1: \begin{abstract}

2: Particle swarm optimization (PSO)

3: is a widely used 

4: nature-inspired meta-heuristic for solving

5: continuous optimization problems.

6: However, when running the PSO algorithm, one encounters the 

7: phenomenon of so-called

8: stagnation, that means in our context, the whole swarm starts to converge

9: to a solution that is not (even a local) optimum.

10: The goal of this work is to point out

11: possible reasons why the swarm stagnates at these

12: non-optimal points.

13: To achieve our results, we use the newly

14: defined \emph{potential} of a swarm.

15: The total potential has a portion for every dimension of

16: the search space, and it drops when

17: the swarm approaches the point of convergence.

18: As it turns out experimentally, the swarm is very likely to come

19: sometimes into ``unbalanced'' states, i.\,e., almost all

20: potential belongs to one axis. Therefore, the swarm becomes

21: blind for improvements still possible in any other direction.

22: Finally, we show how in the light of the potential and these

23: observations, a slightly adapted PSO rebalances

24: the potential and therefore increases the quality of the

25: solution. Note that this is an extended version of \cite{SWb:13}.

26: \end{abstract}

27: