Anda di halaman 1dari 16

Int. J. Experimental and Computational Biomechanics, Vol. X, No.

Y, xxxx

Human postural control during standing posture with


a muscle-tendon actuator
Daniel Boari Coelho*
Human Motor Systems Laboratory,
School of Physical Education and Sport,
University of So Paulo,
Av. Prof. Mello Moraes, 65,
So Paulo, SP. 05508-030, Brazil
E-mail: daniboari@usp.br
*Corresponding author

Marcos Duarte
Biomedical Engineering,
Federal University of ABC, Brazil
Office 608-1, Block A, Santo Andr, Brazil
E-mail: marcos.duarte@ufabc.edu.br
Abstract: The way humans control the balance in the upright posture is not yet
fully understood and complications due to deficiency in the postural control are
of relevant concern. In order to achieve an early diagnosis on disorders
affecting human postural control, the theoretical modelling coupled with
computer simulation emerges as an option used by researchers. Looking for a
model for a greater understanding of human postural control system, this paper
presents a new computational model dedicated to the study of human upright
posture, with the presentation of a model of muscle-tendon actuator. The way
in which the subsystems of human control, such as the dynamic muscle, neural
control, visual, vestibular and somatosensory contribute to the stable posture
was also focused. This model simulate the responses of fluctuations in the
centre of mass and centre of pressure and offers the possibility of measuring
neural activation, decomposition of reactive torque and participation of each set
of muscle groups to maintain posture. It reproduces, in a model of feedback, the
existing cross-correlation between experimentally observed and the neural
signal and the centre of pressure.
Keywords: postural control system; ankle strategy; inverted pendulum; sensory
system.
Reference to this paper should be made as follows: Coelho, D.B. and
Duarte, M. (xxxx) Human postural control during standing posture with a
muscle-tendon actuator, Int. J. Experimental and Computational
Biomechanics, Vol. X, No. Y, pp.000000.
Biographical notes: Daniel Boari Coelho is a graduate at Engineering,
Masters at Biomedical Engineering and PhD candidate in Neuroscience
Program, University of So Paulo.

Copyright 200x Inderscience Enterprises Ltd.

Comment [t1]: Author: Please


reconstruct the sentence to make the
intended meaning clear to readers.

D.B. Coelho and M. Duarte


Marcos Duarte

Introduction

The way humans control the balance in the upright posture is not yet fully understood and
complications due to deficiency in the control of posture are of relevant concern.
Simplifying, during static stance a minimum and necessary task of postural control
system is to maintain the vertical projection of centre of gravity (COG) or centre of mass
(COM) of the individual within the base of support, defined by the base of the foot. The
limits of stability depend on the area of the support-base, the vertical projection of COM,
COM height, and weight of the mass to be controlled. The limits of stability also depend
on the interaction between position and velocity of COM. Therefore, if the individual is
very close to the edge of the support-base and velocity of COM is high, it is more
difficult to recover the stability compared to the individual in the centre of the
support-base, with equally high speed. A real limit of stability in normal subjects is
determined by biomechanical and neuromuscular constraints such as strength and speed
of muscular response. The control of upright posture uses information on monitoring the
position of COM of each body member (Horstmann and Dietz, 1990; Morasso et al.,
1999; Morasso and Schieppati, 1999) through orientation from sensory information
(Gurfinkel et al., 1995b, 1995a).
Stability is achieved by generating moments of force on the joints of the body to
neutralise the effect of gravity or any other disturbance in a continuous and dynamic
process during a stay in a given posture. Physically, the body that interact to generate
forces are: support interface (feet/floor), body parts and internal structures such as bones,
muscles, joints and tendons. To regulate balance, the system needs information about the
relative positions of body segments and the magnitude of the forces acting on the body.
For this purpose, three classes of sensors can be used by the body: somatosensory, visual
and vestibular. These receptors act in complex, integrated, redundant and differently for
each disorder on the human body. The passive properties of the musculoskeletal system,
particularly the stiffness of biological structures also play an important role in
maintaining the balance. The control of postural balance in a person is highly affected by
the nature of the task, environmental conditions and the sensory information available,
and personal conditions.
Much progress was made on understanding postural control but this control system is
still not fully understood. Recently, based on important empirical findings, some
researchers have proposed new forms of control that is different from classical theories of
postural control system (Asslander and Peterka, 2014; Collins and De Luca, 1993; Gatev
et al., 1999; Jacono et al., 2004; Morasso and Sanguineti, 2002; Peterka, 2000; Winter
et al., 1998). There are still many questions about both the properties of muscle (Loram
and Lakie, 2002a, 2002b; Winter et al., 2001) and the demand of the central nervous
system (CNS), such as the involvement of the sensory systems on the stabilisation of
human postural sway. There is still controversy about the type of control selected; some

Comment [t2]: Author: Please


provide biographical details for
Marcos Duarte of not more than
100 words.

