Acoustic imaging for the monitoring of industrial infrastructures remains a common feature to address numerous issues, such as diagnosing faults in rotating machinery or detecting gas leaks. Simultaneously, autonomous vehicles have emerged as a strong tool to automate various maintenance and diagnostic tasks, incorporating increasingly diverse sensors, including microphone arrays. The objective of this work is to present a new control strategy for an autonomous vehicle equipped with a microphone array, aimed at mapping its environment completely blindly. The system must be robust to background noise and to a wide range of sound levels from sources, in order to be effective in an industrial setting. The control strategy proposed in this preliminary work meets these expectations and allows for mapping a large number of sources in a noisy environment, with a wide range of sources sound levels. Its versatile approach allows adapting the localization algorithm used between MUSIC and constrained-MUSIC depending on the situation and acoustic environment, leveraging the information the robot will have collected during its measurement. Its performances are evaluated in a simulation study.
Read full abstract