You should be aware - just look at the discussions on RG - that there is also a rather strong criticism against PSO, and other metaheuristics for that matter - and that for several reasons, among those being that (a) there is very little theory behind it and (b) hence a very weak guarantee on the solutions obtained. A fairly large group of scientists here at RG are critical regarding the outcome of using them - in fact there is very little that you can guarantee! In structural optimization - where in fact I have written a few papers - there is quite often a structure (sorry for the pun) in the problem itself that makes it possible to provide much more efficient and convergent methodologies, and I would suggest that you start from those.

If ACO means ant colony something then NO, that's not a good methodology either. It has very little theory behind it, no natural stopping criteria, and an extremely lousy chance to solve large-scale problems in optimization - where there tend to be rather strong structural elements in the model which can be utilized in design of the optimization methodolgy. In my reading on metaheuristics I have seen very little focus on utilizing the structure of the original problem statement, and everyone who has a good upbringing in mathematical optimization knows that finding a proper structure of the problem is one of the most keys to success on a large-scale problem. Learn mathematical programming - before it's too late. :-)

Here are two papers I would recommend in this context:

Article A survey of non-gradient optimization methods in structural ...

The first one discusses how there's not enough benchmarking and an over-reliance on metaheurisitics in the literature.

Article Performance evaluation of metaheuristic search techniques in...

The second one is a pretty extensive benchmark, albeit with some methodological limitations. (Only meta-heuristics were tested, and robustness was not considered.) The PSO is among the better algorithms, but ACO is not.

Finally, I also test two structural problems in my PhD, including one direct search (DIRECT) and one model-based method (RBFOpt).

Thesis Efficient, Visual, and Interactive Architectural Design Opti...

Btw, I agree with Michael Patriksson that, in principle, one should always strive to develop algorithms that exploit a problem's specific structure. But in practice, there's typically no time for this development, which is why I think it's worthwhile to try and come up with somewhat generalizable recommendations.

Personally, I find that in practice it's often helpful to try a couple of different algorithms (including, but not only, metaheuristics) and to see what works well for a particular problem. In structures, one approach that seems to work well is to first run a global algorithm (RBFOpt), followed by a more local one (SBPLEX).

I can accept the "there's no time" principle in non-academic business, but never in academia. A personal viewpoint, that perhaps not everyone else agrees on, is that the endless flow of new metaheuristics is, to put it simply, quite boring - and I want my academic and practical work to give me some satisfaction back. Truth is, PSO will never tickle me.

FROM MY PERSPECTIVE, PSO OR ACO PROVIDE PROBABILISTIC RESULTS ONLY. MAY BE ATTEMPTED WITH SOME PROVEN ALGORITHM TO STRUCTURAL OPTIMIZATION.

If ACO refers to the Algorithms, Combinatorics, and Optimization Program at Carnegie Mellon University, then THAT's something else - that's a scientific program, in contrast to PSO. :-)