Identification of nonlinear systems using particle swarm optimization technique

System identification in noisy environment has been a matter of concern for researchers in many disciplines of science and engineering. In the past the least mean square algorithm (LMS), genetic algorithm (GA) etc. have been employed for developing a parallel model. During training by LMS algorithm the weights rattle around and does not converge to optimal solution. This gives rise to poor performance of the model. Although GA always ensures the convergence of the weights to the global optimum but it suffers from slower convergence rate. To alleviate the problem we propose a novel Particle Swarm Optimization (PSO) technique for identifying nonlinear systems. The PSO is also a population based derivative free optimization technique like GA, and hence ascertains the convergence of the model parameters to the global optimum, there by yielding the same performance as provided by GA but with a faster speed. Comprehensive computer simulations validate that the PSO based identification is a better candidate even under noisy condition both in terms of convergence speed as well as number of input samples used.