to the microcontroller.
File sigPWM sigADC sigVS0 sigVS1 sigVS2 sigVS3

Rotary potentiometer provides a signal ADC (in milivolts)


during swing and stance phase. VS0 and VS1 are recorded
digital signals from the switches mounted under the heel
and the toes part of the right leg insole. Signals from the
lower insole of the right leg are VS2, VS3.
Secondary functions for the electronics in control
application is to ensure that the speed and overload
detection are as desired by closed loop control.

3.2 Communication and Graphical User Interface

Using USART the controller transmits data packet to the


PC through the RS232 serial interface. A graphical
program module written in MATLAB receives the data
and visualizes it in its own window, giving us the
representation of the signals. The developed software
allows different mathematical operations with the data, Figure 7. Signals from the external sensors
visualization and printing of the results, graphics and
tables.
These sensors data are then used in every step of the
control algorithm (fig.8) in order to optimize the heel to
forefoot transition during the stance phase TL or swing

102
phase TR of walking. The microcontroller receives the 4. Experimental Results
diagnostic information about the system from the sensors
and generates the torque command to the driver. The simulation tool used to simulate the
electromechanical system was Matlab Simulink. A
Simulink block diagram model is a graphical
representation of a mathematical model of the dynamic
system where the orthosis is represented by its
SimMechanics model and the actuator is represented by
its transfer function [6].
We simulate the model in Inverse Dynamics mode to
compute the joint torque required to rotate the foot in
desired position. Our aim is to analyze the torque,
necessary to produce at the ankle joint 15 degrees
rotation. During the simulation the geometry of the
orthosis is presented as a double pendulum.

Figure 8. Control algorithm


Figure 10. Control Representation of the orthosis geometry as a
To get a more exact torque it is important to have a double pendulum in SimMechanics
feedback and PID control in order to maintain stability
when a foot load is applied [7]. Once we know the computed torque, we can calculate the
required dynamic motor torque and to decide which is the
correct motor with appropriate parameters for joint
actuation.

Tz = Td Tc Tg , (3)

Td = ( J c + md 2 ) q&& + kq& + mgd sin q , (4)


Figure 9. PID controller with feedback where Td is the driving torque; Tc the torque caused by
the friction; Tg - torque caused by the gravity; Jc is the
The controller reads the system state y, by a rotational foot body inertia moment; q generalized coordinate.
potentiometer, subtracts the measured angle from a Tthe weight g of the feet and the lever arms are
desired reference y0, to generate the error value, e. The determined from the conventional anthropometric tables
error will be managed in three terms - the proportional, Tp, knowing the body weight.
the integral, Ti, and the derivative, Td, terms are summed
to calculate the output based on the PID algorithm: The system parameters are:
m1 = 0.7 kg, m2 = 0.20 kg, m3 = 0.10 kg,
de(t )
t
1
u (t ) = k p e(t ) + e( )d + Td (1) l2 = 0.025 m,
Ti 0 dt
J33 = 3.7 kg cm2 d = 0.15 m,
Approximating the integral and the derivative terms to get
the discrete form this gives the controller: where l is the length of the body; d is the distance between
the ankle joint and the centre of mass of the body; m is the
n sum of the masses of the foot m1, the orthosis m2 and the
u ( n) = K p e(n) + K i e( k ) + K d ( y ( n) y ( n 1)) (2) actuator m3.
k =0
Finally, there is a relationship between the inertia J of a
where n is the discrete step at time t. body about ankle axis and its rotational inertia J33, about

103
an axis through its center of mass, parallel to the ankle angles, foot (heel and toe) contacts and foot velocities.
axis. The relation is shown by equation Angle sensor (rotary potentiometer) detects joint position
and provides a signal ADC during swing and stance
J = J33 + md2 = 3.7 +0.7225 = 0.01612(kg.m2) (5)
phase. The tactile sensors acted like a switch, turning off
Once we know the actuator parameters and computed when the foot was over the sensor and turning on when
torque, we can verify that this is the correct answer of the the leg moved away from it.
system simulation by analyzing driven angular motion for
During walking the microcontroller collects sensor data in
the articulation of the ankle joint (foot) in Matlab
four VS buffers, ADC buffer and PWM duty-cycle buffer.
Simulink (fig.11).
In monitoring mode the data are transmitted to the PC
Dynamic2::ANGLE1_ANGLE
through the RS232 serial interface. A graphical program
130
module written in MATLAB receives the data and
120 visualizes it in its own window, giving us the
110
representation of the signals (fig.12). VS0 and VS1 are
M e a su re

100
90
80
recorded digital signals from the tactile sensors mounted
70 on the orthosis. Signals from the lower insole of the
healthy leg are VS2 and VS3.
60
0 5 10 15 20 25 30 35
Time (Sec)

Dynamic2::Velocity (deg/sec)

42,37527548
32,37527548
22,37527548
12,37527548
M e a su re

2,375275483
-7,624724517 0 5 10 15 20 25 30 35
-17,62472452
-27,62472452
-37,62472452
Time (Sec)

Dynamic2::Acceleration (deg/sec^2)

52,62589887
42,62589887
32,62589887
22,62589887
M e a su re

12,62589887
2,625898875
-7,374101125 0 5 10 15 20 25 30 35
-17,37410113
-27,37410113
-37,37410113
-47,37410113
Time (Sec)

Dynamic2::ServoMotorNetLoad (mm N)

1,85152E+11

1,65152E+11

1,45152E+11
M e a su re

1,25152E+11

1,05152E+11

85151969830

65151969830
0 5 10 15 20 25 30 35
Time (Sec)

Figure 11. Simulation results: angular motion, velocity,


acceleration and torque for the ankle joint

In order to test the control algorithm and system


functionalities a laboratory model of orthosis with hinge
joint and attached laterally direct drive actuator was
designed [8]. A healthy subject equipped with the sensors
mounted under the heel and the toes part of the insole
(TR1, TR2 for the right leg and TL1, TL2 for the left leg),
performs different trials of slow and normal level
walking. During walking the motion of the orthosis is
observed - if the time and phase parameters of the orthosis
coincide with these of the right leg.
The experimental results show that the sensors work
Figure 12. Visualization of human motion data in MATLAB.
together to detect walking over one given interval of time Data are collected on four VS buffers, ADC buffer and PWM
and to collect the following parameters: ankle joint duty-cycle buffer

104
Results indicate that sensing shows reliable and accurate The system can be used for assisting the human walking
gait phase detection possibilities. First results indicate that in cases of injured ankle-foot complex and biomechanical
data from the sensors already gave enough redundant analysis of human locomotion.
information for reliably detect initial contact (heel strike)
and toe-off in normal walking. Detecting heel-up and
stance, and ankle joint angle are available using this Acknowledgement
configuration of the sensors. During walking angles in the
orthosis ankle joint could be estimated within 3 degrees The study was supported by European Social Fund
accuracy using only a rotational potentiometer. (Grants BG051PO001-3.3.04/40).

5. Discussion References

The features of normal walking are important for overall [1] J. Blaya, H. Herr, Adaptive Control of a Variable-
autonomous control and are not achievable with most Impedance Ankle-Foot Orthosis to Assist Drop-foot Gait,
mechanically passive orthoses. We assumed that normal IEEE Transactions on Neural Systems & Rehabilitation
gait is symmetrical and that deviation from a reference engineering, vol. 12(1), 2004, 24-31.
pattern is a sign of disability. To analyze asymmetry, the
step time of the affected side (TR) was subtracted from [2] S. Kaneoko, H. Yano, A new concept of dynamic
the unaffected side (TL). The difference in stride lengths orthosis for paraplegia: the weight bearing control (wbc)
should be zero for symmetric gait. orthosis, Prosthetics and Orthotics Intern., 1997, pp. 222
228.
The objective of the control algorithm was to select
appropriate ankle torque value for each state. That would [3] N. Wasylewski, B. Ruthenberg, An experimental
result in an improved gait in terms of biological realism device for investigating the force and power requirements
and symmetry between affected and unaffected sides. of a powered gait orthosis, Journal of Rehabilitation
Specifically, the aim of the controller was to achieve a Research and Development, 1997, pp 203213.
biologically realistic maximum flexion angle during the
swing phase and a biologically realistic early stance [4] I.P. Veneva, Autonomous Control System for Gait
flexion/extension at the heel strike critical to overall drop Analysis, Proc. 9th International Robotics Conf., Varna,
foot. Bulgaria, 2007, pp. 230-236
The developed device provides broad information for both
control and gait analysis. The data from the sensors are [5] I.P. Veneva, Y.E. Toshev, Control algorithm for
used in every step from the control algorithm. The ankle-foot orthosis and applications, Proc. 9th
experimental data discussed in this paper can be used in International Robotics Conf., Varna, Bulgaria, 2007, pp.
cases of the drop foot treatment and lower limb 225-230
rehabilitation to enhance the AAFO functional
performance and to improve the patient gait. [6] I.P. Veneva, G. Boiadjiev, Mathematical Model of
Electromechanical Controlled Module in Orthopedics,
Proc. 6th Int. Conf. on Bionics and Prosthetics,
Biomechanics and Mechanics, Mechatronics and
6. Conclusion Robotics, Varna, Bulgaria, 2008, pp. 22-26.
The adaptation schemes presented here automatically
[7] I.P. Veneva, PID Control for Active Ankle-foot
modulate actuator torque in order to optimize the heel-to-
orthosis Positioning, Edition Machine Union, Year V,
forefoot transition throughout the stance and the swing
Vol.6, 2008, ISSN1310-3946, pp. 52-57
phases of walking. The proposed possibilities to integrate
active systems into AFOs hinge joint provide gait control
[8] I.P. Veneva, Design and Implementation of Device for
during real-use situations and enhance the functional
Control of Active Ankle-Foot Orthosis, Proceedings of
performance using only local sensing and computation.
the 11th National Congress on Theoretical and Applied
The system controls the orthosis functionalities, records Mechanics, Borovets, Bulgaria, 2009
the data received from sensors during the gait and
transfers recorded data to graphical user interface for
visualization and future analysis.

105

You might also like