Sunteți pe pagina 1din 6

2018 International Conference on Computing, Mathematics and Engineering Technologies – iCoMET 2018

Control Synthesis of Single link Biomechanical


Model using Feedback Linearization

Nadia Imran, A.M.Mughal, M.Najam ul Islam


Electrical Engineering Dept. Bahria University Islamabad, Pakistan
nadia.imran@bui.edu.pk, amahmood1109@gmail.com , najam@bahria.edu.pk

Abstract— Postural stability is the capability of an individual Researchers found that both active and passive mechanisms
to maintain the position of a human body in upright posture, more are involved in postural equilibrium. When the CNS receives
specifically to balance its center of mass (COM) within precise commands from brain or reflex loop; active torque is generated
limitations of physiological space and stability limits. In this through muscle contraction [5].
paper, we study an application of nonlinear control of a single link
biomechanical model relating to postural stability. We construct To understand and improve body movement mechanism
feedback compensator to regulate and estimate the movement of motion analysis is done using mathematical models of human
biomechanical model for postural stability and fall prevention. musculoskeletal assembly. The researchers have used the
The compensator design utilizes the strength of feedback inverted-pendulum models for analysis of bio-mechanical
linearization method to regulate the 2nd order system equations. movements because of its resemblance with human body and its
Moreover, feedback estimator mimics a nonlinear estimation of inherent instability.
central-nervous-system (CNS) to investigate the stability of
biomechanical movements. We simulate this scheme in The authors in [6] studied the single link bio-mechanical
MATLAB/Simulink environment and observe that the model with PID controller. They developed the physiological
compensator with optimal gains resulted in optimized torques in controller with short latencies in position, velocity and force
the presence of noisy measurements from joint sensor. feedbacks. Researchers in [7] determined the balanced stability
boundaries calculated from COM position and COM velocity
Keywords— Postural stability; Feedback Linearization; for various times to ankle joint torques. The authors in [8]
Biomechanical Movements. developed the bond graph model of the muscular skeletal system
analyzing the designed system with H2 and H∞ compensators
after adding uncertainties and white noise disturbance at sensory
I. INTRODUCTION feedback loops. They used 3-link, four-segment model to study
the function of knee-joint in maintaining balance in upright
The study and analysis of biomechanical movements is of
posture.
great interest to researchers from various disciplines. The human
upright posture has been studied comprehensively in clinical and A common approach for the control of nonlinear systems is
theoretical settings. to use linear control technique for linear approximation of the
nonlinear system at an operating point. This technique is
Postural stability plays a significant role in the management
generally used because the choice for linear control technique is
of upright posture as well as retaining balance or equilibrium
wider. However, this approach generally works in a small
while performing different actions and voluntary movements.
neighborhood of operating region where the linear
Postural stability is the capability of an individual to maintain
approximation is valid. The feedback linearization approach is
the position of a body in upright posture or specifically its center
based on transformation of nonlinear system into an equivalent
of mass (COM); within precise limitations of space which are
linear one, which is valid for all functional regions [9]. This
stated as stability limits. A person is said to be in stable posture
method cancels the nonlinearities in the system and the resulting
as long as the person’s weight vector passes through the base-
linear system is then combined with linear control technique for
of-support (BOS). Some individuals are not able to maintain
effective control strategy [10, 11].
stability even though this condition is met [1]. Various motor
control mechanisms and information is required for better This paper uses the nonlinear plant model developed in [4].
control of postural movements [2]. Number of studies have The feedback linearization technique has been used to provide
shown that visual system enhances the postural balance [3]. It better control strategy. We synthesized the design of controller
has been observed by researchers that the upright posture is and observer based on vector filed method combined with linear
maintained in the presence of both the internal perturbations quadratic regulator and linear quadratic estimator for optimal
(breathing and hand-limb movements) and external gains. The simulation results for angular position, velocity, and
perturbations (kick, push) to the BOS; by constant input torque joint input torque verify the design scheme.
provided by anti-gravity muscles in lower-leg [4].

