# Particle Swarm Optimization (PSO) Sample Code using Java

UPDATE November 17, 2014: The full source code can now be downloaded from github: https://github.com/therealmanalu/pso-example-java

UPDATE May 03, 2012: The MegaUpload was long gone. If you still need the code, please contact me directly on my email at gandhi.mtm [at] gmail [dot] com. Thank you.

UPDATE April 23, 2011: The sample code can be downloaded from here. I hope the code will be useful for you to understand the topic more clearly.

Yes, I’m still coding and I’m proud of it :)

This post assume that the reader has already known about Particle Swarm Optimization (PSO) method, and hence I wouldn’t spare a space to explain about it. But if you’d like to ask the method’s component on the code I provide below then I’ll be gladly explain it in greater length. [This post is written by Gandhi Manalu – gandhim.wordpress.com]

In this post, I’ll describe and provide a sample code of PSO method for solving a very simple function optimization. Let say that the function to be minimized is as follow:

$f(x,y) = (2.8125-x+xy^4)^2+(2.25-x+xy^2)^2+(1.5-x+xy)^2$

With the following constraints: $1<=x<=4; -1<=y<=1$

In order to solve this problem using PSO, we’ll need these classes:

• Position: to represent Position part of the particle
• Velocity: to represent Velocity part of the particle
• Particle: the particle itself [This post is written by Gandhi Manalu – gandhim.wordpress.com]
• SimplePSO: the main control of the program
• PSOConstants: an interface to define parameters used in the PSO

Since we’re going to solve two-variable function optimization, we’ll need to provide two-dimensional position and velocity.

For the Position we have:

package org.gandhim.pso;

/**
*
* @author gandhim
*/
public class Position {
private double x;
private double y;

public Position(double x, double y) {
this.x = x;
this.y = y;
}

public double getX() {
return x;
}

public void setX(double x) {
this.x = x;
}

public double getY() {
return y;
}

public void setY(double y) {
this.y = y;
}
}

For the Velocity we have:

package org.gandhim.pso;

/**
*
* @author gandhim
*/
public class Velocity {
private double x;
private double y;

public Velocity(double x, double y) {
this.x = x;
this.y = y;
}

public double getX() {
return x;
}

public void setX(double x) {
this.x = x;
}

public double getY() {
return y;
}

public void setY(double y) {
this.y = y;
}
}

And the Particle is as follow:[This post is written by Gandhi Manalu – gandhim.wordpress.com]

package org.gandhim.pso;

/**
*
* @author gandhim
*/
public class Particle {
private Position location;
private Velocity velocity;
private double fitness;

public double getFitness() {
calculateFitness();
return fitness;
}

public void calculateFitness() {
double x = this.location.getX();
double y = this.location.getY();

fitness = Math.pow((2.8125 - x + x * Math.pow(y, 4)), 2) +
Math.pow((2.25 - x + x * Math.pow(y, 2)), 2) +
Math.pow((1.5 - x + x * y), 2);
}

public Position getLocation() {
return location;
}

public void setLocation(Position location) {
this.location = location;
}

public Velocity getVelocity() {
return velocity;
}

public void setVelocity(Velocity velocity) {
this.velocity = velocity;
}
}

Pay attention to the calculateFitness() method, it is where we put the function evaluation.

Now, we’re ready for the main process of the PSO. In this class, we’ll need several methods:

• initializeSwarm() – to initialize the swarm used in the method
• execute() – the main part of the process[This post is written by Gandhi Manalu – gandhim.wordpress.com]

Here’s the code (partial code):

private void initializeSwarm() {
Particle p;
Random generator = new Random();

for (int i = 0; i < SWARM_SIZE; i++) {
p = new Particle();
double posX = generator.nextDouble() * 3.0 + 1.0;
double posY = generator.nextDouble() * 2.0 - 1.0;
p.setLocation(new Position(posX, posY));

double velX = generator.nextDouble() * 2.0 - 1.0;
double velY = generator.nextDouble() * 2.0 - 1.0;
p.setVelocity(new Velocity(velX, velY));

