Is NS3 the correct simulator for me?

[Soumya Banerjee]

I am currently investigating  a novel methodology for  distributed un-supervised clustering (using GA) in wireless sensor networks.

When I began my my research (ME thesis) 6 months back I initially started working with NS3. But on discovering that till then the energy module is not integrated in lr-wpan and also due to the relative complexity of a NS3 simulation, I shifted to a high level model using matlab.

I have finished my initial proposal and my Thesis guide has approved of it . Now I am thinking on how to do a more  detailed analysis of my methodology and i realized for that I need to use a network simulator.  

My interest primarily lies in the un-supervised learning part of the problem not in the network application of it. Considering the details covered in a NS3 simulation I was wondering if there is some other comparatively less detailed simulator more applicable for me ?

[Tommaso Pecorella]

Shortly put: you can use whatever simulator you want, even Matlab.

The higher level the simulator, the less you will be able to consider the side-effects of your system.

E.g., you may "think" that a node is able to know the energy of a neighbour node. With ns-3 you'll have to deal with the fact that info have to be embedded in a packet, packets have to be sent and so on.
Still considering the energy, you already noted that Matlab can't really evaluate the energy consumed by the nodes, and so far ns-3 can't either (because the MAC is not using the SLEEP state).

What is the best simulator for you ? I don't know. I could say that you have to work at the lowest level you can, i.e., with a network simulator like ns-3 or omnet++. Eventually, you should enhance and expand a simulator if it doesn't fit your needs. The complexity and the difficulty in using the tool is a price to pay to have realistic results. The easier and the less complex the tool, the less precise and useful will be the results.

Now, the choice is yours.

Cheers,

T.

[Soumya Banerjee]

Thanks for your prompt reply.
The side-effects is what i am concerned about. I will discuss my dilemma with my supervisor tomorrow. I was looking for names of simulation software that is not as low level as NS3 . Can you possibly help me there?

[Tommaso Pecorella]

Hi,

to be honest... no, I can't help you. The alternatives coming to my mind (e.g., Omnet++) are different, but not simpler. 

As a side note, it's more or less like waling into a Range Rover auto dealer and saying "I'm a bit scared by the price of the car. Can you suggest some other brand with similar product but cheaper?"

Cheers,

T.

[Soumya Banerjee]

I guess that is a fair analogy. Thanks for your input though.