We present a new methodology for optimal placement of random centers, for approximating certain given data by using smoothing radial basis functions. A new technique is presented to optimize both the number of radial basis centers and its optimal placement using some variants of advanced multiobjective genetic algorithms.