Proceedings of the 23rd International

Conference on Automation &

Computing, University of Huddersfield,
Huddersfield, UK

PID Parameters Optimization

Using Ant-colony Algorithm
for Human Heart Control
M.Aabid1, A.Elakkary2, N.Sefiani3,
1.2.3
LASTIMI Laboratory: "Laboratory of Systems Analysis, Information Processing and Integrated Management"
1.2.3
Mohammadia Engineering School Rabat, Mohammed V University
Rabat, Morocco
1
[email protected], [email protected], [email protected]

Abstract – This article aims to optimize metaheuristics for before proceeding to the animal experimentation stage.
continuous variable problems often encountered in the The objective of this article is to apply optimization
engineering field and its application in the medical field for methodology to obtain a better mathematical model as
the resolution of heart disorders. The work, based on the well as a simple, compact and expensive physical
model of Yanagihara, Noma, and Irisawa (YNI) consists of
analogue model of the total artificial heart.
understanding and isolating the human cardiovascular
system using a hydro-electromechanical (HEM) approach, II. HYDRO-ELECTROMECHANICAL HUMAN
we also include the methodology of optimization by colonies
CARDIOVASCULAR DESIGN
of ants (ACO) in order to control the heart. Then, different
approaches to metaheuristic design are used following this The heart measure, about 13 cm long, 8 cm wide, weighs
perspective and the algorithms developed by MATLAB code about 500g and beats 3 billion times in a lifetime.This
are applied to a problem of regulation and control of the function is provided by the nodal tissue figure 1 which is
human heart perturbed by brain problems. For the desired called the natural pacemaker.
electrical SA node, which means the desired flow by
adapting the contraction of the ventricle and the atrium, the
PID coefficients (Kp, Ki, Kd) are tuned using ACO by
MATLAB code by acting on the hydro- electromechanic
model In SIMULINK controlled by PID after applying the
disturbance. The results of this experiment really present a
good response after the application of perturbation to the
cardiac system compared to our previous published articles
which rely on the methodology of Ziegler–Nichols for the
revelation of the parameters Proportional–Integral–
Derivative for the PID controller.
Figure 1: The nodal tissue for Normal Heart
Keywords: Ant Colony Optimization, Human Heart, Control
and Regulation, Pacemaker, Sympathetic Nervous System,
Several linear and non linear model are developed
Brain, HEM
[5][15][16] most of those models are based on YNI
I. INTRODUCTION model [21] others design a mathematical model for the
artificial generation of signals electrocardiogram
The heart is the central organ for the cardiovascular
(ECG)[17].
system. So since the 1960s, much work has been done in
this field using physical models of the real system.
Computer analyses as well as mathematical models have
been widely used in the simulation of cardiovascular
systems. The Windkessel model is a famous example of
such a discrete model.Although computational analyses
of numerical models have taken the place of physical
models in many cases, physical electrical models are still
indispensable to test the reliability and control of Figure 2: Normal Heart Beat from ECG
artificial hearts. In order to obtain accurate results, the
performance and functions of the artificial heart must be This work is based on the hydro-electromechanical model
evaluated using physical hydro-electromechanical odels of the human heart [5][3] which is the basis of several
works of the mathematical and physical models described End Repeat T + +
in scientific works [8] [11] [21]. For this reason, the new Repeat on the N individuals of the population
model based distinguishes between large arterial vessels Calculate ai Repeat on d dimensions
and peripheral arterioles, and includes the inertial effects of Vij = _ (vij + _1_1 (pij - xij) + _1_2 (pgj - xij))
blood. And an excitation pulse is sent to the node SA each Vij = vij + aij, Xij = xij + vij
0.85 seconds. After a mathematical analysis of the End
electrical circuit they obtained the transfer function below If f (xi) <f (pi) then pi = xi , If f (pi) <f (pg) then g = i
for the core: End
( )= End if stop criterion
( + )+ . The main idea is a direct transfer of all the solutions found
at one iteration to the next iteration and then the
III. ANT COLONY OPTIMIZATION ALGORITHM
calculation of the amount of pheromone needed for the
A. Elaboration of the ACO algorithm next iteration. In traditional ant colonization algorithms, it
Among this method of optimization Fig 4, the algorithms is the pheromone that is transferred directly from one
of ant colonies form a class of recently discovered iteration to another. Then the authors propose in [26] a
problems for difficult optimization. These algorithms are modification of how the pheromone changes, so as to
based on collective deposition and tracking behavior allow a longer trace of the good solutions for a longer
observed in ant colonies. time, then explicitly eliminate their influence of the
pheromone matrix. The three previous strategies are
Initialize modified in [27], introducing an elitist notion: The best
ants are the only ones allowed to change the pheromone at
each iteration, and when a change is detected, the old
For each iteration (i=1 to Ni)
good solutions are not forgotten, but modified to the best,
so that they can become new reasonable solutions and it
For each Ant (A=1 to NA) gives us The best BEST_CASE case.D. Merkle and M.
Middendorf study the dynamics of ACO (Ant Colony
Optimization) in [28] and then propose a deterministic
Move Ant to randomly selected
node model, based on an averaging of the expected behavior of
the ants. The new equations proposed are:
Calculate the cost [τ ij ]α [η ij ]β
pij =
 [τ
h∈s
ij ]α [η ij ]β
Update local pheromone trail
The pheromone is indicated by T and the distance
between cities is η of the blow J it is elements of the
iteration for the solutions box and q is a variable between
No Is Last [0, 1].
Ant?
B. Tuning of the PID controller by the ACO algorithm
Yes Two approaches are possible to design optimization
Get best solution from above heuristics Continues Fig 5 following this methodology of
iterations
ACO . The first is to create a multi-agent system where
communication plays a central role, as a process for the
Update global pheromone emergence of a coherent global behaviour of the system.
The second approach describes ant colonization
algorithms as methods that manipulate a sampling of a
Is Last No probability distribution for detection of the best Kp ,Ki
Iteration and Kd parameters for tuning our PID in real time and
also when the PID is used to regulate the heart disturbed
Yes
by external noises like Disease , Anxiety or fight….
Get best solution

