Empirical Study of an Unconstrained Modified Particle Swarm Optimization
This document has been relocated to http://scholarsmine.mst.edu/ele_comeng_facwork/1010
There were 1 downloads as of 27 Jun 2016.
In this paper, an unconstrained modified particle swarm optimization (UMPSO) algorithm is introduced and studied empirically. Four well known benchmark functions, with asymmetric initial position values, are used as testing functions for the UMPSO algorithm. The UMPSO is a variation of the canonical PSO in which the velocity and position is unconstrained, an additional strategic component is added, and the social component term has been modified. The strategy component is used instead of varying parameters or mutation to enhance diversity in the swarm during the search. The UMPSO algorithm is then compared to results obtained from the constrained canonical PSO (CPSO) and the unconstrained canonical PSO (UPSO). The results show that UMPSO algorithm with no maximum velocity and position, and no minimum velocity and position value that performs better than the CPSO and the UPSO for the Sphere, Rosenbrock, Rastrigrin, and Griewank benchmark functions.