Question About Multiobjective Genetic Algorithm Optimisation

Hi,

Firstly i want to explain what i am trying to do:

I am trying to optimize a cellular network by using multiobjective genetic algorithm.

I want to optimise number of base stations , coverage and multi coverage.

my population consist of possible base station locations in X and Y coordinates. which some of them commissioned and others not depending on some constraints.

i want my gene to have lets say 10 possible base station locations with only 5 them of them commissioned by this way i want to keep some junk elements in gene to get a better solution surface.

with three optimisation object i will get a pareto surface.

What i need help on how should i feed the gamultiobj function?

It is easy for coverage and multicoverage because they are just 1-by-1 values. But for basestation number and positions i dont know what to do.

Even though i read the manuals for the gamultiobj manual i cant get my head around this problem. Any help/tips will be great help for me.

[david]

hello,where you finally able to achieve this? if u were pls paste code and a little explanation please. thank you

Answer 1

[Hari]

Hi,

I understand that you are trying to use a multi-objective genetic algorithm (gamultiobj) in MATLAB to optimize a cellular network by considering three objectives: minimizing the number of commissioned base stations, maximizing coverage, and maximizing multi-coverage. Your challenge is representing the gene such that it holds potential base station locations while also distinguishing which are active (commissioned).

I assume that you already have a list of possible base station locations (X,Y), and for each individual in the population, you want to evolve both the commissioning decision (binary: 0 or 1) and evaluate the resulting network performance.

In order to implement this in gamultiobj, you can follow the below steps:

Step 1: Define the chromosome structure

You can represent the gene as a vector where:

• Each gene corresponds to one base station location.
• The value is a binary (0 or 1) indicating whether that station is commissioned or not.

Step 2: Write a custom fitness function

In this function, for a given binary vector x:

• Count the number of commissioned base stations (objective 1).
• Calculate total coverage based on commissioned stations (objective 2).
• Calculate multi-coverage (areas covered by >1 station) (objective 3).

Your fitness function should return a 1x3 vector with values for all objectives.

Step 3: Set up the gamultiobj options

Set:

• 'PopulationType','bitstring' to indicate binary genes.
• Use 'UseParallel',true if coverage computations are intensive.

Step 4: Configure the call to gamultiobj

Set the number of variables to the number of candidate base station locations.

Step 5: Post-process results

The output will be a set of Pareto-optimal binary vectors, each representing a trade-off between number of stations, coverage, and multi-coverage. You can extract coordinates using the known list of locations.

Refer to the documentation of "gamultiobj" function for additional options:

https://www.mathworks.com/help/gads/gamultiobj.html

Refer also to "multiobjective optimization" workflows:

https://www.mathworks.com/help/gads/multiobjective-optimization.html

Hope this helps!