Comparative Performance Investigations of Stochastic and Genetic Algorithms Under Fast Dynamically Changing Environment in Smart Antennas

In a mobile communication systems, the number of observation data (snapshots) used for covariance matrix estimation can be insufficient, which often occurs due to fast dynamically changing environment or signal characteristics are rapidly changing. In these situations, the performance of the standard adaptive algorithms such as LMS are known to degrade substantially. In this paper, we propose the use of a Genetic Algorithm (GA) to perform the adaptation control of the system parameters under dynamically changing environments The GA-based beamformer has nearly optimal interference cancellation under dynamic conditions, and makes the output SINR consistently close to the optimal one regardless of the number of snapshot used. Other advantages of the GA is its simplicity and fast convergence provided that the parameters are appropriately chosen, which makes it a practical algorithm for beamforming in smart antenna. Simulation results validate substantial performance improvements relative to other standard adaptive algorithms. Although, the use of GA is not new in smart antenna technology, the performance evaluation of the genetic optimization under fast dynamically changing environment has not been investigated to the best of my knowledge and it is of great practical significance.