Dear Maryam,

the general steps of GA are:

1- determine the parameters that  you think they will affect the considered phenomena

2- generate random sets of input data (population) each set contains values for the selected parameters

3- test each set of the population and determined its error in terms of (SSR)

4- sort the population according to SSR

5- chose survives (sets with lower SSR values), generally, survives are about 10-30% of the population

6- complete the population to its original size by randomly mixing the parameters values of survives to generate new population members

7- repeat steps from 3 to 6 until you satisfied with the error percent (SSR)

you may find the book in the link below useful.

http://www.boente.eti.br/fuzzy/ebook-fuzzy-mitchell.pdf

Also, there is an add-in tool for excel to carry out all GA steps automatically , see the link below

http://www.solver.com/genetic-algorithms-excel?utm_source=bing&utm_medium=cpc&utm_campaign=Tier%201%20-%20Tutorials&utm_term=genetic%20algorithm%20pdf

thank you very much for answering

I understands the steps of the genetic but I'm facing difficulties in the cost optimization of cellular network using GA (the number of base station reduction)

Dear Maryam,

I'm a structural engineer, so I'm not qualified to tell you what is the optimum positions for cellular phone towers, but I'm always interesting in physics, so let me think with you and correct me if I'm wrong.

I assume that you consider signal strength (and may be signal quality) as fitting function, it is logic, when a tower transmits a signal, the energy will propagate in the shape of sphere and the tower in its center, the sphere will expand as the signal propagate. but the energy of the signal is constant, so the intensity of the energy on the surface of that sphere will reversely Proportional with its radius squared( the distance to the tower squared). if there is a minimum accepted value for the intensity of signal energy (signal strength) that leads directly to maximum accepted distance from the tower. of course this analysis is valid if the media around the tower is homogeneous. 

based on previous analysis the optimum solution is to arrange the towers so that the spacing between each tower is exactly the maximum allowable distance. many millions year ago, bees solved this issue using the hexagonal grid. Hence, it is very logic not get any optimization using this model, case closed.

So, you have to change the case to get different solution, on other words, you have to change the boundary conditions and break that homogeneity to get another optimum solution.

Based on that I suggest to consider some zones with different conductivity  in your model to simulate actual obstacles such as high raise buildings or mountains, also can consider Z coordination to simulate actual topography. those factors will make the media non-homogeneous, hence, the optimum solution will not be the hexagonal grade but it will depend on the distribution of the obstacles and the topography of the site.

I hope this helps, again I'm not a specialist in this field, so correct me if I'm wrong.

good luck

your information are so great and helpful

that's exactly what I'm planned to do but i faced difficulties in taking zones with obstacles so in order to keep in the time limit that i have i took the other path (homogeneous)if i finished it i'll try to do on a real environment 

thank you very much

best regards

Dear Maryam,

Thank you for your encouraging reply. Regarding the optimization algorithm , let me suggest the following one:

1- define the boundaries of the study zone 

2- locate some towers randomly in the study zone, number of towers should be more than the required

3- calculate signal strength at mid point between each towers, if it is more than the minimum accepted value move the towers away from each other along the connecting line till signal strength become the minimum accepted value. on other hand  if signal strength is less that the accepted minimum value move the towers toward each other along the connecting line till signal strength become the minimum accepted value.

4- Repeat step 3 between every two towers sequentially till the system settled on the optimum arrangement.

5- number of towers inside the study zone in the minimum required  number, see the attached figure

you can develop this model for non-homogeneous environment by dividing the study zone into sub-zones ( may be using orthogonal grid) and give each sub-zone a different conductivity to simulate obstacles.

for homogeneous environment, you can calculated the signal strength using close form equation, So you just need an excel sheet, but for non-homogeneous environment you may use finite element software something like ANSYS or ABAQUS.this will take a lot of effort, the other alternative is to use the built in functions in MATLAB to carry out the calculations ( you may need to write some code to control the functions)   

Any way I think it is an interesting research point, good luck

thank you very much Sir that is great

I have already done the first three steps so i think i will try to do non-homogeneous part and see what will happen

best regards

Hi

I think you have a problem with  the fitness function. You have to introduce a related criterion with the number of the base stations. Later this criterion will have a direct impact on the evolution of the fitness in the GAs. In other words the objective to minimize or maximize it should be expressed  directly  in the objective function.

Best regards