Human postural control during standing posture

evidences suggest that posture is stabilised by anticipatory control in cooperation with the
elastic properties of muscle. These new models and theories have not yet been fully
tested and discussed, and have diverging aspects and this search is still relevant in the
scientific community. There are therefore, many questions to be answered, like what is
the participation of the CNS in postural control, which values of the muscular properties
guarantee the stability of the system.
The purpose of this paper is the development of computational models to simulate the
human postural control in upright stance based on different theories, testing some of the
latest empirical findings on human postural control. In this work we specifically address
five questions.
1

What is the effective intrinsic, mechanical ankle stiffness during quiet standing and
when balancing the inverted pendulum?

Can stability be maintained with small values of intrinsic ankle stiffness?

What mechanism for controlling the body COM is suggested by these results?

Is the PD controller capable of facilitating robust balance during quiet standing,


despite long neurological time delays?

Which is the contribution of the available sensory information to human standing?

Methods

We will use the inverted pendulum model controlled by different forms of control which
may include mechanisms for feedback and feedforward. This modelling will allow a
better understanding of the new findings and theories on human postural control. For this
purpose, we will reproduce computationally the body in its sagittal plane, modelling it as
a rigid body, with only a COM and pressure, of two segments, comprising the foot and
the rest of the body, united by a joint like hinge. At first, the model will be considered on
its wide scope, containing the sensory systems and control systems, without specifying
which the components of these systems are. Disruption of this system will be done by
introducing a noise with Gaussian distribution of appropriate range, reproducing the
environmental factors that interact with humans (Conforto et al., 2001; Fukuoka et al.,
1999), analysing what is the response of this model to the disturbance. To play the neural
activity, a model for feedback controller with a proportional, integral and derivative
(PID) will be applied. As the mathematical model of posture involves the
musculoskeletal component, the CNS model will be introduced to control muscle
involvement in this movement, with a stiffness and viscosity itself. Muscle activity will
try to replicate the dynamics of muscle contraction, using the model of Zajac (1989),
according to the adaptation of Menegaldo et al. (2006, 2003). On a second instance, the
sensory systems are specified in visual, vestibular and somatosensory, analysing the
interference in which these various components create the control of human postural
balance. Observing the human response to changes in one or more of their sensory
systems, we observe the gain and the involvement of visual, vestibular and
somatosensory in stability in human posture.

D.B. Coelho and M. Duarte

All simulations were performed using the Simulink software version 5.0 of Matlab
6.5 (MathWorks Inc., Natick, MA USA), considering a duration of 100 seconds. The
simulation parameters used were fixed pitch algorithm with a resolution of differential
equations associated with the model Dormand-Prince (ode5), 0.001 of tolerance. The
disturbance used is a Gaussian function block generated by Matlabs band limited white
noise (white noise) with zero mean, standard deviation of six and sample time of 0.4
seconds, going through a low pass filter of first order with time constant of 0.5 s,
generating a maximum torque (Td) in the ankle joint of about 6 Nm. The model
parameters consist of neural controller gains (KP, KD, KI), passive properties of muscle
(KPpas, KDpas), sensorial delay (d) and electromechanical delay of muscle-tendon actuator
(m) delay do efferent signal (efe). The balance problem can completely be described by
the movement of the whole-body COM, the distance l from the axis of rotation to the
COM remains constant, and the excursions of the COM are small with respect to l. The
gravity force vector mg is located at the COM, pointing vertically downward. The forces
on the feet can be represented by a single ground reaction force vector, located at the
centre of pressure (COP). A schematic representation of the model is shown in Figure 1.
Figure 1

Representation of the model including a muscle-tendon actuator in the postural control


system (see online version for colours)

The position of COM of the human body can be approximated to be located above the
second sacral vertebra, 5 to 10 cm posterior to the ankle joint, simulations were
conducted considering an angle of the inverted pendulum of 0 and 4 leaning forward in
relation to the vertical.
One can summarise the operation of the model as follows:

The model takes as disturbance torque (Td), which is shifting the reference position,
until the postural control system acts on the muscles producing a torque muscular
reaction to the disturbance (Tc).

The torque disturbance produces a posture change in angular position of the


biomechanical model, which is sensed by a sensory system, modelled with a sensory
delay (d) of about 100 ms (Schieppati and Nardone, 1997).

There is a comparison of the angular position signal with a reference angle, and
therefore an error on this angle, which is sensed by the CNS, designated here as a
PID controller and the controller gains neuromuscular dimensionless, represented by

Human postural control during standing posture

KP, KD and KI, adjusted by the tuning method of Continuous Oscillations


Ziegler-Nichols (Ziegler and Nichols, 1942).

The PID controller returns a value of a neural excitation signal (u(t)).

The time delay (efe) of the neural excitation signal is 80 ms, due to the distance
travelled by the efferent neural control signal of the actuators with muscle-tendon
(Mrachacz-Kersting et al., 2007).

The excitation signal are imputed to neural models of muscle-tendon actuators, here
represented by transfer functions identified from the models proposed by Menegaldo
et al. (2006, 2003), generating a torque control (Tc). In this simulation, we tested two
types of actuators: a model that included the identification system agonist-antagonist
muscle (anterior tibialis + posterior tibialis, and soleus gastrocnemius), and another
that modelled the muscle-tendon actuator plantar flexors (soleus + gastrocnemius +
anterior tibialis) and dorsiflexion (posterior tibialis). The latter presents a switched
situation such that the neural excitation provided by the PID controller is positive,
the dorsiflexion group is encouraged, and in the reverse case, the plantar flexor group
is stimulated. Despite this switch, the modelling provides moments of co-contraction
due to delays and dynamic muscle model, where the two muscle groups receive
neural signals.