978-1-5386-1370-2/18/$31.00 ©2018 IEEE


The joint-angular position and the angular-velocity are
state variables. The input torque τ constitutes the muscle force
II. MATHEMATICAL BACKGROUND
which then translates to joint torque. The joint position and
A. Nonlinear plant model angular velocity which are the outputs of this system; translate
into fascicle-length and fascicle velocity as inputs to the
In our research we have considered the human body as a muscle-spindle assembly.
single segment with ankle joint and static base of support (BOS)
shown in (Fig.1). This segment can rotate about the ankle-joint. B. Feedback Linearization
The forces which are acted upon by this system are the force of
gravity and the net torque produced by the MTCs (Muscle- We have used the feedback linearization method to construct
Tendon Complexes) around the ankle-joint. feedback compensator. The control provided by this scheme
comprises of two parts. The first part involves the smooth state
In our model, m represents the mass of body, I is the inertia of space transformation z=T(x) that eliminates the plant’s
body, k is the length of segment from COM to a joint, is the nonlinearities and the other part in combination with linear
joint angle which represents the position of body with respect to control technique controls the subsequent linear system [10].
ground and is the joint input torque. The nonlinear inverted
pendulum system is represented by the following equation: The controllability matrix of the nonlinear state space can be
calculated by taking the derivative of vector field with respect
= (1)
to vector field (lie bracket).

0
[(ad0 f, j) (ad1 f, j)] = (3)
0

The above controllability matrix is full rank. Now to


construct the state transformation we take the last row of
inverse of controllability matrix to calculate the transformation
matrix. Afterwards the lie derivative can be calculated as:

( ) ( + )
= (4)
( ( )) ( + )

The transformation z=T(x) is calculated such that x=T-1(z)


Fig.1. Single link Biomechanical Model and T(0)=0.

Nonlinear plant model can be written in state space form as:


0 T-1(z) = (5)
= + τ (2)

The values of biomechanical body mass m, the length of


y= [1 0] x segment k from COM and Inertia I of body have been taken
This nonlinear plant is of the form from ref [6] given in Table.1. The transformation can be written
in state space form as:
= f(x) + J(x) u
y = h(x) 0
= 573.9109 0.0207 + (6)
1
We assume that
f (0) = 0 and J (0) = 0 In the next section we will combine the transformed system
with linear quadratic regulator to construct control for the single
Where the function f: Rn →Rn link model based on feedback linearization method.
The input function J: Rn→Rn× m
Where J = [J1(x) J2(x) ----- Jm (x)] and C. Controller Design
The nonlinear system defined in eq.1 has been linearized at
Ji: Rn→Rn i= (1, 2, -------n)
standing upright posture i.e. = 90° with ground as a frame of
The output map h: Rn→Rp reference. The equilibrium points for standing (upright) posture
and without any movement are = 0 and = 0. The
linearized model is given by:
If f can be defined overall Ck, where k=1, 2, 3-------∞ then
we can write f belongs to . The functions f, Ji and h are
assumed to be vector fields on Rn.
∆ 0 1 ∆ 0 transformation matrix we pick the last column of inverse of
= 0 ∆ + ∆ (7) observability matrix /

This linearized model is an unstable system. We can control


the body movement through the feedback of states. The state = [ (ad0 f , ) (ad1f , )] (14)
feedback control the input torque which is further responsible
for stable body movement. For optimal controller design we The nonlinear state space equation is obtained in z-coordinates
calculated the gains through linear Quadratic regulator. The using inverse transformation as follow
Riccati equation for this design problem is given by:

MA+AT M- MBR-1BTM+Q=0 (8) = f(x) | x=T(z) + g(x) | x=T(z)


u = Az+ (y,u)
Where each term represents a matrix. The input-weighting y= Cz
matrix R and the state-weighting matrix Q are chosen such as
to minimize the performance index. The controller gain K is 0 +
calculated by: = + (15)
0
K = R-1 BTM (9)
The linearized model obtained in equation (7) is subjected to
500 0
Where Q = LQE for calculating the optimal gain L for observer design. We
0 500
R = 0.01 have chosen the values of Q and R in such a way as to make the
poles of estimator faster than controller.
The gain K obtained from above equation gets optimal The state estimator in z- coordinates has the form
solution for the steady state response, initial and final position
of the body, and minimizes the error in states and control input = (A-LC) ̌ +Ly + (y, u) (16)
τ. The stabilizing state feedback Controller using feedback
linearization method in z coordinates is defined by the We get the state estimator for the nonlinear system defined by
following equation eq.1

