Optimization of array geometry for direction-of-arrival estimation using a priori information

This paper focuses on the estimation of the direction-of-arrival (DOA) of signals impinging on a sensor array. A novel method of array geometry optimization is presented that improves the DOA estimation performance compared to the standard uniform linear array (ULA) with half wavelength element spacing. Typically, array optimization only affects the beam pattern of a specific steering direction. In this work, the proposed objective function incorporates, on the one hand, a priori knowledge about the signal’s DOA in terms of a probability density function. By this means, the array can be adjusted to external conditions. On the other hand, a modified beam pattern expression that is valid for all possible signal directions is taken into account. By controlling the side lobe level and the beam width of this new function, DOA ambiguities, which lead to large DOA estimation errors, can be avoided. In addition, the DOA fine error variance is minimized. Using a globally convergent evolution strategy, the geometry optimization provides array geometries that significantly outperform the standard ULA with respect to DOA estimation performance. To show the quality of the algorithm, four optimum geometries are presented. Their DOA mean squared error is evaluated using the well known deterministic Maximum Likelihood estimator and compared to the standard ULA and theoretical lower bounds.