Figure 3: Functional Scheme for ACO Algorithm

The algorithm use for the optimization in our case to

construct a optimal PID parameters for Heart control in
disturbed case is is :
Initialize S = {xi, vi} in a cube [-X, X] d G, t = 0 Repeat Figure 4: Optimization heuristic to for PID parameters Tuning under
on the N individuals of the population with Heart under disturbances
Pi = xi If f (pi) <f (pg) then g = i
This aloghorithm is done by Matlab but the design is Scientists have estimated that the normal adult human
elaborated on Simulink so the Run of Matlab code of brain is made up of one hundred billion neurons and
the Optimization select the best PID parameters Kp, Ki, With each neuron to be connected thousands of other
Kd and affect them to PID on the simulink design in neurons the brain uses parallel distributed processing
real time. where treatment occurs simultaneously through billions
of neurons distributed in the brain [12][13]. The
brainwaves if its overdosed is disturbance and
controller in the same time in for the human heart Fig 9
[10] Brainwaves contain four groups known as Delta,
Theta, Alpha and Beta frequency bands [13]. Delta is
with the highest amplitude with the lowest frequency
Figure 5 : PID parameters assignation by the ACO code in MATLAB band and Beta is the lowest amplitude with the highest
frequency band.
IV. ANALYSE OF CARDIOVASCULAR SYSTEM
WITHOUT NOISE TABLE II. VALUES OF THE PD PARAMTERS

The basis of our working model [3] using MATLAB / CATEGORIE OF BRAIN WAVE TAKED LIKE DISTURBANCE
Simulink simulation. Which is based on four BETA (14-100 Hz) Arousal, Anxiety ,disease, fight
mathematical components Fig 8, the main elements are
ALPHA (8-13.9 Hz) Pre-sleep, pre-waking, drowsiness
pacemaker, the heart, the sensor and the circulatory
system.The sino-atrial node containing the natural THETA (4-7.9 Hz) Deep mediation, Hypnagogic
simulator is the basis of the action producing the DELTA (1-3.9 Hz) Loss of body awareness, Dreamless
rhythmic action of the heart by pulses propagated along
a node, this action will be propagated to the ventricle by The frequency of the brain is used in this article like a
the myocardium The ECG signal is characterized by a disease which means a disturbance in the reality of
recurring sequence of three waves, namely:P-wave : electrical activity of cardiovascular system, so we try to
0.07 sec, QT-interval : 0.375 sec, PR-interval : 0.20, control this disturbed human rate heart to mimic the
QRS-complex : 0.139 sec normal heart activity [10].
The parameter used for the generator on this
B. Application of ACO to the Disturbed human
simulation are :
cardiovascular system
TABLE I. LOW THROUGH PARAMETERS OF GENERATOR
The purpose of this part is the synthesis of the PID-ACO
Type of pulse Temp-Base controller. The optimal parameters (Kp, Ki, Kd) are
Amplitude 1 mV
determined according to the performance of the
Period 2 sec
Pulse Width(% of period) 1 sec cardiovascular system (cancelling the static error of the
Phase delay 0-1 unit heart signal in real time of the stroke and also reducing the
exceedance, the response time and the rise time) Of the
PID controller Figure 11 using the Ant Colony
V. APPLICATION OF ACO TO THE Optimization (ACO) method to obtain an accurate, rapid,
CARDIOVASCULAR SYSTEM DISTURBED BY stable cardiovascular system and robust.
THE BRAIN

A. Brain System Disturbance

The brain is an organ that plays the role of a controller Output

of emotion, thought, movement and several functions in Heart
the human body. ∑ PID
REF -

Circulatory Barorecptor
System

Figure 7 Interaction between the Brain and the Human Heart in the
Real case
Proportional action Kp improves system time and
reduces steady-state error. Despite full control, by
Figure 6 Interaction between the Brain and the Human Heart in the reducing the steady state error, it can generate a bad
Real case transient response to the heart controlled by the PID /
ACO.The derivative action Kd will have the effect of PR-interval (mv) 0.1 0.02
increasing the damping in the system, reducing the
overrun and improving the transient response.The gain The PID controller with parameters P, I and D optimized
KP, Ki and Kd of the PID controller are generated by by Ant Colony Optimization (ACO) has been applied to
the ACO algorithm for a given installation. The the perturbed heart Fig 12 and has lost its normal
population is represented by a matrix (n x 3), where the functionality, which makes it possible to adapt the
ant selects the optimum parameters Kp, Ki and Kd of perturbed cardiac system to its normal work. It is observed
the PID control system by minimizing the objective that the perturbed cardiovascular system mimics the
function and the simulink model of disturbed and cardiovascular reference Fig system when its controlled
controlled cardiovascular système is like in figure below by ACO/PID controlled.

Figure 8 Simulink presentation of cardiovasculare Système disturbed

by Brain waves and controlled by ACO algorithm by Matlab
Fig 10 result of cardiovascular System disturbed by Brain waves and
TABLE III :. FLOW THROUGH PARAMETERS OF THE controlled by ACO algorithm
ISTURBANCE GENERATOR
To test the robustness of our algorithm we apply several
Value
cases of perturbations generated by the brain to model the
Wave Form random
errors. In addition several Gaussians of different cases
Amplitude 50 were applied as follows :Case with brain waves DELTA
Frequancy 90 Hz Loss of body awareness, Dreamless is demonstrated like
below :
After this disturbance the heart lose his functional
function and the whole cardiovascular system is working Wave Mean
Variance Frequancy
out of the normal Form value:
Value Gaussian 0 5 1.2 Hz
TABLE V :PARAMETERS OF THE DISTURBANCEGENERATOR
TABLE VI :PARAMETERS OF ACO/PID
Kp Ki Kd
Value 5.8144 12.8137 19.2181
Fig 9 Interaction between the Brain and the Human Heart in the Real
case
After perturbation the heart loses these functionality Fig
13 in such a way that the response both to a large margin
of amplitude 50mv / -50mv which causes the system to
oscillate so that the QRS 0.08 Sec value and amplitude is
between 0.6 And - 0.5 when the normal cardiovascular
system is about 0 so we conclude that the heart is really
disrupted "as a real heart attack or anxiety and the Yp -
Ym = error (e) = 0.2.

TABLE IV. COMPARISON BETWEEN PERTURBATED

AND CONTROLLED HEART Figure 11 cardiovascular System disturbed by Brain waves Loss of
Heart/Parameters Perturbated Control body awareness and controlled by ACO/PID
led
QRS(Sec) 0.08 0.139 Case with brain waves THETA of Hypnologic is demonstrated
QRS(mv) 0.6/-0.5 0.8/-0.2 like below :
P-wave(Sec) Notexisted 0.07
TABLE VII. PARAMETERS OF THE DISTURBANCE GENERATOR
QT-interval(Sec) Variable 0.375
PR-interval (Sec) 0.03 0.2
 Wave Mean Fig 14 cardiovascular System disturbed by Brain waves of Arousal and
Variance Frequancy controlled by ACO/PID
Form value:
Value Gaussian 0 5 7 Hz
A comparison has been made Fig19 with the four
perturbations generated by the brain, which yields
scientific results so that ACO / PID is so robust for most
cases of disturbances between 0-96 HZ in another way To
eliminate most of the natural disturbance generated by the
dysfunction of the nervous system

Figure 12 cardiovascular System disturbed by Brain waves Hypnologic

and controlled by ACO/PID
Figure 15 Comparison cardiovascular System disturbed by Brain waves
Case with brain waves ALPHA of Pre-sleep is demonstrated of 4 cases of disturbances and controlled by ACO/PID
like below :
A comparison with our previous work based on ziegler-
TABLE VIII . PARAMETERS OF THE DISTURBANCE nichols [3] was carried out which gives that the ACO /
GENERATOR
Wave Mean
PID stabilizes the cardiovascular system better than PID
Variance Frequancy tunned by Ziggeler-Nicols because it has a margin of
Form value:
Value Gaussian 0 5 13.9 Hz Amplitude error of 0.03mv and do not delete the
frequency disturbance because the heart still suffers a
frequency disruption of 33.34 Hz
TABLE X
PARAMETERS OF ACO/PID and ziegler-nichols
Kp Ki Kd
ACO / PID (MSE) 5.8144 12.8137 19.2181
ACO / PID (ITAE) 5.001 12.9500 19.2181
ACO / PID (IAE) 5.8144 13.654 19.2181
ACO / PID (ISE) 5.766 13.654 19.2181
Ziggeler-Nicols PID 40 0.1 0.01

Fig urev13 cardiovascular System disturbed by Brain waves of Pre-sleep

and controlled by ACO/PID

Case with brain waves BETA of Arousal is demonstrated

like below :
TABLE IX. PARAMETERS OF THE DISTURBANCE
GENERATOR
Wave Mean
Variance Frequancy
Form value: Fig ure 16 Comparison cardiovascular System disturbed and controlled
Value Gaussian 0 5 96 Hz by ACO/PID and Ziggeler-Nicols PID

ACKNOWLEDGMENT
We have presented in this work examples of application
of the optimization method based on artificial ants
colonies (ACO) for the synthesis of an optimal PID
regulator. During the optimization process, possible
solutions are evaluated through a function that involves
the performance criteria calculated on the overall
cardiovascular system response. After simulations on the
human heart, the results show that the ACO can give an
effective search for an optimal PID regulator insensitive to Control Systems Magazine, Vol.18, No.6, pp.26-34, December
1998.
perturbations in the form of torque discharge generated by
[13] Y. M. Randall C. O'Reilly, Computational Exploration in
the brain and which present diseases such as arousal, pre- Cognitive Neuroscience: Understanding The Mind by Simulating
sleeping The PID-ACO controller is the best because it The Brain, 2000.
has satisfactory performances and has good robustness. [14] E. E. Konofagou, J. Luo, D. Saluja, D. Cervantes, J. Coromilas,
and K. Fujikura, “Noninvasive electromechanical wave imaging
Also our job is to regulate the impact of the emotional
and conduction-relevant velocity estimation in vivo,” Ultrasonics,
disturbance on the cardiac system and to regulate it after vol. 50, pp. 208–215, 2010.
its perturbation by a External emotional element. Indeed [15] A.M. Cook, J.G. Simes, “A Simple Heart Model Designed to
the emotions influence the reality by the human being on Demonstrate Biological System Simulation”, IEEE Transactions
on Biomedical Engineering, Vol. Rme-19, No. 2, March 1972.
the cardiac it is for this reason we began by the [16] Y. Wu, P.E. Allaire, G. Tao, D. Olsen,” Modeling, Estimation, and
determination of the frequencies of the perturbations Control of Human Circulatory System With a Left Ventricular
generated by the brain like human diseases then we have Assist Device”, IEEE Transactıons on Control Systems
to determine their frequencies and then we Have applied Technology, Vol. 15, No. 4, July 2007.
[17] S.L.M. Abad, N.J. Dabanloo, M.R. Mohagheghi, “Different
these disturbing frequencies to the human heart which Approaches for linear and non-linear ECG Generation”,
causes a dysfunction in human cardiovascular International Conference on Biomedical Engineering and
functionality. The PID-ACO controller was used to Informatics, 2008.
correct the disturbed signal and then to have a cardiac [18] E.H. Maslen, G.B. Bearnson, P.E. Allaire, R.D. Flack, M. Baloh,
E. Hilton, M.D. Noh, D.B. Olsen, P.S. Khanwilkar,J.D. Long,
signal emitting the heart signal in the normal state. As a “Feedback Control Applications in Artificial Hearts”, IEEE
result, our model can be used for medical purposes in Control Systems Magazine, Vol.18, No.6, pp.26-34, December
order to deal with the many cardiovascular problems that 1998.
rely on the sudden disruption of normal human heart [19] W. V. Shi and M. C. Zhou, “Recent advances of sensors for
pacemakers,” in Proc. IEEE Int. Conf. Network., Sens. Control,
functionality. Delft, The Netherlands, 2011, pp. 520–525.
REFERENCES [20] J. Keener and J. Sneyd, Mathematical Physiology. New York:
SpringerVerlag, 1998.
[1] The health statistics is the association’s 2015 Heart Disease and [21] Sapna Katiyar, Avneesh Mittal, Abdul Quaiyum Ansari, T.K.
Stroke Statistics Update, which is compiled annually by the Saxena “Ant Colony Algorithm Based Adaptive PID Temperature
American Heart Association, the Centers for Disease Control and Controller” January 2011, 7th Int. Conf. on Trends in Industrial
Prevention, the National Institutes of Health and other government Measurements and Automation (TIMA 2011), At CSIR, Chennai