U=-f (z1, z2)-k1z1-k2z2 (10) 0 −25.4575 26.4575 11.88


= ̌+ y+ +
1 −27.4393 27.4393 0
0.0207
The controlled input torque resulting from LQR is used as (17)
0
an input to the nonlinear model which regulates the body y= z2
movement. Both the Q and R play a very important role in
optimal controller design. Using the inverse transformation the state estimator can be
represented in original coordinates as follow
Solving the equation (10) in z coordinates yields the controller
equation in the following form 0 1 27.4393 0 0
= - (y- )+ +
0 0 26.4575 11.88 0.0207
U= -570.596 sin (0.0207z1)-24.15z1- 7.63z1 (11)
(18)
To obtain the above equation in original coordinates we apply
inverse transformation E. Compensator Design
In the control law implementation we used the state
U= -570.596sin(x1) – 1167x1 – 368.9x2. (12) estimates rather than the states themselves. With combined
estimator-controller in a loop the plant’s nonlinearities may
cause difficulties in stabilizing the system, however the strength
D. Observer Design
of the vector field method alters the output of the system in a
In this section we design full state observer which gives noticeable way. The initial conditions of the estimator were set
estimation of the position angle and angular velocity of the to zero. The estimator gives the estimation of states in similar
biomechanical model. The state estimator impersonate the way as the commands generated by CNS. The combined
states profiles generated by CNS for stable body movement. We controller–estimator, compensator is defined by equation (12)
will again consider the nonlinear plant defined by equation (1). and (18). To incorporate the state estimation effects on the
Using feedback linearization technique the observability matrix stability of closed loop system the controller in equation (12) is
is calculated as defined as:

( ℎ) 0 1 U= -570.596 sin ( 1) – 1167 1 – 368.9 2. (19)


= [ad0 f ad1f] = (13)
( ℎ) 1 0
The states of controller 1 and 2 are the estimated states scaled
The observability matrix is full rank which shows that the by optimal control gains which generates the controlled input
system is completely observable. To calculate the signal U.
Fig.2 Simulation scheme for postural control of a single link model using Feedback Linearization

III. SIMULATIONS AND RESULTS

The Nonlinear Biomechanical model (Fig.1) is an


unstable system. We synthesized the full order compensator
based on feedback linearization for the nonlinear model. The
scheme has been combined with LQR/LQE for Optimal
gains. For simulations of the model we used the values
defined in [6]. We implemented the controller and estimator
defined in equation (18) and (19). The MATLAB simulation
scheme constructed in Simulink is shown in Fig.2. The
responses are observed to be completely stable.
Fig.3 shows the phase portraits of angular position and
angular velocity in the phase plane for nonlinear plant
without feedback linearization control. These trajectories are
plotted for different initial conditions of angular position of Fig. 3. Phase portraits of uncontrolled plant
the body; starting from inner most with initial conditions of
0.1, 0.2, 0.3 and the outermost for 0.4 radians. The phase
portraits reveal the existence of closed trajectory in the phase
plane which concludes a stable limit cycle since all the
neighboring trajectories are attracted to it. The stable limit
cycle shows that this biological system exhibits self-
sustained oscillations, which are controlled through feedback
controller design.
Fig.4 shows the phase portraits of the controlled plant.
These trajectories have been plotted for initial conditions of
0.1, 0.2, 0.3 and 0.4 radians. It is clear from the phase
portraits that for different initial angular positions the
trajectories are attracted to zero. The controlled plant is now
asymptotically stable. The body being perturbed comes back
Fig. 4. Phase portraits of controlled plant.
to stable upright position. The angular velocity and position
both settles to zero with controlled input.
Fig.5. Angular position error profile Fig.7. Input Joint Torque profile

ankle joint. Our Controlling scheme intend to settle down the


plant much faster as compared to the previous study [12-13],
where the simulation results settles the single link model in 6s.
Fig.6 shows the simulation results for angular velocity of
the controlled plant. In this case we get maximum overshoot
of 0.27 rad/sec. This response settles down to zero in steady
state. Including nonzero initial velocity means incorporating
reaction forces into the model.

Fig.7 shows the input torque profile of the controlled plant.


The body in upright posture is given perturbation of 0.1 rad.
The estimator output gives estimation of states in such a way
as the commands generated by central nervous system.
The Muscle-Tendon-Complexes (MTC) generates torque of
180 Nm to resumes the upright posture of body in 3sec. The
torque profile settles to zero within 2sec.