}
}

public void execute() {
Random generator = new Random();
initializeSwarm();

evolutionaryStateEstimation();

int t = 0;
double w;

while (t < MAX_ITERATION) {
// calculate corresponding f(i,t) corresponding to location x(i,t)
calculateAllFitness();

// update pBest
if (t == 0) {
for (int i = 0; i < SWARM_SIZE; i++) {
pBest[i] = fitnessList[i];
}
} else {
for (int i = 0; i < SWARM_SIZE; i++) {
if (fitnessList[i] < pBest[i]) {
pBest[i] = fitnessList[i];
pBestLoc.set(i, swarm.get(i).getLocation());
}
}
}

int bestIndex = getBestParticle();
// update gBest
if (t == 0 || fitnessList[bestIndex] < gBest) {
gBest = fitnessList[bestIndex];
gBestLoc = swarm.get(bestIndex).getLocation();
}

w = W_UP - (((double) t) / MAX_ITERATION) * (W_UP - W_LO);

for (int i = 0; i < SWARM_SIZE; i++) {
// update particle Velocity
double r1 = generator.nextDouble();
double r2 = generator.nextDouble();
double lx = swarm.get(i).getLocation().getX();
double ly = swarm.get(i).getLocation().getY();
double vx = swarm.get(i).getVelocity().getX();
double vy = swarm.get(i).getVelocity().getY();
double pBestX = pBestLoc.get(i).getX();
double pBestY = pBestLoc.get(i).getY();
double gBestX = gBestLoc.getX();
double gBestY = gBestLoc.getY();

double newVelX = (w * vx) + (r1 * C1) * (pBestX - lx) + (r2 * C2) * (gBestX - lx);
double newVelY = (w * vy) + (r1 * C1) * (pBestY - ly) + (r2 * C2) * (gBestY - ly);
swarm.get(i).setVelocity(new Velocity(newVelX, newVelY));

// update particle Location
double newPosX = lx + newVelX;
double newPosY = ly + newVelY;
swarm.get(i).setLocation(new Position(newPosX, newPosY));
}

t++;
}
}

And the last is the interface for storing the constants:

package org.gandhim.pso;

/**
*
* @author gandhim
*/
public interface PSOConstants {
int SWARM_SIZE = 30;
int DIMENSION = 2;
int MAX_ITERATION = 300;
double C1 = 2.0;
double C2 = 2.0;
double W_UP = 1.0;
double W_LO = 0.0;
}

I made the interface just for the sake of flexibility.

A sample of running the program is as follow:

As we can see from the result, the program found the solution of the problem for (x=3.0 and y=0.5). You might have noticed that the program is not optimized yet, for example it could have been stopped when it already found the solution.[This post is written by Gandhi Manalu – gandhim.wordpress.com]

Actually, there are other PSO’s components that have not been implemented in this program, for example constraints handling. But since the problem we solved here is a very simple one it doesn’t really need the constraint handling.

I hope that this post is useful for you.

## 50 thoughts on “Particle Swarm Optimization (PSO) Sample Code using Java”

1. dnaveena says:

hai gandhi
I m naveena lakshmi from Malaysia.I m doing my Ph.d on pso for software testing.
I m new for this PSO so i want to know how to find best solution for my problem .
If u send ur mail id for my mail id i can send my problem through attach files.
anapa.naveena@gmail.com this is my id.
Thank u

2. simonustc says:

it is a good example to study, could you send your entire code to me ?
my email is wzj008@gmail.com. many thanks to you.

3. ilmig says:

hi! it’s possible to have all source code?
fabio.

4. biganer2008 says:

Hello,could you send me a copy of code by e-mail because the program are short of some code. My e-mail is menghua523@163.com . Thank you very much!

