A behavioral-based approach to Particle Swarm Optimization

When optimization algorithms are applied to non convex functions, they are generally affected by the problem of local minima, i.e. optimal solutions that do not correspond to the global minimum of the cost function. Particle Swarm Optimization algorithms are no exception. In order to overcome this problem and to speed up the convergence of the algorithm, in this paper a novel PSO algorithm is proposed. The presented algorithm relies on the idea that the particles exploring the search space can be divided in subgroups, each of which with a peculiar behavior, such as to enlarge the explored area while refining the actual solution. In order to prove the effectiveness of the proposed algorithm, optimization benchmark functions known from the literature have been used in Matlab simulations and the results have been compared with analogous results gathered by using the standard PSO algorithm.