The model of muscle-tendon actuator has an electromechanical delay of 80 ms


(Conforto et al., 2006; Hopkins et al., 2007) generating a delayed torque control.

A passive muscle torque is added to the active torque control, generating a reactive
torque disturbance. This torque due to passive muscle properties is measured in
accordance with the position and angular velocity and is represented here by the
passive properties of muscle [KPpas KDpas equal to respectively 64% mgh [Nm/rad]
(Casadio et al., 2005) and 175 Nms/rad].

Then, the muscle torques are inserted into the biomechanical model in order to
compensate the action of inertial and gravitational forces and torque disturbances.

The cycle repeats itself, so as to maintain the posture around a reference position.

In a second step, we assessed the influence of the systems involved in sensory perception
and distribution of the limbs to maintain posture by modelling the sensory systems based
on van der Kooij (2001) and Zupan (2002).

Nonlinear identification of muscle-tendon actuator of the lower Lomb

We modelled the dynamics of muscle contraction as proposed by Zajac (1989), testing


parametric structures auto-regressive (ARX), auto-regressive moving average (ARMAX)
and output error (OE).
For a mathematical model that relates an output variable dynamically with one
entrance, it is necessary a significant correlation between these variables. In this model,
there are two input variables, the level of neural excitation (u (t)) and the length of the
muscle-tendon actuator. For the identification of muscle-tendon actuator, the input

D.B. Coelho and M. Duarte

variable is intricately linked to the output (in this case the force) the level of neural
excitation (u (t)), keeping the muscle-tendon length and equal to a fixed length standard
for identifying, considering the upright posture by aligning vertically. Whereas muscle
length normalised time-invariant, the identification of the muscular system disregards the
influences of passive muscle properties, but only the dynamics of neural activation. Such
a procedure will contribute to a better understanding of the influences of each component
of muscle strength to maintain posture. The output of the model is the torque applied to
the ankle joint given by the regression curves of the geometry of the lower limb of
Menegaldo et al. (2006, 2003).
For this identification, we tested two types of input signals, called PRBS (pseudo
random binary sequence) and natural excitement, varying randomly between the values
of peak to peak. The inconvenience of the PRBS signal is to vary only between two
levels, not equivalent to that observed experimentally, so we chose to shape the input
signal u(t) with natural excitement. The residence time of the signal was equal to 0.004
seconds, obeying the rule to be equal to four times the sampling interval. For this case,
we used a sampling time equal to 0.001 seconds and the system was simulated for a
period equal to 100 seconds.
Due to the complexity of the muscle-tendon model, nonlinear identification
techniques were employed, as proposed by Wiener (Pearson and Pottmann, 2000). The
functions of static nonlinearity of the Wiener model was established using a step in neural
excitation signal u(t) with amplitudes of 0.1 to 1 seconds each in the range 01 and
compared the outputs of model Menegaldo et al. (2006, 2003) with those obtained by
linear identification. From these results, it was determined a polynomial of tenth order.
Subsequently, the identification was made to a muscle-tendon actuator grouped into
plantar flexor (soleus + gastrocnemius + posterior tibialis) and dorsiflexion (anterior
tibialis). The sign of neural excitation u(t) was characterised as a natural excitement with
residence time of 0.04 seconds and sign sampling time of 0.01 seconds.
Finally, we sought to determine the characteristics of the muscular system
agonist-antagonist (anterior tibialis + posterior tibialis + soleus and gastrocnemius). For
this model, the amplitude of neural excitation was 1 to 1, switched from 1 to 0 for the
antagonist muscle system and 0 to 1 for the agonist muscle, with the same properties of
the signal above.
The models with better AIC Akaike index and lower levels of function-loss and FPE
(Akaike, 1974) are presented.

System model control including model of vestibular system

The vestibular system has its predominance in the detection of speed changes in the
position of the centre head joint for a reference. Thus, we included a system for feedback
of speed considering the participation of such a system, with the transfer function
according to equation (1) (Fernandez and Goldberg, 1971).
Gvest =

600 s 2
600 s 2 + 106 s + 1

(1)

Human postural control during standing posture

System model control including model of visual and somatosensory


systems

The work of Chen et al. (2002) presented a mathematical model of the visual system, as
the transfer function of equation (2), which can be introduced in the maintenance of
postural sway to the upright posture.
Gvisual =

0.8167
0.6450 s + 1

(2)

The transfer function of the somatosensory system was given by the work of Peterka and
Loughlin (2004), according to equation (3).
Gsomatossensorial = 1

(3)

For this model, we sought to identify the weight assigned to each component of the
sensory system involved in maintaining posture.
For the simulations, we used an anthropometric model considering a mass of 80 kg
and height of 1.80 cm using an anthropometric scale of Lafond et al. (2004).