5. guilhermeramos says:

Hello!

Would you mind sending me the entire code please? I’m developing an article and it would be very helpfull to me. Thanks in advance!
my e-mail is: guilhermejf1987@gmail.com

6. hi….
i need to calculate optimize cost…..
can u plz help me

7. prashantku says:

HI GANDHI,

REALLY A INTERESTING POST.
I WANTED TO KNOW CAN THIS BE USED FOR THE OPTIMIZATION OF NEURAL NETWORK…CAN U HELP PLEASE.
MY EMAILID PRASHANT.TALKIN@GMAIL.COM

MANY THANKS
PRASHANT

8. satcse says:

Hi Gandhi,I Found your work to be really great and appreciable.I myself is working on PSO right now.SO,if you send the entire code,it would be of great help to me.

9. Hello!
I’m working on a project with pso in python. pls would u mind sending me entire code? so i can use it as a guide. i really would be grateful. Thanks!

10. hi,
i am doing final year project in PSO for job scheduling. i want this full coding. plz send this code to my mail id. my mail id is raviabirami23@gmail.com.

11. Hey,
Your code is truly Fantastic. Could you please send it to my email id archana.shiva@gmail.com. I am sure it would be of real help to my studies.

Thank you very much….

13. Hello~~
It is a good example to study, could you send your entire code to me ?
my email is coman192001@gmail.com. many thanks to you.

14. Would you mind sending me the entire code please? it would be very useful for my thesis about PSO on Nanobot positioning… can u help me please!!! Thanks
my email: valentina.mannara@gmail.com

15. hi …!
i’am happy if you can send me the entire code !I think this code would be helpfull to me ! Thanks a lot !
My e-mail :ngongocquang90@gmail.com

16. Hello!
Great post my friend, I’ve just begun learning about PSO and it has been great help for me. Unfortunately the link you gave for source code doesn’t work for me. Would you be so kind and send it to me?
My email is: eldae25@gmail.com
eldae

• No need anymore, I figured it out myself ;)

17. cleoruth says:

hello!

Good day! Can you send me the copy of the whole source code of the PSO -sample. Through this email Sandy_1663@yahoo.com. This will greatly help our thesis. Thank you so much. And God Bless

18. Hi

Thanks for the nice write up

19. Hi Gandhi..I am a student, actually I doing a benchmark beetween differents no linear functions, but I am mathematical, so the development is something new for me, but is very interesting know how we can levarage this type of knowledge and use this computational approach inside of my school. Therefore I would like to have all the code that you did to work whit this in my project. If you can send me it, I am aprecciate. jorge.valderrama.romero@alumnos.upm.es

20. Dear all, I’m sorry for this very late reply. If you need the program, you could just contact me on my email at gandhi.mtm [at] gmail [dot] com. Thank you.

21. liaoleo38 says:

Dear Gandhi:
Would you mind sending me a copy of the code?
My email i:dlnx38@hotmail.com

22. arifa509 says:

hi,
i am doing final year project in PSO for image retrieval. i want this full coding. plz send this code to my mail id. my mail id is arifa.tehseen@gmail.com.

• Hi just send to my email written on top of this article.

23. I find the article interesting aand would be happy if Gandhi sends me the full code. Thanks.

24. Pretty much all of the response you have is people asking for the full code. Why not just post it here or make a link to dropbox?

25. Good day,
Link to the source code doesn’t work. I would ask if you could please email me the code @ onlinelogin@webmail.co.za but if you fixed the link it would probably save yourself a lot of time and effort.

Thanks for the great post! It has helped me understand the layout and structure of a PSO much better.

26. l104048 says:

can you please send me the code,…. i will be greatful to you

• l104048 says:
27. hai this is kumar i am doing my m.tech project on pso for the project your code is very helpful to me so can u plz mail me your hole code to me. my mai id is kkumarbabu07@gmail.com