ESTIMATION OF DIRECTION OF ARRIVAL FOR ADAPTIVE BEAMFORMING

Abstract

Estimation of Direction of Arrival (DOA) of sources is a basic component of adaptive beamforming. The objective is to steer the main beam in the desired direction, while nulls are allocated in the direction of unwanted signals. It is an area of research which has got direct applications in radar, sonar, seismic exploration, mobile communication etc. Besides DOA estimation, amplitude, frequency and range are the other important parameters that need to be estimated. This dissertation is a contribution towards the above mentioned areas. These contributions are mainly divided into two parts. In first part, our contribution is to develop efficient schemes to jointly estimate the amplitude and DOA of the far field sources. Specifically, we have targeted the joint estimation of amplitude and 2-D DOA (elevation & azimuth angles) of far field sources impinging on 1-L and 2-L shape arrays. In the second part, we deal with near field sources impinging on uniform linear and centro-symmetric cross shape arrays. The basic tool applied to estimate these parameters are meta-heuristic or nature inspired algorithms, which are tailored and trained to solve the problem in hand. These techniques include Genetic algorithm, Particle swarm optimization, Differential evolution and Simulated Annealing. In order to improve the performance, the global search optimizers (meta-heuristic techniques) are hybridized with rapid local search optimization methods such as Pattern search, Interior point algorithm and Active set algorithm. We have used two fitness functions for the far field, as well as, for the near field sources. Initially, we have used Mean Square Error (MSE) as a performance evaluation criterion. This fitness function is based on maximum likelihood principle. The second fitness function is multi-objective, which is the combination of MSE and correlation between desired and estimated vectors after normalization. Both of the fitness functions are easy to implement and need a single snapshot to generate the results. They also avoid any ambiguity among the angles that are supplement to each other. The proposed hybrid schemes are compared with the individual responses of these algorithms and also with the traditional classical techniques available in the literature. The comparison parameters are chosen as the estimation accuracy, convergence, robustness against noise, MSE and proximity effects. To get the near optimum statistics, a large number of Monte-Carlo simulations are carried out for each scheme.