sources. [22] Dorigo M., Maniezzo V. and Colorni A., (1996) Ant System:
[2] Fatimah Ibrahim, Noor Azuan Abu Osman, Juliana Usman and Optimization by a colony of cooperating agents. IEEE
Nahrizul Adib Kadri, Performance Evaluation of Coifman Transactions on Systems, Man, and Cybernetics.
Wavelet for ECG Signal Denoising, IFMBE Proceedings, vol 15, [23] Huseyin Atakan Varol, Zafer Bingul (2004) A New PID Tuning
pp 419-422,2007. Technique Using Ant Algorithm, Proceeding of the 2004 American
[3] M.Aabid, A.Elakkary, N.Sefiani “Stabilization of Human Heart Control Conference, Boston, Massachusetts
using PID controller” Journal of Theoretical and Applied [24] M. Guntsch and M. Middendorf. Applying Population Based ACO
Information Technology,Vol 91.NO.1, 15 Octobre 2016 to Dynamic Optimization Problems. In Third International
[4] Learning dependencies among fetal heart rate features using Workshop ANTS, pages 111–122. Springer
Bayesian networks Shishir Dash, J. Gerald Quirk, and Petar M. [25] M. Guntsch and M. Middendorf. A Population Based Approach
Djuri´c pp6204-6207, 1 September, 2012 for ACO. In 2nd European Workshop on Evolutionary
[5] G. Macones, G. Hankins, C. Spong, J. Hauth, and T. Moore, “The Computation in Combinatorial Optimization, pages 72–81.
2008 National Institute of Child Health and Human Development Springer Verlag, LNCS 2279, 2002.
workshop report on electronic fetal monitoring: Update on [26] M. Guntsch, M. Middendorf, and H. Schmeck. An Ant Colony
definitions, interpretation,and research guidelines,” Journal of Optimization Approach to Dynamic TSP. In L. Spector, E. D.
Obstetric, Gynecologic, Goodman, A. Wu, W. B. Langdon, H.-M. Voigt,M. Gen, S. Sen,
[6] Yalcinkaya, F. ; Dept. of Electr. & Electron. Eng., Kirikkale M. Dorigo, S. Pezeshk, M. H. Garzon, and E. Burke, editors,
Univ., Krkkale, Turkey ; Kizilkaplan, E. ; Erabas, A. Proceedings of the Genetic and Evolutionary Computation
Mathematical modelling of human heart as a Conference (GECCO-2001), pages 860–867, San Francisco,
hydroelectromechanical system Published in: Electrical and California, USA, 7-11 2001. Morgan Kaufmann.
Electronics Engineering (ELECO), 2013 8th International [27] D. Merkle and M. Middendorf. Studies on the Dynamics of ACO
Conference on Date of Conference:28-30 Nov. 2013 Algorithms. In the Genetic and Evolutionary Computation
[7] World Health Organization, Cardiovascular diseases (CVDs) Fact Conference, pages 105–112.(GECCO), New York, 2002.
sheet N°317 Updated January 2015 [28] D. Merkle and M. Middendorf. ACO with the Relative Pheromone
[8] Y. M. Randall C. O'Reilly, Computational Exploration in Evaluation Method. In 3rd European Workshop on Scheduling and
Cognitive Neuroscience: Understanding The Mind by Simulating Timetabling and 3rd European Workshop on Evolutionary Methods
The Brain, 2000. for AI Planning, pages 325–333. LNCS2279, 2002.
[9] Sri. Biswarup Neogi, Dr. Achintya Das “Mathematical Modeling
and Simulating of Cardiovascular System Transfer Function with
Deterministic Approach of Heart Rhythm Signal Equation”,
International Conference on Systemics, Cybernetics and
Informatics, 2008 (ICSCI-2008) Vol.1of 2 Page-604
[10] Ader, R., Felten, D., & Cohen, N. (1990). Interactions between the
brain and the immune system.Annual Review of Pharmacology
and Toxicology30: 561–602.
[11] Wei-Ning Lee, Kana Fujikura, and Elisa E. Konofagou
,”Electromechanical Wave Imaging of Normal and Ischemic
Hearts In Vivo Jean Provost”pp 625-629,2010
[12] E.H. Maslen, G.B. Bearnson, P.E. Allaire, R.D. Flack, M. Baloh,
E. Hilton, M.D. Noh, D.B. Olsen, P.S. Khanwilkar,J.D. Long,
“Feedback Control Applications in Artificial Hearts”, IEEE