In this paper we are going to show that applying the twinkling technique on a naive random search algorithm is frequently more powerful than any algorithm using specific research techniques unless they use information provided by the gradient or the Hessian. In order to illustrate this result we have made the choice of the study of a mechanical system characterized by the non-linear nature of the optimization space. This system is basically an open kinematics chain that represents a robot which has to go through various different trajectories defined by a set of temporally equidistant points. In fact, we are going to show that the genetic algorithm, the simulated annealing algorithm, the particle swarm algorithm, the random search algorithm, need to use comparatively, a huge number of function evaluations in order to perform the same result quality.

This content is only available via PDF.
You do not currently have access to this content.