Bayesian DOA estimator based on modified ant colony optimization

Bayesian maximum a posterior probability density DOA estimator (BM DOA estimator) is known to be the best estimator in DOA estimation for narrow band sources. However, the exponentially increasing computation burden of the BM estimator, due to multidimensional grid search and integrates, makes it very difficult to use the BM estimator in real-time systems. In this paper, a computation feasible ant colony optimization method (ACO) is applied to lighten the computation burden. In addition, in order to overcome the drawbacks of ACO, such as low convergence speed and being easily trapped in local optimum, chaos initialization and local search are integrated into the classic ACO method, to form a novel method named MACO. Based on MACO, a novel BM DOA estimator named BM_MACO with even lower computational complexity is proposed. It is shown via simulations that both methods could keep the good performance of the original BM DOA estimator and reduce the computation evidently. Due to the initialization via chaotic sequences and local search in the optimization procedure, BM_MACO method reduces the sensitivity of parameters, and thus outperforms the BM_ACO for its higher precision and less computation.