Results

The parameters of the controller, here representing the CNS, present as units of measure:
1/rad for the proportional gain Kp, s/rad for the derivative gain Kd and 1/(rad.s) for the
integrative Ki. If the neural excitation provided by the PID controller is between 0 and 1,
the dorsiflexion group is stimulated; in the reverse situation (between 0 and 1), the
plantar flexor group is stimulated. Given the variety of inter-relationships among the
model parameters and sway measures, we predicted that model parameters could be
estimated from a particular set of sway measures using an optimisation procedure
(Menegaldo et al., 2006).

Nonlinear identification of muscle-tendon actuator of the lower Limb

Table 1 shows transfer functions for the models muscle. Note that the functions that have
better representation are characterised by second-order without dead time.
The functions of nonlinear static Wiener model are characterised by second-order
polynomials, with order parameters greater than zero, according to Table 2. In the
absence of an order parameter zero, it is observed that the initial values of linear and
nonlinear are equivalent.
Table 1

Linear identification of muscle-tendon actuator in the lower limb

Muscle
Gastrocnemius
Soleus

Linear function
Gca =

Gca =

35.68 s + 65,200.0
s 2 + 1,127.9 s + 4,156.0

127.9 s + 198,500.0
s 2 + 108.6 s + 3, 272.0

D.B. Coelho and M. Duarte

8
Table 1

Linear identification of muscle-tendon actuator in the lower limb (continued)

Muscle

Linear function

Tibialis anterior

Gca =

138.2 s + 137,100.0
s 2 + 268.9 s + 9,359.0

Tibialis posterior

Gca =

86.49 s + 74, 260.0


s 2 + 275.4 s + 9,503.0

Table 2

Nonlinear function in the Wiener model

Muscle

Nonlinear function

Gastrocnemius

Tc = 0.1119 T 2 + 0.0025 T

Soleus

Tc = 0.0274 T 2 + 0.0024 T

Tibialis anterior

Tc = 0.1173 T 2 + 0.0025 T

Tibialis posterior

Tc = 0.2174 T 2 + 0.0024 T

The transfer function of the grouped muscles gastrocnemius, soleus and tibialis posterior,
is shown in equation (4).
3.12 s + 106.28
0.0052 s 2 + 0.14 s + 1
And the function of static nonlinearity : Tc = 0.0221 T 2 + 0.0012 T

Gca =

(4)

Such identification is shown next to the Menegaldo et al. (2006, 2003) template with a
correlation coefficient of 0.61.
The agonist-antagonist system of the lower limb during stance in relation to the ankle
joint can be identified by equation (5).
Gca =

243.3 s + 1.745 105


s 2 + 233.4 s + 5,384

And the function of static nonlinearity : Tc = 0.0349 T

(5)
2

+ 1.592 T 10.744

The correlation coefficient of this model, in relation to the Menegaldo et al. (2006, 2003)
is 0.52.

System model control for position feedback

The displacement of COP and COM were calculated for a reference angle of 4. Gain
factors of the PID controller were adjusted: KP (proportional gain) equal to 6.0, KD
(derivative gain) of 0.5 and KI (integrative gain) of 0.8. The parameters of the PID
controller, here representing the CNS, such that the displacements of the COP to present
their properties (velocity and RMS value) equivalent to those described in the literature
(de Freitas et al., 2009; Duarte and Freitas, 2005). Figure 2 shows the signal of the neural
stimulation.

Human postural control during standing posture


Figure 2

Sign of neural excitation for an agonist-antagonist muscle system (see online version
for colours)

There is predominance in the performance of the plantar flexor group (given by the
negative sign of the control) over the action of the dorsiflexors. This is plausible, noting
that the angle of the reference posture is aligned 4 forward in relation to the vertical.
The decomposition of the torque signal at the ankle joint can be performed in control
torque (phasic and tonic) and passive muscle torque, with the tonic control signal of
about 20 Nm toward the plantar flexor group.
This signal has a COP of 0.8 cm RMS and COP speed of 1.3 cm/s.
In a second simulation, the angle of reference was changed to 0, with the same
analysis being performed. Factors gains of PID controller were adjusted KP (proportional
gain) equal to 10.0, KD (derivative gain) of 0.7, without integrative component. The
signal had amplitude of COP displacement similar to when the position of the pendulum
was leaning forward. However, the significant difference lies in the dynamics of neural
activation. The average of the signal, the value 0, changes drastically when compared to
the model with the pendulum tilted forward. As the reference position of the posture is in
the value of 0, the neural stimuli activate both plantar flexor muscle groups as
dorsiflexors. In this angular position, the decomposition of the torque control components
and phasic muscle passive properties were similar in amplitude, mean and standard
deviation. The tonic component of the control, in turn, is not present in this model.
When the muscular system was replaced by agonist-antagonist actuator clustered
plantar flexor and dorsiflexion, so that there is a selective stimulation of the dynamics of
neural activation, keeping the body aligned to 4 from the vertical, the controller
parameters change to three in the proportional gain, derivative gain at 0.7 and two
integrative gain. The change happens to the neural excitation signal, which has its mean
shifted towards the plantar flexor group. For this model, there is activation of the
dorsiflexors to maintain posture.

D.B. Coelho and M. Duarte

10

System model control including a model of vestibular system

As proposed in the work of van der Kooij (1999), the weight of information from the
vestibular system that best represented the postural sway was 0.2. For this value of gain
of the vestibular system, the values of the controller were 14 for the proportional gain,
derivative gain for the 5 and 2.7 for the integrative gain. These values of parameters of
neural control cause a tonic torque about 20 Nm in the direction of plantar flexor group.
There is a greater neural activity in this model compared to models that did not include
models of the vestibular system. This increase in neural activity does not cause greater
amplitude of oscillation of the COP, keeping an RMS value of 0.7 cm and average speed
of 1.6 cm/s.

10 System model control including a model of sensorial systems


For RMS values of COP of 0.9 cm and average speed of 1.7 cm/s, the neural controller
gains were around 14 for the proportional gain, five for the derivative gain and five for
the integrative gain. These values were obtained for postural balance; the gains of the
sensory system being 0.1 for the vestibular system, 0.3 for the visual system and 0.6 for
the other sensory systems (included the somatosensory system). Such weights of the
sensory systems are similar to the experimental and simulated results of van der Kooij
(1999).
Figure 3

RMS, speed and frequency (frequency band with 80% of spectral power) of the
oscillation COP in the anteroposterior direction for different models of the postural
control system

As can be seen in Figure 3, RMS, average speed and frequency of oscillation of COP in
the anteroposterior direction of the proposed models correspond to the experimental
results for the quiet stance. These results validate the proposed models.

Human postural control during standing posture

11

The selection of the controller required to stabilise the postural balance depends on
your application and the parameters used in the proposed models as shown in Figure 4.
Figure 4

Neural controller parameters for different models of the postural control system

11 Discussion
The proposed model and simulation results indicate that the developed model proved
consistent in several important aspects, such as the generation of torque control required
in relation to the disturbance suffered by the CNS stimulation of the muscle groups
suitable for the torque control required for stabilisation, muscle forces and torques within
the physiological range and data in the literature for the maintenance of upright posture
and fluctuations of the COM and COP, in agreement with those observed experimentally
in the upright position and described in the literature. Such models, although simplified,
contribute to the greater understanding of balance control, since they include various
parameters of the body such as muscles and tendons, previously ignored or drastically
simplified (Peterka, 2000; Peterka and Loughlin, 2004).
One of the advances of the proposed model is to consider the angle of the reference
COM of the human body in relation to the ankle other than 0. The non-vertical
alignment between the COM and ankle joint reference causes changes in the COP signal,
controller parameters and torques involved in the stabilisation of posture. There is also a
significant difference in neural activation signal, with the average value of neural
excitation signal for 0 is zero, providing an alternation between the activity of the
plantar flexor and dorsiflexion groups. This fact does not correspond to the data in the
literature, which show a predominance of activity of the plantar flexor group at the
expense of the dorsiflexors. When the reference value is between 3 and 5, the prevalence
of plantar flexor muscle group is stressed.
However, the way humans control and maintain their posture is under discussion in
the scientific community, and the findings should be considered carefully when
interpreting the biomechanical and electromyography signals. Kohn (2005) highlights the

12

D.B. Coelho and M. Duarte

importance of correct interpretation of the cross-correlation function for the conclusion of


the need for a feedforward control to maintain posture. The models proposed in these
paper exhibit characteristics of feedback and can reproduce the delay between the signal
of electromyography and COP found experimentally in the work of Gatev et al. (1999).
They prove that the apparent simplicity in maintaining the erect posture involves
a complex and dynamic integration of sensory-motor mechanisms. The goal of
sensory-motor integration is the appropriate amount of torque corrective to resist gravity
and external perturbations. All models showed a similar characteristic findings and the
postural sway found in the literature (Figure 3).
An important component in the maintenance of posture, widely discussed in the
works of Winter and Morasso (Morasso and Sanguineti, 2002; Peterka and Loughlin,
2004) is the muscular system. Representative work by Zajac (1989), Winters (1995) and
Menegaldo et al. (2006, 2003) who modelled the muscle-tendon actuator system as
elastic or viscoelastic. Winters proposed a model containing extrafusal and intrafusal
muscle fibres with the contractile structure of the model in series and in parallel with the
elastic tissue lightly damped, and provide a good approximation of the muscle dynamics,
encompassing characteristics of muscle-reflex actuator. However, the model appears to
be relatively sophisticated concerning the dynamic muscle-tendon, opting for a more
simplified one, and can be characterised by a transfer function of second order. We opted
to use a models muscular (Zajac, 1989), as proposed by Menegaldo (2006, 2003). The
introduction of elements and elastic viscous element in series with the contractile allows
the model to cover all operation regimes muscle.
The system identification of muscle-tendon actuators transfer functions for SISO, and
models that faithfully reproduce their behaviour has several advantages. The main
muscle-tendon actuators involved in the maintenance of posture were identified, resulting
in transfer functions of second order without dead time correlating the dynamics of neural
activation and torque at the ankle joint. A model of the actuators responsible for plantar
flexion-agonist-antagonist and the system was performed, showing high correlation
coefficients with the models proposed by Menegaldo.
It is important to emphasise the importance of the stiffness of the muscle-tendon
structure of the human body in the maintenance of quiet upright posture (when we try to
stay as still as possible). This stiffness acts like an elastic against the moment of
gravitational force, which tends to cause the body to fall forward. While the estimated
contribution of the restorative moment of force due to stiffness varies widely in the
literature, it is estimated that this moment of force is of the order 6590% of the
magnitude of the moment of gravitational force (Casadio et al., 2005; Loram and Lakie,
2002a, 2002b). That is, more than half the time of force responsible for keeping us on
foot would be due to a purely passive component, without direct participation of the
nervous system. The equation governing the dynamics of human balance is inherently
unstable as the method of Routh-Hurwitz stability. This means the need of a controller
that adds a pole in the transfer function, a fact which is not obeyed by a pure proportional
controller. This statement refutes the possibility of a control purely by muscle stiffness to
stabilise human posture. Our simulations indicate that the physiological value of stiffness
is insufficient to stabilise. By adding a derivative component, equivalent to the viscosity,
physiologically muscle stability can be achieved with unrealistic values of gain. For
parameters of stiffness and viscosity proposed in the scientific community, stabilisation
occurs at amplitudes that do not match the disturbance suffered by the body.