Fig.6. Angular velocity error profile


IV. CONCLUSION
Fig.5, 6 and 7 shows the simulation results for angular
position, velocity and joint torque of controlled plant. The Postural stability analysis through feedback linearization
postural reaction is initiated in response to external or internal provides new perspective to the subject study. The
perturbation to the BOS. During postural regulation the COM combination of this vector field method with LQR/LQE
is located behind the heel at the start of the movement. This increases the strength of compensator by penalizing inputs.
stance represents a person leaning with the wall statically i.e. In this study the feedback-linearization with optimal gains
zero joint velocity. The upright posture is the final position at provides better approach for postural stability as compared to
which the movement terminates. Thus initial and final simple linear design schemes presented in literature. This
velocities are zero in our simulations. Our simulations intend approach is also effective in the presence of neural delays
to position the COM over BOS at the movement termination. with almost similar effects [10].
It is obvious from the position response in Fig.5 that in steady This scheme may be adopted for study and analysis of
state this profile settles to zero. The simulation results are for different voluntary movements, such as sit-to-stand task. The
initial perturbation of 0.1 rad. The body comes back to stable sit-to-stand (STS) model which are previously studied with
upright position within 3 sec. with COM positioned over the
optimal controller design and fuzzy modeling can be better [8] Asif M. Mughal and Kamran Iqbal, “Bond graph modeling of
understood with this nonlinear vector field method. physiological motor control with H∞ controller design”, IEEE
We will extend our model to three-link biomechanical International Conf. of Control Applications”, pp947-952,
model and design controller for sit-to-stand movement being Munich, Germany, Oct. 4-6, 2006.
modelled as three inverted pendulums. We will also introduce [9] Zak, Stanislaw H. "Systems and control." chapter 7, pp367-
392, (2003).
sensory-delays at joint-angles and delays at all joint-torques. [10] N. Imran, Asif M. Mughal. "Postural stability of a single
The study can be further extended to construct the control for link biomechanical model using feedback linearization",
bio-mechanical model in the presence of reference- ICAMS NUST, 2011.
trajectories and feed-forward-torques. [11] Ur Rehman, Obaid, Ian R. Petersen, and Baris Fidan.
"Feedback linearization-based robust nonlinear control design
for hypersonic flight vehicles." Proceedings of the Institution
of Mechanical Engineers, Part I: Journal of Systems and
Control Engineering 227.1 (2013): 3-11
APPENDIX [12] Z. Madiha, and A. Mahmood. Bond Graph Modeling and PID
Controller Stabilization of Single Link Biomechanical
Model. International Conference on Modeling and Simulation,
Table.1 Islamabad, Pakistan, 25–27 November, 2013.
[13] Mughal A.M. (2016) Recent Applications of Bond Graph
Parametric values of Biomechanical elements Modeling. In: Real Time Modeling, Simulation and Control of
Dynamical Systems. Springer, Cham.
No Biomechanical model terms
Mass of Human body m 75Kg
1
2.633Nm2
Inertia I
2 /Kg2
Distance from the
k 0.78m
3 center of mass to joint
Acceleration due to
g 9.8m/s2
4 gravity

REFERENCES
[1] Napoli, Alessandro, et al. "Automated assessment of postural
stability system." Engineering in Medicine and Biology Society
(EMBC), 2016 IEEE 38th Annual International Conference of
the. IEEE, 2016.
[2] Mughal, Asif M., and Kamran Iqbal. "Fuzzy reduced order
observer-controller design for biomechanical sit-to-stand
movement." Systems, Man, and Cybernetics (SMC), 2016
IEEE International Conference on. IEEE, 2016.
[3] Mughal, Asif M., and Kamran Iqbal. "Bond Graph Modelling
and Optimal Controller Design for Physiological Motor
Control System." International Journal of Modelling and
Simulation 33.2 (2013): 93-101.
[4] Roy, A. Iqbal, k. "Optimization of an inverted Pendulum based
Biomechanical Model with Multiple Feedback Delays."
International IASTED Conference on Intelligent Systems and
Control Applications pp362-367, Marina Del Rey, CA, USA,
March 1-3, 2004.
[5] Iqbal, Kamran, and Asif M. Mughal. "Active control vs.
passive stiffness in posture and movement
coordination." Systems, Man and Cybernetics, 2007. ISIC.
IEEE International Conference on. IEEE, 2007.
[6] Iqbal, Kamran, and Anindo Roy. "Stabilizing PID controllers
for a single-link biomechanical model with position, velocity,
and force feedback." Transactions of the ASME-K-Journal of
Biomechanical Engineering 126.6 (2004): 838-843.
[7] Mughal, Asif M., and Kamran Iqbal. "A fuzzy biomechanical
model with H2 control system for sit-to-stand movement."
American Control Conference, 2006. IEEE, 2006.

S-ar putea să vă placă și