Maximum likelihood DOA estimation by real-valued genetic algorithm

The problem of obtaining accurate direction of arrival (DOA) estimation of narrow-band sources lying in the far field of the array is one of the central problems in radar, sonar and seismology. In this paper a real-valued genetic algorithm is used to obtain the global optimal solution of the maximum likelihood (ML) DOA estimation. It overcomes the local optima problem existing in some ML DOA estimation algorithms, and improves the estimation accuracy. The proposed real-valued genetic algorithm is composed of real-valued crossover and mutation operators constructed with the information of real number field and objective function. It is an ideal method for searching for the global solution of non-linear real variable functions. Simulation results of noncoherent and coherent sources DOA estimation show that the proposed algorithm is better in accuracy over some conventional DOA estimation algorithms.