Human postural control during standing posture

13

Since to maintain posture, a permanent diversion is tolerable, so the proportion could


be used. As excessive oscillations should be eliminated in the sign of projection of the
COG, how derivative had to be added to the controller. We notice however, that the way
was only added to full control when the system had a considerable deviation permanent.
This occurred at values of different reference angle of neutral axis from the vertical.
Considering the reference value of the ankle angle other than 0, the introduction of the
full mode proved to be of fundamental importance because it eliminates the shift is
permanent and efficient processes for very fast and dominated by dead time. As Peterka
and Loughlin (2004) showed the importance of dead time for the system stability, the
integrative mode had to be employed.
We observed that an increase in the value of the proportional gain is introduced as
models of the sensory systems. This is due to the fact that the proportional mode is very
useful for processes of order n with a dominant pole and responds quickly to changes in
both desired value for disturbances in charge.
The application of derivative mode tends to accelerate the response process, which
was observed during on in increase in constant derivative controller. The mode controller
with proportional-derivative is effective for systems of order two or higher. It results in a
quicker and less permanent shift, increasing the stability of the mesh.
The magnitude of passive muscle strength was significant when compared to overall
muscular strength, which is similar to data found in literature (Bottaro et al., 2005).
We notice the existence of a tonic component of torque due to the neural control for a
reference angle of 4, whose value was similar for the different models. The participation
of the torque due to passive muscle properties increases, as models of sensory systems
were introduced. The inclusion of a delay in signal propagation, as here the components
of afferent, efferent and electromechanical delay, presents a breakthrough in achieving a
system that reproduces the postural sway. The issue of complexity and biomechanical
control does not qualitatively change the overall results. Such models are more reliable
when compared to physiological ones, because they include delays due to afferent,
efferent and electromechanical. The introduction of a time delay on the efferent pathways
generates greater instability in the system, to be included on the command line. Despite
this, the values of delays used were those found in literature.
Despite the redundancy of information about the sensory systems, the proposed
model involving the representations of visual, vestibular and somatosensory considered
that the nervous system uses all available sensory information. However, the proposed
model is capable of simulating the contributions and earnings of the sensory systems,
consistent to experimental results (van der Kooij et al., 2001) in maintenance of posture.
In the model, all delayed sensory information is integrated in such a way that a best
estimate of body orientation is obtained. This model approach agrees with the present
theory of the goal of human balance control, i.e., to achieve effective weighting of all
sensory information to maintain a stable vertical and horizontal alignment of the body
with respect to the individuals intent, experience, instruction and environment. The
influence of sensory manipulation on total body sway during quiet stance is studied by
changing the reliability of individual sensory systems. Knowledge of the precision and
gain of the different sensory systems result a minimum variance estimate of spatial
orientation. In a well-lit environment with a firm base of support, healthy persons rely on
somatosensory (70%), vision (10%) and vestibular (20%) information (Peterka, 2002).
Our findings suggest that the maintenance of a vertical orientation of the body in the
upright standing position is under the continuous control of vestibular, visual and

14

D.B. Coelho and M. Duarte

proprioceptive inflow, in that they contribute in establishing a reference system for the
body vertical. As a general organising principle, we suggested a hierarchy of the sensory
feedback signals. Results suggested that in the sway-referenced condition normal subjects
altered their postural strategy by strongly weighting feedback from plantar somatosensory
force sensors.

