Bat Algorithm: Application to Adaptive Infinite Impulse Response System Identification

The problem of system identification concerns with the design of adaptive infinite impulse response (IIR) system by determining the optimal system parameters of the unknown system on the minimization of error fitness function. The conventional system identification techniques have stability issues and problem of degradation in performance when modeled using a reduced-order system. Hence, a meta-heuristic optimization method is applied to overcome such drawbacks. In this paper, a new meta-heuristic optimization algorithm, called bat algorithm (BA), is utilized for the design of an adaptive IIR system in order to approximate the unknown system. Bat algorithm is inspired from the echolocation behavior of bats combining the advantages of existing optimization techniques. A proper tuning of control parameter has been performed in order to achieve a balance between intensification and diversification phases. The proposed BA method for system identification is free from the problems encountered in conventional techniques. To valuate the performance of the proposed method, mean square error, mean square deviation and computation time are measured. Simulations have been carried out considering four bench-marked IIR systems using the same-order and reduced-order systems. The results of the proposed BA method have been compared to that of the well known optimization methods such as genetic algorithm, particle swarm optimization and cat swarm optimization. The simulation results confirm that the proposed system identification method outperforms the existing system identification methods.