Passive acoustic sensing with arrays has applications in many fields, including atmospheric monitoring of low frequency sounds (i.e., infrasound). Beamforming of array signals to gain spatial information about the signal is common, but the performance is often degraded due to limited resources (e.g., number of sensors, array size). Such sparse arrays create ambiguities due to reduced resolution and spatial aliasing. While previous work has focused on either maximizing array resolution or minimizing spatial aliasing, the current study demonstrates how evolutionary algorithms can be utilized to identify array configurations that optimize for both properties. The non-dominated sorting genetic algorithm II (NSGA-II) was used with the beamwidth and maximum sidelobe level as the fitness functions to iteratively identify a group of optimized synthesized array configurations. This group is termed a Pareto-front and is optimized such that one fitness function cannot be improved without a decrease in the other. These optimized solutions were studied for a single frequency (8 Hz) and a multi-frequency (3 to 20 Hz) signal using either a 36-element or 9-element array with a 60 m aperture. The performance of the synthesized arrays was compared against established array configurations (baseline) with most of the Pareto-front solutions outperforming these baseline configurations. The largest improvements to array performance using the synthesized configurations were with fewer array elements and the multi-frequency signal.
Petrin et al. (Fri,) studied this question.