12 Conclusions
In order to facilitate early diagnosis of disorders affecting the human postural control, the
theoretical modelling coupled with computer simulation emerges as an option used by
researchers. Intending to present a model that represents a breakthrough in the
understanding of human postural control system, this paper reviews the main theories of
postural control, the current models of postural sway and presents a new computational
model dedicated to the study of human upright posture, with the introduction of a model
of muscle-tendon actuator. This model reproduces the responses of oscillations of the
COP and COM and offers the possibility of measuring the neural activation, the
decomposition of reactive torque and the participation of each set of muscle groups to
maintain posture. It reproduces, in a model of feedback, the existing cross-correlation
between experimentally observed and the neural signal and the COP. The findings on the
electrical signal preceding the biomechanics of the projection of COG only are
insufficient to complete the necessity of a feedforward control to control posture, because
this representation can be obtained by a simple model for feedback.

Acknowledgements
D. Coelho is thankful to FAPESP for her scholarship (#05/57437-5).

References
Akaike, H. (1974) A new look at the statistical model identification, IEEE Transactions on
Automatic Control, Vol. 19, No. 6, pp.716723.
Asslander, L. and Peterka, R.J. (2014) Sensory reweighting dynamics in human postural control,
Journal of Neurophysiology, article in press, doi: 10.1152/jn.00669.2013.
Bottaro, A., Casadio, M., Morasso, P.G. and Sanguineti, V. (2005) Body sway during quiet
standing: is it the residual chattering of an intermittent stabilization process?, Human
Movement Science, Vol. 24, No. 4, pp.588615, doi: 10.1016/j.humov.2005.07.006.
Casadio, M., Morasso, P.G. and Sanguineti, V. (2005) Direct measurement of ankle stiffness
during quiet standing: implications for control modelling and clinical application, Gait &
Posture, Vol. 21, No. 4, pp.410424, doi: 10.1016/j.gaitpost.2004.05.005.
Chen, K.J., Keshner, E.A., Peterson, B.W. and Hain, T.C. (2002) Modeling head tracking of visual
targets, Journal of Vestibular Research, Vol. 12, No. 1, pp.2533.
Collins, J.J. and De Luca, C.J. (1993) Open-loop and closed-loop control of posture: a
random-walk analysis of center-of-pressure trajectories, Experimental Brain Research,
Vol. 95, No. 2, pp.308318.

Human postural control during standing posture

15

Conforto, S., Mathieu, P., Schmid, M., Bibbo, D., Florestal, J.R. and DAlessio, T. (2006) How
much can we trust the electromechanical delay estimated by using electromyography?,
Conference Proceedings of the IEEE Engineering in Medicine and Biology Society, Vol. 1,
pp.12561259, doi: 10.1109/IEMBS.2006.259335.
Conforto, S., Schmid, M., Camomilla, V., DAlessio, T. and Cappozzo, A. (2001) Hemodynamics
as a possible internal mechanical disturbance to balance, Gait & Posture, Vol. 14, No. 1,
pp.2835.
de Freitas, P.B., Freitas, S.M., Duarte, M., Latash, M.L. and Zatsiorsky, V.M. (2009) Effects of
joint immobilization on standing balance, Hum Mov Sci., Vol. 28, No. 4, pp.515528, doi:
10.1016/j.humov.2009.02.001.
Duarte, M. and Freitas, S.M. (2005) Speed-accuracy trade-off in voluntary postural movements,
Motor Control, Vol. 9, No. 2, pp.180196.
Fernandez, C. and Goldberg, J.M. (1971) Physiology of peripheral neurons innervating
semicircular canals of the squirrel monkey. II. Response to sinusoidal stimulation and
dynamics of peripheral vestibular system, Journal of Neurophysiology, Vol. 34, No. 4,
pp.661675.
Fukuoka, Y., Tanaka, K., Ishida, A. and Minamitani, H. (1999) Characteristics of visual feedback
in postural control during standing, IEEE Transactions on Rehabilitation Engineering, Vol. 7,
No. 4, pp.427434.
Gatev, P., Thomas, S., Kepple, T. and Hallett, M. (1999) Feedforward ankle strategy of balance
during quiet stance in adults, The Journal of Physiology, Vol. 514, No. Pt 3, pp.915928.
Gurfinkel, V.S., Ivanenko Yu, P., Levik, Yu, S. and Babakova, I.A. (1995) Kinesthetic reference
for human orthograde posture, Neuroscience, Vol. 68, No. 1, pp.229243.
Gurfinkel, V.S., Ivanenko, Y.P. and Levik, Y.S. (1995) The influence of head rotation on human
upright posture during balanced bilateral vibration, Neuroreport, Vol. 7, No. 1, pp.137140.
Hopkins, J.T., Feland, J. B. and Hunter, I. (2007) A comparison of voluntary and involuntary
measures of electromechanical delay, The International Journal of Neuroscience, Vol. 117,
No. 5, pp.597604, doi: 10.1080/00207450600773764.
Horstmann, G.A. and Dietz, V. (1990) A basic posture control mechanism: the stabilization of the
centre of gravity, Electroencephalography and clinical neurophysiology, Vol. 76, No. 2,
pp.165176.
Jacono, M., Casadio, M., Morasso, P.G. and Sanguineti, V. (2004) The sway-density curve and the
underlying postural stabilization process, Motor Control, Vol. 8, No. 3, pp.292311.
Kohn, A.F. (2005) Cross-correlation between EMG and center of gravity during quiet stance:
theory and simulations, Biological Cybernetics, Vol. 93, No. 5, pp.382388, doi:
10.1007/s00422-005-0016-x.
Lafond, D., Duarte, M. and Prince, F. (2004) Comparison of three methods to estimate the center
of mass during balance assessment, Journal of Biomechanics, Vol. 37, No. 9, pp.14211426,
doi: 10.1016/S0021-9290(03)00251-3S0021929003002513 [pii].
Loram, I.D. and Lakie, M. (2002a) Direct measurement of human ankle stiffness during quiet
standing: the intrinsic mechanical stiffness is insufficient for stability, The Journal of
Physiology, Vol. 545, No. Pt 3, pp.10411053.
Loram, I.D. and Lakie, M. (2002b) Human balancing of an inverted pendulum: position control by
small, ballistic-like, throw and catch movements, The Journal of Physiology, Vol. 540,
No. Pt 3, pp.11111124.
Menegaldo, L.L., de Toledo Fleury, A. and Weber, H.I. (2006) A cheap optimal control
approach to estimate muscle forces in musculoskeletal systems, Journal of Biomechanics,
Vol. 39, No. 10, pp.17871795, doi: 10.1016/j.jbiomech.2005.05.029.
Menegaldo, L.L., Fleury Ade, T. and Weber, H.I. (2003) Biomechanical modeling and optimal
control of human posture, Journal of Biomechanics, Vol. 36, No. 11, pp.17011712.
Morasso, P.G. and Sanguineti, V. (2002) Ankle muscle stiffness alone cannot stabilize balance
during quiet standing, Journal of Neurophysiology, Vol. 88, No. 4, pp.21572162.

