WAH-i: Optimising Microphone Array Geometry for Customised Localisation Accuracy

Abstract

Accurate spatial localisation of free-flying echolocating bats is foundational for resolving fine-scale flight and echolocation behaviour, prey interception, and spatial decision-making in natural environments. Acoustic localisation using microphone arrays is widely employed for this purpose, yet array geometries in field studies are typically chosen heuristically rather than systematically optimised. As portable multichannel ultrasonic recording systems become increasingly accessible, principled design guidelines are needed to ensure reliable localisation performance under practical deployment requirements. I introduce an iterative array optimisation algorithm that designs microphone geometries by maximising localisation reliability within a predefined three-dimensional field of interest. The method evaluates candidate geometries using simulated acoustic emissions and time-difference-of-arrival localisation, quantifying performance as a volumetric pass rate: the proportion of source locations that meet a user-defined accuracy threshold. Microphone positions are iteratively perturbed and accepted based on improvements to this task-level metric, while enforcing practical constraints on array aperture, inter-sensor spacing, and deployability. Across canonical polyhedral geometries, random initialisations, and arrays comprising four to twelve microphones, optimisation consistently produced rapid early gains followed by geometry-dependent performance plateaus under the specified stopping criterion and iteration budget. Under fixed-aperture constraints, increasing the microphone count yielded diminishing returns, and optimised low-order arrays-particularly four-microphone configurations-matched or exceeded the volumetric localisation performance of higher-order arrays with suboptimal geometry. Analysis of optimisation trajectories further revealed that convergence dynamics scale with array order, whereas volumetric performance under the tested conditions is dominated by geometry rather than sensor number. These results demonstrate that array geometry is the primary determinant of volumetric localisation reliability, and that efficient, portable arrays can be systematically designed using optimisation rather than heuristic rules. The proposed framework is broadly applicable to bioacoustic localisation problems beyond echolocating bats, including avian tracking, passive acoustic monitoring, and conservation-oriented sensing, and provides a general approach for designing task-optimised acoustic sensor arrays for a wide range of applications.