Optimization of array geometry for identifiable high resolution parameter estimation in sensor array signal processing

Abstract

This paper is concerned with the optimization of array geometry using genetic algorithms as the optimization tool. Recent advances in array processing have been focused on developing high resolution algorithms for estimating signal parameters. The problem of optimal design of the array geometry has been neglected and therefore addressed in this paper. An optimal array geometry will correspond to one with the lowest Cramer-Rao bound (CRB) and which gives rise to minimal ambiguities at low SNR. An approach using genetic algorithms (GA) to minimise the CRB, subjected to the ambiguity constraint is proposed and implemented. By utilizing the parallel search capability of the GA, this approach constitutes an efficient design tool for the design of an array of any size and configuration. An alternative using simulated annealing is also proposed. Both approaches are shown to produce optimum array geometries that are superior to the conventional circular array in terms of accuracy and identifiability.