16

D.B. Coelho and M. Duarte

Morasso, P.G. and Schieppati, M. (1999) Can muscle stiffness alone stabilize upright standing?,
Journal of Neurophysiology, Vol. 82, No. 3, pp.16221626.
Morasso, P.G., Baratto, L., Capra, R. and Spada, G. (1999) Internal models in the control of
posture, Neural Networks, Vol. 12, Nos. 78, pp.11731180.
Mrachacz-Kersting, N., Fong, M., Murphy, B. A. and0 Sinkjaer, T. (2007) Changes in excitability
of the cortical projections to the human tibialis anterior after paired associative stimulation,
Journal of Neurophysiology, Vol. 97, No. 3, pp.19511958, doi: 10.1152/jn.01176.2006.
Pearson, R.K. and Pottmann, M. (2000) Gray-box identification of block-oriented nonlinear
models, Journal of Process Control, Vol. 10, pp.301315.
Peterka, R.J. (2000) Postural control model interpretation of stabilogram diffusion analysis,
Biological Cybernetics, Vol. 82, No. 4, pp.335343.
Peterka, R.J. (2002) Sensorimotor integration in human postural control, Journal of
Neurophysiology, Vol. 88, No. 3, pp.10971118.
Peterka, R.J. and Loughlin, P.J. (2004) Dynamic regulation of sensorimotor integration in human
postural control, Journal of Neurophysiology, Vol. 91, No. 1, pp.410423, doi: 10.1152/jn.
00516.2003.
Schieppati, M. and Nardone, A. (1997) Medium-latency stretch reflexes of foot and leg muscles
analysed by cooling the lower limb in standing humans, The Journal of Physiology, Vol. 503,
No. Pt 3, pp.691698.
van der Kooij, H., Jacobs, R., Koopman, B. and Grootenboer, H. (1999) A multisensory
integration model of human stance control, Biological Cybernetics, Vol. 80, No. 5,
pp.299308.
van der Kooij, H., Jacobs, R., Koopman, B. and van der Helm, F. (2001) An adaptive model of
sensory integration in a dynamic environment applied to human stance control, Biological
Cybernetics, Vol. 84, No. 2, pp.103115.
Winter, D.A., Patla, A.E., Prince, F., Ishac, M. and Gielo-Perczak, K. (1998) Stiffness control of
balance in quiet standing, Journal of Neurophysiology, Vol. 80, No. 3, pp.12111221.
Winter, D.A., Patla, A.E., Rietdyk, S. and Ishac, M.G. (2001) Ankle muscle stiffness in the control
of balance during quiet standing, Journal of Neurophysiology, Vol. 85, No. 6, pp.26302633.
Winters, J.M. (1995) An improved muscle-reflex actuator for use in large-scale
neuro-musculoskeletal models, Annals of biomedical Engineering, Vol. 23, No. 4,
pp.359374.
Zajac, F.E. (1989) Muscle and tendon: properties, models, scaling, and application to
biomechanics and motor control, Critical Reviews in Biomedical Engineering, Vol. 17, No. 4,
pp.359411.
Ziegler, J.G. and Nichols, N.B. (1942) Optimum settings for automatic controllers, Transactions
of the ASME, Vol. 64, No. 11, pp.759765.
Zupan, L.H., Merfeld, D.M. and Darlot, C. (2002) Using sensory weighting to model the influence
of canal, otolith and visual cues on spatial orientation and eye movements, Biological
Cybernetics, Vol. 86, No. 3, pp.209230.

Anda mungkin juga menyukai